April 29, 2025 • 57 mins
With the median company now using three salary data sources—and four or more for larger organizations—it’s clear that more data is the norm. But more doesn’t always mean better. Are you truly getting the most out of your investment in compensation data? Or just adding complexity without clarity?
In this episode, we break down what an effective compensation benchmarking strategy really looks like—and how to use multiple data sources strategically, not just simultaneously.
Resources:
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
Welcome everybody. Thanks for joining us today. My
(00:02):
name is Ruth Thomas. I'm ChiefEvangelist and SVP of Marketing
here at Payscale. So I help todrive our thought leadership and
oversee our marketing teamsthrough all the work that we do
here at Payscale.
Sarah, over to you.
Speaker 2 (00:18):
Good morning, good afternoon, good evening and good
night. I'm Sarah Hillenmeyer.I've been building data and AI
enabled tools in HR tech,medicine and other industries
for more than fifteen years. AtPayscale, I lead our AI and data
insights teams within ourresearch and development
organisation.
Speaker 1 (00:36):
Thank you, Sarah. Okay, so for the last couple of
years, I think we've run a shortdata thought leadership webinar
series. So we're doing the sameagain this year. Really, we aim
to bring you some key insightsaround how to manage data
considerations for when you arecreating a data strategy. We aim
to bring together thoughtleaders, data experts like
(01:00):
Sarah, practitioners, customers,partners to join us in that
discussion.
This is the first of thoseepisodes of this year's series.
We have another webinar comingup on May, where we will also be
sort of building on the contentthat we've covered today. But in
today's session, we reallywanted to cover these things
(01:20):
here. So whether you're pricingone job or whether you're
pricing thousands of jobs,having consistency in terms of
your approach to using data andhow you use that data to support
the business goals that you wantto do is important. So we're
going to start the session offsharing you with a model that we
think about when we think aboutbuilding a data strategy within
(01:44):
an organization.
And then there are some otherhot topics that often come up in
conversation as we think aboutbuilding a data strategy. So
data transparency, data biasesand data and AI. So we're going
to cover off those three in thissession as well, sharing
thoughts and kind of like wherewe're seeing in practice and
some kind of key innovationsthat we're seeing starting to
(02:07):
happen in the market. So that'sthe aim of today's session. So
let's get started.
I'm going to kick us off interms of thinking about what
makes a good data strategy. Andwe think about it as having
three key components. So we'regoing to walk you through this
Sarah and I. So the first is theinput data, the data that you
(02:28):
potentially purchase fromproviders or you get free from
government sources or from theweb. And Sarah is going to talk
you through about the variousthings you should be thinking
about as you bring, select thatdata.
Then there's the kind of how areyou going to use that data
internally. So I'll run youthrough that, like what are the
considerations you think aboutas you bring that together
internally to support yourcompensation processes. And then
(02:52):
in today's world of paytransparency, the discussion
around what you should or shouldnot communicate in terms of, you
know, the data that you're usingand how you're using it is
becoming more pertinent. Sowe'll talk you through some
guidelines and some kind of thethings that we're seeing
happening across our customerbase. So I think Sarah, you're
going to start us off here withquestions to answer as you
(03:14):
evaluate any data source.
Speaker 2 (03:16):
Great. When you're considering using a data source,
whether it's a free data sourceor something that you're
thinking about purchasing, theseare the questions that I
recommend answering and keepingtrack of. The most important
question is whether the datasource contains pay data about
the jobs of your organization.You want to be able to find
most, let's say 80% or more ofthe roles that you're trying to
(03:40):
price within the data sourcethat you're considering. We
generally recommend that foreach role you're pricing, you
find a job in the data sourcethat's at least a 70% match in
terms of the skills, keyresponsibilities, experience,
and reporting structure.
The next thing to consider isthe repeatability. How easy is
(04:02):
it to price jobs in a consistentway year over year or time
period over time period? Are thechanges that you're seeing in
the data that you're buyingreflective of the market? Or is
it a potentially noisy datasource that's pulling from
different populations atdifferent times and showing you
variation that isn't necessarilywhat you would want to bake into
(04:23):
your merit increases and yourmarket increases for your
employees that year. Next up,this is more of a checkbox from
my perspective, although we cantalk about how to evaluate
methodology and explainability.
But any data source that you buyor use, you should understand
where the numbers are comingfrom. How are they calculated?
(04:45):
If they're calculated, ifthey're not calculated and
they're collected empirically,where did the data come from?
When was it collected? Whathappened to it from that point
of collection to the point thatit's in your hands?
That's an important thing toknow. Even if you don't feel
like you can totally evaluateit, you should know the answer
to that question. Number four ishow old are the data? When were
(05:08):
those data points collected?Have they been aged since the
time of collection?
Aging, if you haven't heard it,is a term that refers to
adjusting the market numbers bysome fixed percentage per month
or per year to account formarket trends that have changed
the pay since the time that thedata were collected. Number five
(05:30):
is biases. We're going to talk alot more about this later, but
you should have a sense of theunderlying biases in the data
sample and how that might impactthe way that you need to use
that data going forward. Andlastly, and Ruth will touch on
this more, as you think throughwhich data sources to use for
(05:50):
particular tasks, think aboutwho else can see the data. Is it
something that your employeescan access?
That might change the way thatyou think about using that data
and it might influence yourcompensation strategy and your
communication strategy aroundthat data versus having
something that you know thatyour employees don't have access
(06:10):
to. Ruth, can you talk usthrough the processes of how to
deal with that data once youhave it?
Speaker 1 (06:18):
Yeah, so you've got all that wonderful data, you've
sourced your data. I guess thenext step is what are you going
to do with it and how are yougoing to use it internally?
Probably the biggest decisionpoint is how you're going to
aggregate multiple sources ifyou're using more than one
source of data. We're not goingto talk today about the
(06:39):
different types of data sourcesthat are available. We will
potentially cover that inanother version or in another
episode of this series.
That's something you'd like usto cover as well, leave some
comments in the Q and A and wecan make sure that we cover that
in more detail. But there arelots of kind of considerations
in terms of aggregation, interms of potentially, what
(07:00):
percentile are you going to usewithin each of the survey
sources that you're using? Areyou going to weight them
differently? So do these sourceshave more importance or less
importance for you? And doesthat change by the different
talent groups that you havewithin the organization?
How are you going to deploy? Howare you weight? Which is more
important for you? Sarahmentioned ageing, so you know as
(07:25):
you bring the different datasources together you're going to
need to think about your agingstrategy so that you're looking
at them all essentially from thesame time frame as much as is
possible. And then you know areyou going to use the same
methodology for aggregationacross different job groups or
different talent groups?
So that's an important questionbecause you may have talent
(07:47):
groups that you know, wherethere is a high demand for
talent or that the labor marketis moving very fast or there is
a hot demand for certain skills.And so therefore, how are you
deciding to use data to thinkabout pricing jobs, to think
about building pay ranges, maybeslightly different for that
group than for another groupwhere they may be sitting in a
(08:09):
more potentially stable talentmarket. And the same when it
comes to thinking about geodifferences as well as like how
are you going to factor thatinto bringing all these data
sources together. So that'sprobably the most important step
here in terms of developing yourinternal processes. But you also
need to think about how you'regoing to handle the data that
(08:30):
you're bringing in relative towhat exists in your organization
today.
So you may have some old rangesaround within the organization.
They may be job based or gradebased. What are you going to do
with this data relative tothose? And then also you really
need to do that comparison,which I know you will do. It's
that comparison of the marketdata with your current
(08:51):
incumbents.
But you may need to sense checkthat a little bit as you refine
the data sources that you'rechoosing and the data sources
that you're aggregating andbringing together. Obviously
many people will use the marketdata to help build pay ranges.
Again, you need to think aboutwhat your strategy is for that
(09:12):
and is that going to beconsistent across different
talent groups. And then today'sage, obviously many of us are
using the data that we get fromthese surveys to post in job
postings in order to comply withpay transparency legislation. So
how are you going to go aboutcalculating the range that
you'll actually list in the jobposting?
(09:34):
Again, not something we're goingto cover in-depth today. We have
a detailed webinar that talksabout that whole process of
selecting job postings. I'll askAmber to include that link in
the follow-up. But it's reallylike I have my market data. I've
made it, I've made somediscerning decisions about the
data I'm bringing into theorganization.
(09:55):
And then thinking about thattheme of consistency, how am I
going to apply that in terms ofhow we will use it within the
organization? I don't know ifthere's anything you wanted to
add there, Sarah. No. No. Okay,cool.
And then the third step in ourmodel, we said, you know, in the
age of pay transparency isthinking about what you're
(10:16):
actually going to communicate.So previously, I mean, I've
worked in compensation for manyyears. This was something that
we did in a back room somewhere.Nobody got to know what we were
doing. And probably the onlyinteraction point, for example,
a line manager might have withthe market data was potentially
through the merit review processor a discussion during the
(10:39):
hiring process.
There is a demand, you know, tomake conversations more two way
and to make conversations moremeaningful about pay for people
to be able to understand whereyou are potentially sourcing
market data. So I'm sure many ofyou on the call have been
through the experience of havingan employee or potentially a
(11:02):
candidate through the hiringprocess come to you and say,
well, this is the data that I'mseeing for the job. Know, in
order to be able to counterthat, you need to be able to
have a conversation about thedifferent types of market data
sources, potentially wherethey're sourcing that data from.
So that might be job boards orGoogle even nowadays, you know,
(11:23):
that's very easy through Googleto get that data for a job
relative to how you're choosingmarket data. And that's going to
force you in a way then to bemore transparent.
And so that's why organizationsare starting to think more
about, you know, how far do wego in terms of sharing data? So,
you know, what is it you'regoing to share? You have various
(11:43):
options. Are you going to sharethe methodology? Are you
actually going to share themarket data ranges?
That's probably one people feel,you know, least likely to share.
And then are you going to tellemployees potentially where they
sit against that range? And thenare you going to explain the
data sources that you've usedand why? That's probably quite
(12:05):
technical conversation for mostof your audience in your
organization. But there isprobably a way that you can tell
that story in terms of, youknow, a more generic level.
It's like we choose market datasources that represent X, Y, Z
and we use that, you know, to dothis. So having a talk track to,
you know, the type of datasources that you're using
without going through a highlytechnical soliloquy about how
(12:29):
you, you know, selected all ofthose processes per like the
stuff that Sarah talked tothrough at the beginning. And
then, just thinking about whatmore do you need to do? So what
regulations apply to us as wethink about communicating pay
and what are our competitorsdoing in terms of how they
communicate about market data aswell. So an interesting
(12:49):
conversation, interested to hearyour thoughts, interested to
hear your questions as we moveinto the Q and A about how
comfortable you feel at thispoint about communicating market
data and what elements of thatyou think that you should be
communicating.
Speaker 2 (13:05):
And we'll talk more about different places to land
on that transparency spectrumand how to migrate from one
section to another if you wantto increase your transparency.
But just to wrap us up here,when we say data strategy or
compensation data strategy,those three pillars are the
things that we're talking about.Which input data are you using?
(13:28):
Does it cover your jobs? Is it arepeatable data set that has
stability in it where the trendsare real?
Can you explain it? Is it fresh?Is the bias something that you
can work with? And again, we'lltalk more about that. And then
how do you process that data togenerate pay decisions for all
of your employees?
(13:49):
And then thirdly, how do youcommunicate that? And obviously,
these three pillars areinterdependent, but we think
it's useful to think about eachof them as separable. You can
think about each of theseseparately, even though the data
choices you make may influenceyour communication strategy or
(14:09):
vice versa. If you're startingout, just start marching down
this line. Pick your data, thinkabout your process, determine
your communications.
Speaker 1 (14:19):
Yeah, so this is a great reference slide for you.
You obviously will get a copy ofthe slides in the follow-up
communications that will go out,but this is probably, you know,
is where we pull together thatmodel for you. A good checklist
slide for you to take back toyour own organisations. Okay, so
we've brought up the topic ofdata transparency. Let's dig
(14:41):
into that a little more.
I talk a lot about paytransparency. You would have to
have been living somewhere verystrange to not know if you're
working in the field ofcompensation and HR that
transparency is a thing. It'sdriving a lot of what we do and
decide around compensationpractice today. And that's
(15:01):
obviously particularly beendriven by legislation. And that
legislation is really aroundpromoting fairness, equity and
trust by ensuring that employeesunderstand how their pay is
determined and how it comparesto pay of their peers.
So that's obviously about openlysharing information and we know
many of you have been on thatjourney. We see that in our
(15:24):
results from our annualcompensation best practice
study. And also through thewebinars that we do, where we
poll you, we're gonna go up onour polls coming up in a minute.
But we know, this is a keyactivity that you've been
working on and thinking abouthow you communicate pay is top
of mind. But we often talk aboutthat in the context of
communicating pay.
(15:45):
So I'm going to share with youthe pay transparency continuum
that we often share here atPayscale. And then Sarah is
going to say, you can take thatand you can do it for data too.
So, when we think about the paytransparency continuum, there's
two ends to this where you'revery secret about pay. And
probably when I started mycareer in compensation, that was
(16:07):
very much where we were. Andthen at the far end of the
spectrum, you've got like beingvery transparent, fully
transparent, potentiallypublishing all your salaries
publicly.
There are very few organizationsthat do that. Where we see most
people falling is in the sharesalary range data for the
current role with the employee.So that tends to be the goal
(16:31):
where most organizations aretrying to get to. There's
obviously quite a lot of workthat needs to happen to get you
to that point. But as we thinkmore and more about career
progression, then the point inthe range where we have all
salary range data sharedinternally, so people have an
understanding of how theircareer can progress, how their
pay can progress, then havingthat wider view of just not
(16:53):
their own pay range, but maybethe pay range of the next roll
up or the next roll up will givethem an idea around career and
pay progression that's importantas well.
So we've talked about this manytimes here at PayScale. This is
our pay transparency continuum,but we're going to slightly
tweak it and you're going totalk us through how that might
apply to a data strategy, Sarah.
Speaker 2 (17:15):
Took the continuum. I love this spectrum analysis. And
I put the data strategy on topof this. So instead of thinking
about sharing pay on thiscontinuum, we're thinking about
sharing or talking about sharinghow the ranges are created and
how those decisions are made andthat decision about how
(17:36):
someone's aid is reached. At thepay secrecy end of the spectrum,
you have no information at all.
You know your pay, but nothingabout how it's created. I would
say at this end of the spectrum,you want at least your comp team
and probably your executiveboard to understand that
compensation strategy. Thatmight mean having it written in
(17:58):
some place, or having a set ofguidelines that everyone who's
working on compensation isfollowing so that you know that
everything's happening the sameway across different
practitioners, but they're notnecessarily sharing that with
managers and certainly not withemployees. On the other end of
that spectrum, have the fullinformation. If you had written
(18:19):
down your comp strategy, here'swhat it would say.
It's which surveys you're using,what jobs you match to, what
other jobs are in a comparablegroup, how the range was created
from that market data. And alongthe way, we have a couple of
other stop points. You can sharesort of high level philosophy
with your managers andpotentially your employees, or
you could share the strategyitself with managers and
(18:42):
employees, even if you don'tshare each data point. So
there's lots of different placesto be here. Right now, we're
seeing folks moving from thedata strategy secrecy into more
of the philosophy sharing,especially with managers and
sometimes with employees to helpdrive those better conversations
(19:03):
that Ruth was talking about.
Where it came from helpsunderstand how I can impact that
and what sort of levers I haveto pull to get paid more or to
grow in my career.
Speaker 1 (19:18):
So a good takeaway for all of you on the call today
is to think about where am I onthose two continuums of pay
transparency? And where might Iaspire to be? What do I have to
do to get there? And a lot ofthat will be influenced by kind
of your overall culture oftransparency within your
organizations. But as Sarahalluded to there, we know
(19:40):
there's a growing expectationfor stronger conversations
around pay and how pay isdetermined.
And we need to really kind ofequip ourselves and equip
managers who are frontlinefacing those questions and those
conversations around, you know,how the pay that an individual
(20:01):
has been derived. Okay, so we'regoing go move to our first poll.
Let's find out how transparentyou are with your data strategy.
So in the poll tab, you're goingto be able to vote there on a
number of options. So do youcommunicate your data strategy
in terms of how you're usingmarket data only with your HR or
(20:23):
senior management to managers?
Do you communicate to everybody?Are you in the move towards
transparency bucket or maybeyou're not transparent yet? Or
you maybe you potentially don'tknow. So we're going to let you
vote away there. It's reallyinteresting for us to see how
you're thinking and how this isevolving.
(20:45):
It's important for us, Sarah, Iguess, as we think about
supporting organisations ingiving them the data sources
that we do and the choices thatthey're making there.
Speaker 2 (20:58):
Absolutely. We work with customers that are in many
different places on the spectrumand understanding what the pain
points are and what they needfrom the software and the data
to to help push in the directionthat that's aligned with their
goals is really helpful. Sothese kinds of polls and talking
to customers and prospects tounderstand these pain points are
(21:21):
very valuable information forus.
Speaker 1 (21:25):
Okay, so I could see people voting away, but I think
we're leveling off. Yep, Amber'ssharing the results. Thank you,
Amber in the background. Right,so this is probably in line with
where we were talking aboutwhere people were on those
continuums. So we have themajority of you are really
communication to HR or seniormanagement only, but there is a
good body of you, 19 and a halfpercent of you nearly who want
(21:48):
to move towards transparency.
And we see a number of you whoare communicating to managers
only, because we said, know,they're really important
participants in this process,but a good number of you, 14%
communicating to all within theorganization. So some of you are
on your journey. Some of you areprobably in that middle part of
(22:08):
the journey that we, that weshared earlier. Okay, now
another question we often getasked is around data bias,
because there's this common mythSarah that there is one market
number out there and there isone data source that you can go
to and it's going to give youthat magic number and that is
(22:29):
going to be the right number ofwhat someone would be paid. And
I love the way you explain databias and why that is not the
case.
So I'm passing it over to you.
Speaker 2 (22:38):
All right. The unicorn myth of having the exact
answer is just there if youcould only find it. I don't mean
to be inflammatory with this,but all market data is biased.
Everything that you'reconsidering, buying, purchasing,
looking up on the internet hassome form of bias. And I'll talk
you through some of the morecommon ones and talk you through
(23:00):
how to handle that bias and whento think about it as a big deal
and when to just sort of knowthat it's there.
Starting at the top end of whatyou might pay for, surveys,
compensation management surveys,and data sources like Peer are
generally collected from bigcompanies. This is because
(23:21):
bigger companies are likely toparticipate in these surveys and
pay for these expensive dataassets. And I'll show you
exactly what that SKU looks likein the next slide and talk to
you about what that means. Otherdata sources that you might be
using, like PayScale's employeereported data or information
from Glassdoor or Levels. Fyi,tends to skew towards younger,
(23:47):
more technical respondents.
There you see more 25 year oldsoftware engineers taking out
those surveys, filling out thatinformation than you do 70 year
old accountants and more thanyou do blue collar worker
working in a union job. So thedata that comes out of that is
(24:07):
aggregated and massaged, but itrepresents that younger, more
technical population moreheavily than it does some of the
other populations that you mightbe hiring for. Interestingly, as
more and more states have addedjob postings, job pay
transparency requirements andlegislation, and more and more
(24:29):
job postings have payinformation, I've been watching
the aggregate job posting datashift a little bit here and
there. As states that generallyhave a lower cost of labor add
pay transparency legislation, Isee the market range at the
national level drop for a job.And then as a state where the
cost of labor is very high, astransparency legislation, I see
(24:52):
the market range for thenational level go up a little
bit because the underlying datamix has shifted as those
different states change theirlaws and as different companies
are then adding data into thataggregate data pool.
So big plug for the Bureau ofLabor Statistics here. They do a
really, really great, verystatistically sound job of
(25:16):
evening out both labor pooldistribution data as well as pay
data. So if you're not checkingthe Bureau of Labor Statistics
website every once in a while tosee what their take is, I
absolutely recommend it becauseit'll give you a much cleaner,
unbiased view of labor in thiscountry, at least, and of pay
(25:40):
across different jobs indifferent professions, at
different industries andlocations. The downside is it's
pretty tough to build acompensation strategy and plan
based only on the Bureau ofLabor Statistics data because
it's not granular enough andit's too old to be super
relevant. But it's a reallygood, free, easily accessible
(26:00):
reference point.
So just to pull the covers backon this a little bit, I'll show
you some of the biases in PEER.And I'm using PEER here, one of
PayScale's data sets, as arepresentative of a
participation based aggregateddata set. So this is my guess is
(26:21):
this is a lot like what the databehind a third party survey
would look like. I happen tohave access to this one, so I
want to show you here, but applythis thinking to any of the
large survey providers thatyou're considering publishing.
On the x axis here, what I haveare different size bins.
(26:43):
So each of these bins representscompanies of particular size. On
the left hand are companies thathave less than five employees,
and on the right side of thisaxis is companies that have at
least a thousand employees. Whatthe y axis is showing is how
much of the workforce isrepresented by people working at
(27:07):
that size company. So Bureau ofLabor Statistics is medium blue.
That's the one I trust.
What you can see here is thatabout 8% of The US workforce
works at a really small company,less than five employees. They
could be self employed. Theycould work with two people at a
very small firm. And about 18%of The US workforce works at a
(27:32):
company that has a thousand ormore employees. So there's kind
of this dip in the middle thatshows the trend over various
company sizes.
But you see quite a few peoplework at very small companies and
less than 20% work at thesereally large firms. When we look
(27:55):
at peer, and again, I'm usingpeer as a representative here,
we see that more than 90% of theincumbents that are included in
the peer data set work at bigcompanies. So we are
overrepresenting data from verylarge companies here compared to
The US labor force in general.The upside is if you work at a
(28:19):
company that has a thousand ormore employees, the data that's
in PEER or a third party surveythat's collected in similar way
is very representative of whatpay is at those larger firms.
The downside is if you work at asmall firm, if you work at a 20
person company, using those dataassets and data sources may not
(28:42):
give you a very clean reflectionof what pay would look like at
that smaller company.
And you may need to go look atBureau of Labor Statistics or
other places to understand howthat large firm pay relates to
what it would look like at asmaller organization that you're
representing.
Speaker 1 (29:05):
Oh, sorry, sorry. I
Speaker 2 (29:11):
wanted to point this out and I won't go into as much
depth, but like other surveys,peer is also biased towards
particular industries. And thiscan be very meaningful to folks,
and sometimes it's not. But ifyou're using an all, all, all
(29:31):
data cut from peer or from adifferent survey provider, it's
important to understand whattypes of incumbents and
organizations are making up thatall, all, all data cut. For a
job like a compensation analyst,and we probably have some on the
phone, you may already knowthis, but compensation analysts
make a little bit less incolleges and universities and
(29:54):
slightly more in health care,But the ranges are quite
overlapping and really industrydoesn't impact the pay of a
compensation analyst very much.We see compensation analysts get
hired from one industry into aposition in another industry and
then move back to a thirdindustry all the time.
The talent pool has a skill setthat's useful to lots of
(30:14):
different companies. On theother hand, a job like a project
manager makes significantly moremoney in construction or in an
architecture firm than they doin finance. It's almost like
it's a different job, but it'scalled the same thing. Like 40%
more if you work in architecturethan in finance as a project
manager. So if you are pricingjobs and you pick up a survey
(30:37):
that is mostly architecture andconstruction, but you're trying
to price jobs in finance, youcould end up with a view of a
project manager role that saysit should be paid significantly
more than what the market rateactually is in finance, even
though it kind of looks okay forComp Analysts.
So something to be aware of andsomething that we can tell you
(31:02):
how to handle. But basicallyevery data source you get will
have a bias like this, wheresome industries are more
represented and some are lessrepresented. So what you're
Speaker 1 (31:14):
doing is to make sure that you discuss with your
vendors on that. So you're goingto talk us through this, I
think.
Speaker 2 (31:20):
Absolutely. Just ask ask your vendors. I'll I'll send
you that file if you wanna seeit for peer, and understand
where those those data arecoming from. And my guidance
here is if you're using an allall cut, which I love that
there's lots of data in there,that's a robust thing to do. If
the bias of the underlyingdataset doesn't match the
(31:44):
firmographic details of yourorganization, you could have a
problem.
The way to combat that problemis to be more specific, to use a
specific scope or data cut thatis oriented towards firms like
you. So even if you're using adata set like Peer and you know
it's got a lot of healthcare andretail in it, if you're in
(32:05):
materials and mining or you'rein finance, trimming down that
list and looking specifically atorganizations that are like you
will help make sure that thebias of the underlying data set
isn't impacting your marketranges and sending you often
into a range that's really notappropriate to the labor pool
(32:26):
that you're hiring from.Occasionally, there's not data
available for the industry thatyou're really interested in. And
so what we see folks do there isuse some kind of adjustment
factor to say, well, I'm tryingto price a dentist in the
construction industry. There'sno other dentist in the
(32:48):
construction industry, but Iknow that dentists in the
healthcare industry make alittle bit more than they do in
other industries.
There are actually dentists inother industries, even though it
doesn't sound like there wouldbe. They're dentists and
insurance. And I know that theymake a little bit less. So I
might take that overall numberand say, I think it's probably
(33:08):
less than health care because myindustry is likely to pull more
from talent that would also workin the insurance industry or
similar than it is to pulldentists that work in the health
care industry. So you can makesome adjustments.
And comp pros that have beendoing this for a long time often
have a back of the envelope wayof doing this for lots of
(33:29):
different scenarios.
Speaker 1 (33:34):
Okay, so we've covered data biases. We'd like
to hear if this is somethingthat you've been thinking about
in terms of your data strategy.So we have another poll going in
the polls tab. Do you monitoryour data processes for bias? Is
this something you docontinuously, occasionally,
you're planning to, you weren'taware of the issue and hopefully
(33:55):
you are now after Sarah's greatexplanation or you don't know.
So we're just interested to seeif this is something that
resonates with you and somethingthat you're thinking about as
you work with different datasources. And I can see you're
busy voting away there and I cansee some questions coming in. So
we are going to move to the liveQ and A soon. So any questions
(34:17):
you've got specifically aboutthe biases that we've talked
about so far about transparency,get those into the Q and A and
Sarah and I will be able toaddress those very shortly. We
have one more topic to coverbefore we move into the live Q
and A.
But let's have a look at whatyou are saying about do you
monitor your data processes forbias?' So occasionally, so we've
(34:40):
got nearly 38.5% of peoplesaying they occasionally monitor
and we've got 20% of you who aresaying, nearly 20% of you saying
continuously and then some ofyou who are planning to do it.
So it's good to see, I thinkwe've got a group of mature
compensation practitioners hereon the line. This is something
that's a topic that you're awareof. Hopefully this is just a
(35:02):
good refresher on why it'simportant to think about those
biases as you work with yourdata sources. So thank you very
much for responding to thatpoll.
Okay, so you can't talk aboutanything to do with HR or
anything to do with anythingtoday without talking about AI,
to be honest. And we've beentalking about AI as it relates
(35:24):
to data and data strategy, aswe've done this webinar series
over the last this is the thirdyear we've kind of run it. And
there's definitely been a hugemove in terms of progress, I
think, Sarah. We've beenthinking about it over that time
very actively. We've beenthinking about it in terms of
how we incorporate it into thedata solutions that we have here
(35:47):
at Payscale.
But Sarah is going kind of talkyou through where we're at today
in terms of how you should thinkabout AI as part of your data
strategy.
Speaker 2 (35:58):
Yeah, as Ruth said, although I've been doing AI
since before it was cool andPayScale's been doing AI since
before it was cool, 2023 was theyear that generative AI exploded
into the public discourse. Andin 2024, we saw the use of AI in
HR kind of pop up and start togo more mainstream. This year,
(36:21):
HR professionals are embracingAI, particularly for tasks like
compensation management and jobdescription creation. As part of
an annual poll that we do calledCompensation Best Practices
Report, which you can get to viathis link, we gather responses
from 3,600 comp and HRprofessionals a mixture of
(36:42):
PayScale customers and alsofolks that don't use our tools.
This year, we found that 71% ofthose respondents have a
positive sentiment about usingAI for market pricing.
And similarly, a growingmajority of respondents are
interested in AI for other usecases, like recommending pay
increases, helping supportlegislative compliance, and for
(37:06):
documentation andcommunications. In the
compensation world, as I alludedto a little bit earlier with my
back of the envelopecalculations, data driven
techniques to extrapolate fromexisting data are our bread and
butter. Many of the methods thatare common in our field, taking
a weighted combinations ofunderlying data sources or aging
(37:28):
the data or calculating andusing differentials, are really
just simple implementations ofAI. Even doing a Google search
is basically using AI. So in thenow and certainly in the future,
I expect that AI tools, enabledtools will make these methods
that we have been using foryears more robust, more
(37:52):
scalable, more repeatable, andmore accessible to folks that
don't have years and years ofexpertise and know the data
inside and out.
With tools like Explore, whichwe launched this week, and the
underlying AI powered dataasset, we're using AI to find
the same patterns that we've allbeen measuring and using as part
(38:13):
of the art of compensation. Butleveraging AI and the data at
scale to do it instead of doingit one off each with their own
methodology. This helps uscorrect the biases in the
underlying data sources, and itimproves the accuracy of
calculating a pay market rangewhen the data are sparse. So if
(38:34):
you want to hear more aboutthis, connect with me on
LinkedIn or schedule a demo andwe'll show you what those tools
look like. The same techniquesalso help us discover some
patterns that are prettyinteresting and insights about
pay and compensation strategythat haven't been measurable
before.
(38:54):
So compensation decision makers,whether that's a compensation
analyst, an HR generalist, a VP,an exec, or a board need
education on how artificialintelligence can make
compensation related decisionsmore accurate and more
efficient. AI can assistemployers with job management,
(39:15):
including writing and analyzingjob descriptions. And AI is also
being currently used and moretools coming for market pricing
and pay analysis. Over the nextcouple of years, all of these
boxes that don't have stars, Iexpect to see more and more
areas where AI makes ourexisting workflows more
efficient and helps us uncoverkey insights that can drive our
(39:39):
compensation strategies. I willsay the potential for AI to
extend bias rather than mitigateit is still a very real concern,
especially where AI canunknowingly perpetuate
historical salary discriminationand pay disparities for
protected classes.
(39:59):
Ultimately, human interventionis required to ensure that AI
powered algorithms are properlyinformed and producing fair
outputs. I like to say that AIcan scale bias or reduce bias at
scale, But our judgment and ourexpertise helps make sure that
we push towards that reducedbias at scale rather than the
(40:22):
scale bias. I also want to flagfor folks that are doing this
work day in and day out, paygaps are not excused by the
reliance on flawed models. If anemployee in this country, if an
employee can demonstrate thatthey are paid differently from
another employee for equal orsimilar work, the burden shifts
(40:43):
to the employer to demonstratethat that pay differential is
due to a permissible reason,like differences in skills or
differences in performance ordifferences in location or
responsibility, rather thanconscious or unconscious bias or
discrimination, even if an AItool made that recommendation.
(41:04):
So employers, not the tool, arelegally responsible at this
time.
Speaker 1 (41:12):
Great, yeah, thanks for calling that out, Sarah.
Something I was going to pick upon. So I mean, for me, as I
think about it as a compensationpractitioner, there are more
data sources available today andthe level of compensation
decisions that we're needing tomake are more granular. When I
think about the whole skillsconversation and how that plays
(41:32):
into this, really, you know,having advanced decision support
was the only way we were goingto get to the place where we
could start to correlate allthose different information
sources and information datapoints. We do need to get
comfortable as practitioners interms of thinking about how AI
can support us in that processand can help to improve the
(41:54):
decision making that we aredoing whilst being aware all the
time about what type of AI isbeing used.
So if you're talking to a vendorand they're talking about AI,
ask them how it's beingdeployed, make sure you
understand how it's being usedand in what way so that you can
make an informed decision andfeel comfortable about that as
(42:15):
you move forward. Okay, so lastpoll before our Q and A poll.
How are you using AI today? Weknow there's quite a few
different use cases that peoplehave. We've surveyed on this
before in our compensation bestpractice report, but interested
to see how prevalent the use ofAI is in terms of the work that
(42:36):
you're doing today.
So I'm not going to read out allthe options there. If you want
to vote away, I think you canselect all that apply in this
particular poll. But are youusing it in benchmarking,
monitoring for pay equity,collecting intelligence on
skills for recruiting, writingoffer letters, using it to speed
up survey participation, thatwould be nice and year over year
(42:59):
updates. How are you using ittoday? And then we will move in
to the much promised Q and A.
I can see a few questions inthere. So we'll be getting to
those very shortly. Okay. Are welevelling out on votes? Okay,
(43:31):
none of the above or unsureseems to be the majority at the
moment.
Of course, that may be theunsure category. So there has
historically been a lot of AI.It was just probably just called
machine learning at that pointin market data sources. So I'm
pretty sure that if you're usingany form of market data sources,
it's got some form of AI inthere. And it just, you know,
(43:55):
you may not just know it underthe term AI.
We can see people are also usingAI to collect intelligence on
skills for recruiting,upskilling, career pathing and
or compensation. So there wasnearly quarter of you using it
for that. And then some of youusing AI to benchmark, about
(44:15):
just over 10%, nearly 12% of youusing AI to benchmark and price
jobs or predict pay ranges. Soclearly an area where this is, I
think we're going to see thegrowing influence of AI, Sarah.
Don't know how do you feel whenyou look at those poll results?
Speaker 2 (44:34):
It's interesting, you know, and I think if I pulled my
team, which is all people whobuild AI tools, we would get a
similar distribution of placeswhere people are seeing the
value already and places whereit's still kind of hype and
like, I don't know. And then asyou mentioned, Ruth, places
where the old fashioned ways areactually not very dissimilar
(44:58):
from what we think of as AI. Wetalk about pay equity, for
instance, that's been based onregressions for many years.
Regressions fall into themachine learning and AI umbrella
because it's predicting thingsgiven data. So this is really
interesting to me.
And certainly the collectingintelligence on skills, career
(45:21):
pathing and compensation is anarea where AI, especially the
next generation text processingfrom large language models can
really make a big difference inthe kind of information that is
available to us. So fascinatingto me to see the spread and the
(45:42):
different places where peopleare interested or already using
AI tools.
Speaker 1 (45:49):
So yeah, thank you for sharing all your insights,
everybody, through those threepolls. So closing us out,
really, you know, the keymessages that we wanted to get
across today is, you you need adata strategy. And the aim of
that data strategy is really togo for consistency of
methodology so that you have arepeatable process that makes
(46:11):
sense for your business. Aimingfor transparency becoming
increasingly important, youknow, you're going to have more
demands on you to be able toexplain how you determining each
employee's compensation and howthat number was derived either
from managers who are havingconversations with employees or
(46:31):
directly with employeesthemselves. So you think you
getting comfortable with howtransparent you want to be about
the type of data sources thatyou're using as a journey I
think you need to go through.
Do you want to cover off thelast two, Sarah?
Speaker 2 (46:46):
My goal today was to flag the biases issue and to
show you that it's notnecessarily a bad word, that all
of the data assets that youconsider are biased in some way,
and that there are things thatyou can do to mitigate that bias
and to work with it rather thanagainst it. And then to
(47:07):
highlight a couple of key placeswhere we're seeing AI deliver
value to practitioners like youand have you listen to each
other a little bit through thatpoll to get an idea of where
those tools are moving. Ifyou're interested in hearing
more about AI tools, drop a notein the comments. We could do
(47:27):
another webinar that's morefocused on that, either from a
market landscape or thoughtleadership perspective.
Speaker 1 (47:37):
Okay, all right. Thank you. Right, we're going to
move into Q and A now. There isanother poll running in the
background. If you would like tofind out more about how Payscale
can help you address some of theissues that Sarah and I have
talked today, learn more aboutthe AI innovation journey that
we're on here at Payscale, thenif you click in that poll,
(48:00):
someone will reach out to you toget back to you to help you
understand that.
We're going to jump into the Qand A. I'm going to go back to
one of the first questions thatwas asked of us, which was when
we were really talking abouttransparency. And it says: Do
you recommend employers having awritten codified comp data
(48:20):
strategy versus just somethingyou used in practice? Classic
consulting answer would be itdepends and I think it really
does depend. I'd say thatdepends on the size of your
organization.
If you have many compensationpeople and HR people kind of
working on this together acrossthe organisation, then quite
often having some kind ofwritten data strategy that
(48:44):
people can refer to would beuseful. If you're just like a
sole person on your own andyou're doing this as the comp
analyst on your own and you'reworking with your HR team,
probably useful to train them onit. But do you necessarily need
something in writing? Probablynot, I would say. So it really
kind of depends on how manypeople you're trying to ensure
(49:04):
that consistency applies to.
It's like what is right for youto ensure you're applying the
consistency and the decisionsthat you've made around how
you're going to approach data.Now Sarah, there are a couple of
questions about adjustments inthere.
Speaker 2 (49:21):
Yeah.
Speaker 1 (49:21):
Do you want to take those two and talk to those two?
Speaker 2 (49:24):
I will. So we got two questions that are related. One
was, do you have anyrecommendations for selecting
the appropriate adjustmentpercentage for difference in
industry? And the other was,what advice would you give for
making adjustments where thereisn't applicable industry for
those who are newer to comp? I'mgonna give you two answers.
(49:46):
One is go get a demo of Explorebecause we take the hard part
out of it for you. And we've gotlots of experts that do this at
a job by job level so that youdon't have to be an expert to
understand how to do it. ButI'll tell you how to do it
yourself, too. The biggest pieceof advice I can give you here is
(50:09):
don't think about the industrythat you're working in for a
particular role. Think about theindustry or industries that
you're hiring from.
So if you're hiring, accountantsare a good example, cop analysts
are a good example. If you lookat your books and you say, Okay,
my employees that are in thisrole came from a large number of
(50:30):
different industries. And when Ipost this role, the next few
people I'm hiring could alsocome from education or
construction or retail ortechnology. Then in that case,
an all, all, all data cut, noindustry differential is going
to give you the bestrepresentation of what pay is
(50:52):
for that role, because it's notvery dependent on which industry
you're in. On the other hand, ifyou are pricing jobs or a family
of jobs within your organisationwhere you are consistently
pulling from one industry thatis where you're always hiring
from that industry and you wouldnever consider hiring from a job
(51:15):
outside of that industry.
Then there's kind of twobranches in this decision tree.
One is to think about whetherthe job itself is particular to
the industry that you're hiringin. So if you're working in
agriculture and you're hiring aagricultural laborer, all of the
data from any survey that showsyou what the pay is for
(51:38):
agricultural laborers isbasically going to come from
agriculture. It's not going tocome from other industries.
Speaker 1 (51:53):
I've lost your sound, Sarah, I think. Yes. I'll just
check whether it was me or you.
Speaker 2 (51:59):
Can you hear me now?
Speaker 1 (52:00):
Yes, you're back.
Speaker 2 (52:02):
Okay, great. I have a little bloopity bloop. So in
that case, the job itself isrepresenting the pay in the
industry because it's specificto the industry. The trickiest
part is when you have a jobthat's represented across many
industries and the pay variesacross those industries, like a
(52:23):
project manager that I showedbefore, where they make 40% more
in construction than they do infinance. In that case, what
people generally do, and this iswhy I'm so excited about
Explore, but what peoplegenerally do is they look at
average differentials acrosslots of jobs and say, well,
generally jobs pay a little bitmore in this industry than they
(52:47):
do in this industry.
So if I'm pricing in finance, Imight pay 8% more than in the
construction industry becausemost jobs pay a little bit more
in finance. That doesn't workperfectly, as this example
shows, because some jobs paymore in one industry and less in
another industry. But that'sgenerally what the best practice
(53:11):
has been until the tools havegotten better and caught up with
being able to learn more jobspecific information about how
industry affects pay. So let metry to sum it up. I know that
was a lot of information.
One is in many, many cases, youdon't need to worry about
industry at all because you'repulling talent from across
(53:34):
industries and you're basicallypaying a market rate that is not
dependent on an industry. That'scase one. That's probably 70 to
80% of the jobs that you'repricing. Case two is you're
pricing a job that's only foundin a couple of industries and
all of the market data thatyou're looking at is coming from
those industries. They don'tneed a differential.
(53:56):
That's probably 10%. So nowwe're at like 80 to 90% of the
data, no differential. That last10 is the hard one where the pay
varies across industries and thedata that you have isn't totally
represented by the industry thatyou're pulling from. And then
(54:18):
that's where the art part comesin. And I think AI can make a
really, really big difference.
But the general practice overthe last several years has been
to look at what I call a jobagnostic differential of how
does pay generally differbetween these two industries.
You can get that from a dataasset like PayScale has, or you
(54:39):
can look it up on BLS and youcan see how pay varies across
industries.
Speaker 1 (54:45):
Okay, thank you for that answer to those two
questions, Sarah. I think maybewe've got time for one more
question. Someone was callingout that they are having to
start from scratch at theircompany to figure out pay
ranges, but they don't have themoney to potentially buy pay
scale data sources. What's theirrecommendation on how they can
begin? Well, the good news isthat we have a new freemium
(55:07):
offering that is going to belaunched shortly, that will be
able to give you access to thatnew explorer experience that
Sarah has been talking abouttoday.
If you want to find out moreabout that again, click into the
poll or reach out to us and wecan make sure we can fill you in
about that. So that is one way,depending on the number of jobs
that you've got to price,obviously, you know, there are
(55:29):
freemium options around fromsome vendors. Sarah, you talked
about other free data sourcespotentially like the BLS data.
You know, are other data sourcesthat you can use as well. But
I'd also think about thispotentially, you know, depending
on the number of jobs thatyou've got to price within your
organization from an ROIperspective.
(55:49):
Because quite often people go, Ican't afford to buy pay data.
When you need to think aboutactually what is the impact of
not using pay data and the costto your business potentially of
not doing that. You price fivejobs wrong and that could
actually aggregate up to a valueof $2,030,000 US dollars that
you're making a mistake on andeven more potentially. So, you
(56:12):
know, think about framing itfrom the perspective of the ROI
perspective to yourorganization. And particularly
when we're in the slightlycautious macroeconomic
environment that we're in at themoment, where we know many
organizations are highly costsensitive, You know, the right
pay decision made that isaccurate and near term real term
(56:37):
data, know, that's reallyimportant.
So think about how you framethat from the kind of ROI
perspective to your business.And with that, I think we are at
time. So I'm going to ask Amberto come back on stage and close
us out. But thank you all verymuch for your time today. And
thank you, Sarah, for sharingyour insights.
Speaker 2 (56:54):
Thanks for having me, Ruth.
Speaker 3 (56:57):
Yes, thank you both so much for sharing all of your
insights and expertise on thistopic. It was a wonderful way to
kick off our data series. Youwill we'll be doing a few more
of these events on this topic,covering into different areas
over the next couple of months,so stay tuned. Okay. The session
(57:17):
today comes in around recordingand slides.
We will send both the recordingand slides in one to two
business days, as well as someof the follow-up materials that
Ruth discussed and someadditional ones. So please keep
your eye out for those. Withthat being said, we're out of
time. So I will wish you all awonderful rest of your day, and
(57:41):
we hope to see you at a futureevent soon. Thanks, everyone.
Speaker 1 (57:46):
Thank you.
Welcome everybody. Thanks for joining us today. My
(00:02):
name is Ruth Thomas. I'm ChiefEvangelist and SVP of Marketing
here at Payscale. So I help todrive our thought leadership and
oversee our marketing teamsthrough all the work that we do
here at Payscale.
Sarah, over to you.
Speaker 2 (00:18):
Good morning, good afternoon, good evening and good
night. I'm Sarah Hillenmeyer.I've been building data and AI
enabled tools in HR tech,medicine and other industries
for more than fifteen years. AtPayscale, I lead our AI and data
insights teams within ourresearch and development
organisation.
Speaker 1 (00:36):
Thank you, Sarah. Okay, so for the last couple of
years, I think we've run a shortdata thought leadership webinar
series. So we're doing the sameagain this year. Really, we aim
to bring you some key insightsaround how to manage data
considerations for when you arecreating a data strategy. We aim
to bring together thoughtleaders, data experts like
(01:00):
Sarah, practitioners, customers,partners to join us in that
discussion.
This is the first of thoseepisodes of this year's series.
We have another webinar comingup on May, where we will also be
sort of building on the contentthat we've covered today. But in
today's session, we reallywanted to cover these things
(01:20):
here. So whether you're pricingone job or whether you're
pricing thousands of jobs,having consistency in terms of
your approach to using data andhow you use that data to support
the business goals that you wantto do is important. So we're
going to start the session offsharing you with a model that we
think about when we think aboutbuilding a data strategy within
(01:44):
an organization.
And then there are some otherhot topics that often come up in
conversation as we think aboutbuilding a data strategy. So
data transparency, data biasesand data and AI. So we're going
to cover off those three in thissession as well, sharing
thoughts and kind of like wherewe're seeing in practice and
some kind of key innovationsthat we're seeing starting to
(02:07):
happen in the market. So that'sthe aim of today's session. So
let's get started.
I'm going to kick us off interms of thinking about what
makes a good data strategy. Andwe think about it as having
three key components. So we'regoing to walk you through this
Sarah and I. So the first is theinput data, the data that you
(02:28):
potentially purchase fromproviders or you get free from
government sources or from theweb. And Sarah is going to talk
you through about the variousthings you should be thinking
about as you bring, select thatdata.
Then there's the kind of how areyou going to use that data
internally. So I'll run youthrough that, like what are the
considerations you think aboutas you bring that together
internally to support yourcompensation processes. And then
(02:52):
in today's world of paytransparency, the discussion
around what you should or shouldnot communicate in terms of, you
know, the data that you're usingand how you're using it is
becoming more pertinent. Sowe'll talk you through some
guidelines and some kind of thethings that we're seeing
happening across our customerbase. So I think Sarah, you're
going to start us off here withquestions to answer as you
(03:14):
evaluate any data source.
Speaker 2 (03:16):
Great. When you're considering using a data source,
whether it's a free data sourceor something that you're
thinking about purchasing, theseare the questions that I
recommend answering and keepingtrack of. The most important
question is whether the datasource contains pay data about
the jobs of your organization.You want to be able to find
most, let's say 80% or more ofthe roles that you're trying to
(03:40):
price within the data sourcethat you're considering. We
generally recommend that foreach role you're pricing, you
find a job in the data sourcethat's at least a 70% match in
terms of the skills, keyresponsibilities, experience,
and reporting structure.
The next thing to consider isthe repeatability. How easy is
(04:02):
it to price jobs in a consistentway year over year or time
period over time period? Are thechanges that you're seeing in
the data that you're buyingreflective of the market? Or is
it a potentially noisy datasource that's pulling from
different populations atdifferent times and showing you
variation that isn't necessarilywhat you would want to bake into
(04:23):
your merit increases and yourmarket increases for your
employees that year. Next up,this is more of a checkbox from
my perspective, although we cantalk about how to evaluate
methodology and explainability.
But any data source that you buyor use, you should understand
where the numbers are comingfrom. How are they calculated?
(04:45):
If they're calculated, ifthey're not calculated and
they're collected empirically,where did the data come from?
When was it collected? Whathappened to it from that point
of collection to the point thatit's in your hands?
That's an important thing toknow. Even if you don't feel
like you can totally evaluateit, you should know the answer
to that question. Number four ishow old are the data? When were
(05:08):
those data points collected?Have they been aged since the
time of collection?
Aging, if you haven't heard it,is a term that refers to
adjusting the market numbers bysome fixed percentage per month
or per year to account formarket trends that have changed
the pay since the time that thedata were collected. Number five
(05:30):
is biases. We're going to talk alot more about this later, but
you should have a sense of theunderlying biases in the data
sample and how that might impactthe way that you need to use
that data going forward. Andlastly, and Ruth will touch on
this more, as you think throughwhich data sources to use for
(05:50):
particular tasks, think aboutwho else can see the data. Is it
something that your employeescan access?
That might change the way thatyou think about using that data
and it might influence yourcompensation strategy and your
communication strategy aroundthat data versus having
something that you know thatyour employees don't have access
(06:10):
to. Ruth, can you talk usthrough the processes of how to
deal with that data once youhave it?
Speaker 1 (06:18):
Yeah, so you've got all that wonderful data, you've
sourced your data. I guess thenext step is what are you going
to do with it and how are yougoing to use it internally?
Probably the biggest decisionpoint is how you're going to
aggregate multiple sources ifyou're using more than one
source of data. We're not goingto talk today about the
(06:39):
different types of data sourcesthat are available. We will
potentially cover that inanother version or in another
episode of this series.
That's something you'd like usto cover as well, leave some
comments in the Q and A and wecan make sure that we cover that
in more detail. But there arelots of kind of considerations
in terms of aggregation, interms of potentially, what
(07:00):
percentile are you going to usewithin each of the survey
sources that you're using? Areyou going to weight them
differently? So do these sourceshave more importance or less
importance for you? And doesthat change by the different
talent groups that you havewithin the organization?
How are you going to deploy? Howare you weight? Which is more
important for you? Sarahmentioned ageing, so you know as
(07:25):
you bring the different datasources together you're going to
need to think about your agingstrategy so that you're looking
at them all essentially from thesame time frame as much as is
possible. And then you know areyou going to use the same
methodology for aggregationacross different job groups or
different talent groups?
So that's an important questionbecause you may have talent
(07:47):
groups that you know, wherethere is a high demand for
talent or that the labor marketis moving very fast or there is
a hot demand for certain skills.And so therefore, how are you
deciding to use data to thinkabout pricing jobs, to think
about building pay ranges, maybeslightly different for that
group than for another groupwhere they may be sitting in a
(08:09):
more potentially stable talentmarket. And the same when it
comes to thinking about geodifferences as well as like how
are you going to factor thatinto bringing all these data
sources together. So that'sprobably the most important step
here in terms of developing yourinternal processes. But you also
need to think about how you'regoing to handle the data that
(08:30):
you're bringing in relative towhat exists in your organization
today.
So you may have some old rangesaround within the organization.
They may be job based or gradebased. What are you going to do
with this data relative tothose? And then also you really
need to do that comparison,which I know you will do. It's
that comparison of the marketdata with your current
(08:51):
incumbents.
But you may need to sense checkthat a little bit as you refine
the data sources that you'rechoosing and the data sources
that you're aggregating andbringing together. Obviously
many people will use the marketdata to help build pay ranges.
Again, you need to think aboutwhat your strategy is for that
(09:12):
and is that going to beconsistent across different
talent groups. And then today'sage, obviously many of us are
using the data that we get fromthese surveys to post in job
postings in order to comply withpay transparency legislation. So
how are you going to go aboutcalculating the range that
you'll actually list in the jobposting?
(09:34):
Again, not something we're goingto cover in-depth today. We have
a detailed webinar that talksabout that whole process of
selecting job postings. I'll askAmber to include that link in
the follow-up. But it's reallylike I have my market data. I've
made it, I've made somediscerning decisions about the
data I'm bringing into theorganization.
(09:55):
And then thinking about thattheme of consistency, how am I
going to apply that in terms ofhow we will use it within the
organization? I don't know ifthere's anything you wanted to
add there, Sarah. No. No. Okay,cool.
And then the third step in ourmodel, we said, you know, in the
age of pay transparency isthinking about what you're
(10:16):
actually going to communicate.So previously, I mean, I've
worked in compensation for manyyears. This was something that
we did in a back room somewhere.Nobody got to know what we were
doing. And probably the onlyinteraction point, for example,
a line manager might have withthe market data was potentially
through the merit review processor a discussion during the
(10:39):
hiring process.
There is a demand, you know, tomake conversations more two way
and to make conversations moremeaningful about pay for people
to be able to understand whereyou are potentially sourcing
market data. So I'm sure many ofyou on the call have been
through the experience of havingan employee or potentially a
(11:02):
candidate through the hiringprocess come to you and say,
well, this is the data that I'mseeing for the job. Know, in
order to be able to counterthat, you need to be able to
have a conversation about thedifferent types of market data
sources, potentially wherethey're sourcing that data from.
So that might be job boards orGoogle even nowadays, you know,
(11:23):
that's very easy through Googleto get that data for a job
relative to how you're choosingmarket data. And that's going to
force you in a way then to bemore transparent.
And so that's why organizationsare starting to think more
about, you know, how far do wego in terms of sharing data? So,
you know, what is it you'regoing to share? You have various
(11:43):
options. Are you going to sharethe methodology? Are you
actually going to share themarket data ranges?
That's probably one people feel,you know, least likely to share.
And then are you going to tellemployees potentially where they
sit against that range? And thenare you going to explain the
data sources that you've usedand why? That's probably quite
(12:05):
technical conversation for mostof your audience in your
organization. But there isprobably a way that you can tell
that story in terms of, youknow, a more generic level.
It's like we choose market datasources that represent X, Y, Z
and we use that, you know, to dothis. So having a talk track to,
you know, the type of datasources that you're using
without going through a highlytechnical soliloquy about how
(12:29):
you, you know, selected all ofthose processes per like the
stuff that Sarah talked tothrough at the beginning. And
then, just thinking about whatmore do you need to do? So what
regulations apply to us as wethink about communicating pay
and what are our competitorsdoing in terms of how they
communicate about market data aswell. So an interesting
(12:49):
conversation, interested to hearyour thoughts, interested to
hear your questions as we moveinto the Q and A about how
comfortable you feel at thispoint about communicating market
data and what elements of thatyou think that you should be
communicating.
Speaker 2 (13:05):
And we'll talk more about different places to land
on that transparency spectrumand how to migrate from one
section to another if you wantto increase your transparency.
But just to wrap us up here,when we say data strategy or
compensation data strategy,those three pillars are the
things that we're talking about.Which input data are you using?
(13:28):
Does it cover your jobs? Is it arepeatable data set that has
stability in it where the trendsare real?
Can you explain it? Is it fresh?Is the bias something that you
can work with? And again, we'lltalk more about that. And then
how do you process that data togenerate pay decisions for all
of your employees?
(13:49):
And then thirdly, how do youcommunicate that? And obviously,
these three pillars areinterdependent, but we think
it's useful to think about eachof them as separable. You can
think about each of theseseparately, even though the data
choices you make may influenceyour communication strategy or
(14:09):
vice versa. If you're startingout, just start marching down
this line. Pick your data, thinkabout your process, determine
your communications.
Speaker 1 (14:19):
Yeah, so this is a great reference slide for you.
You obviously will get a copy ofthe slides in the follow-up
communications that will go out,but this is probably, you know,
is where we pull together thatmodel for you. A good checklist
slide for you to take back toyour own organisations. Okay, so
we've brought up the topic ofdata transparency. Let's dig
(14:41):
into that a little more.
I talk a lot about paytransparency. You would have to
have been living somewhere verystrange to not know if you're
working in the field ofcompensation and HR that
transparency is a thing. It'sdriving a lot of what we do and
decide around compensationpractice today. And that's
(15:01):
obviously particularly beendriven by legislation. And that
legislation is really aroundpromoting fairness, equity and
trust by ensuring that employeesunderstand how their pay is
determined and how it comparesto pay of their peers.
So that's obviously about openlysharing information and we know
many of you have been on thatjourney. We see that in our
(15:24):
results from our annualcompensation best practice
study. And also through thewebinars that we do, where we
poll you, we're gonna go up onour polls coming up in a minute.
But we know, this is a keyactivity that you've been
working on and thinking abouthow you communicate pay is top
of mind. But we often talk aboutthat in the context of
communicating pay.
(15:45):
So I'm going to share with youthe pay transparency continuum
that we often share here atPayscale. And then Sarah is
going to say, you can take thatand you can do it for data too.
So, when we think about the paytransparency continuum, there's
two ends to this where you'revery secret about pay. And
probably when I started mycareer in compensation, that was
(16:07):
very much where we were. Andthen at the far end of the
spectrum, you've got like beingvery transparent, fully
transparent, potentiallypublishing all your salaries
publicly.
There are very few organizationsthat do that. Where we see most
people falling is in the sharesalary range data for the
current role with the employee.So that tends to be the goal
(16:31):
where most organizations aretrying to get to. There's
obviously quite a lot of workthat needs to happen to get you
to that point. But as we thinkmore and more about career
progression, then the point inthe range where we have all
salary range data sharedinternally, so people have an
understanding of how theircareer can progress, how their
pay can progress, then havingthat wider view of just not
(16:53):
their own pay range, but maybethe pay range of the next roll
up or the next roll up will givethem an idea around career and
pay progression that's importantas well.
So we've talked about this manytimes here at PayScale. This is
our pay transparency continuum,but we're going to slightly
tweak it and you're going totalk us through how that might
apply to a data strategy, Sarah.
Speaker 2 (17:15):
Took the continuum. I love this spectrum analysis. And
I put the data strategy on topof this. So instead of thinking
about sharing pay on thiscontinuum, we're thinking about
sharing or talking about sharinghow the ranges are created and
how those decisions are made andthat decision about how
(17:36):
someone's aid is reached. At thepay secrecy end of the spectrum,
you have no information at all.
You know your pay, but nothingabout how it's created. I would
say at this end of the spectrum,you want at least your comp team
and probably your executiveboard to understand that
compensation strategy. Thatmight mean having it written in
(17:58):
some place, or having a set ofguidelines that everyone who's
working on compensation isfollowing so that you know that
everything's happening the sameway across different
practitioners, but they're notnecessarily sharing that with
managers and certainly not withemployees. On the other end of
that spectrum, have the fullinformation. If you had written
(18:19):
down your comp strategy, here'swhat it would say.
It's which surveys you're using,what jobs you match to, what
other jobs are in a comparablegroup, how the range was created
from that market data. And alongthe way, we have a couple of
other stop points. You can sharesort of high level philosophy
with your managers andpotentially your employees, or
you could share the strategyitself with managers and
(18:42):
employees, even if you don'tshare each data point. So
there's lots of different placesto be here. Right now, we're
seeing folks moving from thedata strategy secrecy into more
of the philosophy sharing,especially with managers and
sometimes with employees to helpdrive those better conversations
(19:03):
that Ruth was talking about.
Where it came from helpsunderstand how I can impact that
and what sort of levers I haveto pull to get paid more or to
grow in my career.
Speaker 1 (19:18):
So a good takeaway for all of you on the call today
is to think about where am I onthose two continuums of pay
transparency? And where might Iaspire to be? What do I have to
do to get there? And a lot ofthat will be influenced by kind
of your overall culture oftransparency within your
organizations. But as Sarahalluded to there, we know
(19:40):
there's a growing expectationfor stronger conversations
around pay and how pay isdetermined.
And we need to really kind ofequip ourselves and equip
managers who are frontlinefacing those questions and those
conversations around, you know,how the pay that an individual
(20:01):
has been derived. Okay, so we'regoing go move to our first poll.
Let's find out how transparentyou are with your data strategy.
So in the poll tab, you're goingto be able to vote there on a
number of options. So do youcommunicate your data strategy
in terms of how you're usingmarket data only with your HR or
(20:23):
senior management to managers?
Do you communicate to everybody?Are you in the move towards
transparency bucket or maybeyou're not transparent yet? Or
you maybe you potentially don'tknow. So we're going to let you
vote away there. It's reallyinteresting for us to see how
you're thinking and how this isevolving.
(20:45):
It's important for us, Sarah, Iguess, as we think about
supporting organisations ingiving them the data sources
that we do and the choices thatthey're making there.
Speaker 2 (20:58):
Absolutely. We work with customers that are in many
different places on the spectrumand understanding what the pain
points are and what they needfrom the software and the data
to to help push in the directionthat that's aligned with their
goals is really helpful. Sothese kinds of polls and talking
to customers and prospects tounderstand these pain points are
(21:21):
very valuable information forus.
Speaker 1 (21:25):
Okay, so I could see people voting away, but I think
we're leveling off. Yep, Amber'ssharing the results. Thank you,
Amber in the background. Right,so this is probably in line with
where we were talking aboutwhere people were on those
continuums. So we have themajority of you are really
communication to HR or seniormanagement only, but there is a
good body of you, 19 and a halfpercent of you nearly who want
(21:48):
to move towards transparency.
And we see a number of you whoare communicating to managers
only, because we said, know,they're really important
participants in this process,but a good number of you, 14%
communicating to all within theorganization. So some of you are
on your journey. Some of you areprobably in that middle part of
(22:08):
the journey that we, that weshared earlier. Okay, now
another question we often getasked is around data bias,
because there's this common mythSarah that there is one market
number out there and there isone data source that you can go
to and it's going to give youthat magic number and that is
(22:29):
going to be the right number ofwhat someone would be paid. And
I love the way you explain databias and why that is not the
case.
So I'm passing it over to you.
Speaker 2 (22:38):
All right. The unicorn myth of having the exact
answer is just there if youcould only find it. I don't mean
to be inflammatory with this,but all market data is biased.
Everything that you'reconsidering, buying, purchasing,
looking up on the internet hassome form of bias. And I'll talk
you through some of the morecommon ones and talk you through
(23:00):
how to handle that bias and whento think about it as a big deal
and when to just sort of knowthat it's there.
Starting at the top end of whatyou might pay for, surveys,
compensation management surveys,and data sources like Peer are
generally collected from bigcompanies. This is because
(23:21):
bigger companies are likely toparticipate in these surveys and
pay for these expensive dataassets. And I'll show you
exactly what that SKU looks likein the next slide and talk to
you about what that means. Otherdata sources that you might be
using, like PayScale's employeereported data or information
from Glassdoor or Levels. Fyi,tends to skew towards younger,
(23:47):
more technical respondents.
There you see more 25 year oldsoftware engineers taking out
those surveys, filling out thatinformation than you do 70 year
old accountants and more thanyou do blue collar worker
working in a union job. So thedata that comes out of that is
(24:07):
aggregated and massaged, but itrepresents that younger, more
technical population moreheavily than it does some of the
other populations that you mightbe hiring for. Interestingly, as
more and more states have addedjob postings, job pay
transparency requirements andlegislation, and more and more
(24:29):
job postings have payinformation, I've been watching
the aggregate job posting datashift a little bit here and
there. As states that generallyhave a lower cost of labor add
pay transparency legislation, Isee the market range at the
national level drop for a job.And then as a state where the
cost of labor is very high, astransparency legislation, I see
(24:52):
the market range for thenational level go up a little
bit because the underlying datamix has shifted as those
different states change theirlaws and as different companies
are then adding data into thataggregate data pool.
So big plug for the Bureau ofLabor Statistics here. They do a
really, really great, verystatistically sound job of
(25:16):
evening out both labor pooldistribution data as well as pay
data. So if you're not checkingthe Bureau of Labor Statistics
website every once in a while tosee what their take is, I
absolutely recommend it becauseit'll give you a much cleaner,
unbiased view of labor in thiscountry, at least, and of pay
(25:40):
across different jobs indifferent professions, at
different industries andlocations. The downside is it's
pretty tough to build acompensation strategy and plan
based only on the Bureau ofLabor Statistics data because
it's not granular enough andit's too old to be super
relevant. But it's a reallygood, free, easily accessible
(26:00):
reference point.
So just to pull the covers backon this a little bit, I'll show
you some of the biases in PEER.And I'm using PEER here, one of
PayScale's data sets, as arepresentative of a
participation based aggregateddata set. So this is my guess is
(26:21):
this is a lot like what the databehind a third party survey
would look like. I happen tohave access to this one, so I
want to show you here, but applythis thinking to any of the
large survey providers thatyou're considering publishing.
On the x axis here, what I haveare different size bins.
(26:43):
So each of these bins representscompanies of particular size. On
the left hand are companies thathave less than five employees,
and on the right side of thisaxis is companies that have at
least a thousand employees. Whatthe y axis is showing is how
much of the workforce isrepresented by people working at
(27:07):
that size company. So Bureau ofLabor Statistics is medium blue.
That's the one I trust.
What you can see here is thatabout 8% of The US workforce
works at a really small company,less than five employees. They
could be self employed. Theycould work with two people at a
very small firm. And about 18%of The US workforce works at a
(27:32):
company that has a thousand ormore employees. So there's kind
of this dip in the middle thatshows the trend over various
company sizes.
But you see quite a few peoplework at very small companies and
less than 20% work at thesereally large firms. When we look
(27:55):
at peer, and again, I'm usingpeer as a representative here,
we see that more than 90% of theincumbents that are included in
the peer data set work at bigcompanies. So we are
overrepresenting data from verylarge companies here compared to
The US labor force in general.The upside is if you work at a
(28:19):
company that has a thousand ormore employees, the data that's
in PEER or a third party surveythat's collected in similar way
is very representative of whatpay is at those larger firms.
The downside is if you work at asmall firm, if you work at a 20
person company, using those dataassets and data sources may not
(28:42):
give you a very clean reflectionof what pay would look like at
that smaller company.
And you may need to go look atBureau of Labor Statistics or
other places to understand howthat large firm pay relates to
what it would look like at asmaller organization that you're
representing.
Speaker 1 (29:05):
Oh, sorry, sorry. I
Speaker 2 (29:11):
wanted to point this out and I won't go into as much
depth, but like other surveys,peer is also biased towards
particular industries. And thiscan be very meaningful to folks,
and sometimes it's not. But ifyou're using an all, all, all
(29:31):
data cut from peer or from adifferent survey provider, it's
important to understand whattypes of incumbents and
organizations are making up thatall, all, all data cut. For a
job like a compensation analyst,and we probably have some on the
phone, you may already knowthis, but compensation analysts
make a little bit less incolleges and universities and
(29:54):
slightly more in health care,But the ranges are quite
overlapping and really industrydoesn't impact the pay of a
compensation analyst very much.We see compensation analysts get
hired from one industry into aposition in another industry and
then move back to a thirdindustry all the time.
The talent pool has a skill setthat's useful to lots of
(30:14):
different companies. On theother hand, a job like a project
manager makes significantly moremoney in construction or in an
architecture firm than they doin finance. It's almost like
it's a different job, but it'scalled the same thing. Like 40%
more if you work in architecturethan in finance as a project
manager. So if you are pricingjobs and you pick up a survey
(30:37):
that is mostly architecture andconstruction, but you're trying
to price jobs in finance, youcould end up with a view of a
project manager role that saysit should be paid significantly
more than what the market rateactually is in finance, even
though it kind of looks okay forComp Analysts.
So something to be aware of andsomething that we can tell you
(31:02):
how to handle. But basicallyevery data source you get will
have a bias like this, wheresome industries are more
represented and some are lessrepresented. So what you're
Speaker 1 (31:14):
doing is to make sure that you discuss with your
vendors on that. So you're goingto talk us through this, I
think.
Speaker 2 (31:20):
Absolutely. Just ask ask your vendors. I'll I'll send
you that file if you wanna seeit for peer, and understand
where those those data arecoming from. And my guidance
here is if you're using an allall cut, which I love that
there's lots of data in there,that's a robust thing to do. If
the bias of the underlyingdataset doesn't match the
(31:44):
firmographic details of yourorganization, you could have a
problem.
The way to combat that problemis to be more specific, to use a
specific scope or data cut thatis oriented towards firms like
you. So even if you're using adata set like Peer and you know
it's got a lot of healthcare andretail in it, if you're in
(32:05):
materials and mining or you'rein finance, trimming down that
list and looking specifically atorganizations that are like you
will help make sure that thebias of the underlying data set
isn't impacting your marketranges and sending you often
into a range that's really notappropriate to the labor pool
(32:26):
that you're hiring from.Occasionally, there's not data
available for the industry thatyou're really interested in. And
so what we see folks do there isuse some kind of adjustment
factor to say, well, I'm tryingto price a dentist in the
construction industry. There'sno other dentist in the
(32:48):
construction industry, but Iknow that dentists in the
healthcare industry make alittle bit more than they do in
other industries.
There are actually dentists inother industries, even though it
doesn't sound like there wouldbe. They're dentists and
insurance. And I know that theymake a little bit less. So I
might take that overall numberand say, I think it's probably
(33:08):
less than health care because myindustry is likely to pull more
from talent that would also workin the insurance industry or
similar than it is to pulldentists that work in the health
care industry. So you can makesome adjustments.
And comp pros that have beendoing this for a long time often
have a back of the envelope wayof doing this for lots of
(33:29):
different scenarios.
Speaker 1 (33:34):
Okay, so we've covered data biases. We'd like
to hear if this is somethingthat you've been thinking about
in terms of your data strategy.So we have another poll going in
the polls tab. Do you monitoryour data processes for bias? Is
this something you docontinuously, occasionally,
you're planning to, you weren'taware of the issue and hopefully
(33:55):
you are now after Sarah's greatexplanation or you don't know.
So we're just interested to seeif this is something that
resonates with you and somethingthat you're thinking about as
you work with different datasources. And I can see you're
busy voting away there and I cansee some questions coming in. So
we are going to move to the liveQ and A soon. So any questions
(34:17):
you've got specifically aboutthe biases that we've talked
about so far about transparency,get those into the Q and A and
Sarah and I will be able toaddress those very shortly. We
have one more topic to coverbefore we move into the live Q
and A.
But let's have a look at whatyou are saying about do you
monitor your data processes forbias?' So occasionally, so we've
(34:40):
got nearly 38.5% of peoplesaying they occasionally monitor
and we've got 20% of you who aresaying, nearly 20% of you saying
continuously and then some ofyou who are planning to do it.
So it's good to see, I thinkwe've got a group of mature
compensation practitioners hereon the line. This is something
that's a topic that you're awareof. Hopefully this is just a
(35:02):
good refresher on why it'simportant to think about those
biases as you work with yourdata sources. So thank you very
much for responding to thatpoll.
Okay, so you can't talk aboutanything to do with HR or
anything to do with anythingtoday without talking about AI,
to be honest. And we've beentalking about AI as it relates
(35:24):
to data and data strategy, aswe've done this webinar series
over the last this is the thirdyear we've kind of run it. And
there's definitely been a hugemove in terms of progress, I
think, Sarah. We've beenthinking about it over that time
very actively. We've beenthinking about it in terms of
how we incorporate it into thedata solutions that we have here
(35:47):
at Payscale.
But Sarah is going kind of talkyou through where we're at today
in terms of how you should thinkabout AI as part of your data
strategy.
Speaker 2 (35:58):
Yeah, as Ruth said, although I've been doing AI
since before it was cool andPayScale's been doing AI since
before it was cool, 2023 was theyear that generative AI exploded
into the public discourse. Andin 2024, we saw the use of AI in
HR kind of pop up and start togo more mainstream. This year,
(36:21):
HR professionals are embracingAI, particularly for tasks like
compensation management and jobdescription creation. As part of
an annual poll that we do calledCompensation Best Practices
Report, which you can get to viathis link, we gather responses
from 3,600 comp and HRprofessionals a mixture of
(36:42):
PayScale customers and alsofolks that don't use our tools.
This year, we found that 71% ofthose respondents have a
positive sentiment about usingAI for market pricing.
And similarly, a growingmajority of respondents are
interested in AI for other usecases, like recommending pay
increases, helping supportlegislative compliance, and for
(37:06):
documentation andcommunications. In the
compensation world, as I alludedto a little bit earlier with my
back of the envelopecalculations, data driven
techniques to extrapolate fromexisting data are our bread and
butter. Many of the methods thatare common in our field, taking
a weighted combinations ofunderlying data sources or aging
(37:28):
the data or calculating andusing differentials, are really
just simple implementations ofAI. Even doing a Google search
is basically using AI. So in thenow and certainly in the future,
I expect that AI tools, enabledtools will make these methods
that we have been using foryears more robust, more
(37:52):
scalable, more repeatable, andmore accessible to folks that
don't have years and years ofexpertise and know the data
inside and out.
With tools like Explore, whichwe launched this week, and the
underlying AI powered dataasset, we're using AI to find
the same patterns that we've allbeen measuring and using as part
(38:13):
of the art of compensation. Butleveraging AI and the data at
scale to do it instead of doingit one off each with their own
methodology. This helps uscorrect the biases in the
underlying data sources, and itimproves the accuracy of
calculating a pay market rangewhen the data are sparse. So if
(38:34):
you want to hear more aboutthis, connect with me on
LinkedIn or schedule a demo andwe'll show you what those tools
look like. The same techniquesalso help us discover some
patterns that are prettyinteresting and insights about
pay and compensation strategythat haven't been measurable
before.
(38:54):
So compensation decision makers,whether that's a compensation
analyst, an HR generalist, a VP,an exec, or a board need
education on how artificialintelligence can make
compensation related decisionsmore accurate and more
efficient. AI can assistemployers with job management,
(39:15):
including writing and analyzingjob descriptions. And AI is also
being currently used and moretools coming for market pricing
and pay analysis. Over the nextcouple of years, all of these
boxes that don't have stars, Iexpect to see more and more
areas where AI makes ourexisting workflows more
efficient and helps us uncoverkey insights that can drive our
(39:39):
compensation strategies. I willsay the potential for AI to
extend bias rather than mitigateit is still a very real concern,
especially where AI canunknowingly perpetuate
historical salary discriminationand pay disparities for
protected classes.
(39:59):
Ultimately, human interventionis required to ensure that AI
powered algorithms are properlyinformed and producing fair
outputs. I like to say that AIcan scale bias or reduce bias at
scale, But our judgment and ourexpertise helps make sure that
we push towards that reducedbias at scale rather than the
(40:22):
scale bias. I also want to flagfor folks that are doing this
work day in and day out, paygaps are not excused by the
reliance on flawed models. If anemployee in this country, if an
employee can demonstrate thatthey are paid differently from
another employee for equal orsimilar work, the burden shifts
(40:43):
to the employer to demonstratethat that pay differential is
due to a permissible reason,like differences in skills or
differences in performance ordifferences in location or
responsibility, rather thanconscious or unconscious bias or
discrimination, even if an AItool made that recommendation.
(41:04):
So employers, not the tool, arelegally responsible at this
time.
Speaker 1 (41:12):
Great, yeah, thanks for calling that out, Sarah.
Something I was going to pick upon. So I mean, for me, as I
think about it as a compensationpractitioner, there are more
data sources available today andthe level of compensation
decisions that we're needing tomake are more granular. When I
think about the whole skillsconversation and how that plays
(41:32):
into this, really, you know,having advanced decision support
was the only way we were goingto get to the place where we
could start to correlate allthose different information
sources and information datapoints. We do need to get
comfortable as practitioners interms of thinking about how AI
can support us in that processand can help to improve the
(41:54):
decision making that we aredoing whilst being aware all the
time about what type of AI isbeing used.
So if you're talking to a vendorand they're talking about AI,
ask them how it's beingdeployed, make sure you
understand how it's being usedand in what way so that you can
make an informed decision andfeel comfortable about that as
(42:15):
you move forward. Okay, so lastpoll before our Q and A poll.
How are you using AI today? Weknow there's quite a few
different use cases that peoplehave. We've surveyed on this
before in our compensation bestpractice report, but interested
to see how prevalent the use ofAI is in terms of the work that
(42:36):
you're doing today.
So I'm not going to read out allthe options there. If you want
to vote away, I think you canselect all that apply in this
particular poll. But are youusing it in benchmarking,
monitoring for pay equity,collecting intelligence on
skills for recruiting, writingoffer letters, using it to speed
up survey participation, thatwould be nice and year over year
(42:59):
updates. How are you using ittoday? And then we will move in
to the much promised Q and A.
I can see a few questions inthere. So we'll be getting to
those very shortly. Okay. Are welevelling out on votes? Okay,
(43:31):
none of the above or unsureseems to be the majority at the
moment.
Of course, that may be theunsure category. So there has
historically been a lot of AI.It was just probably just called
machine learning at that pointin market data sources. So I'm
pretty sure that if you're usingany form of market data sources,
it's got some form of AI inthere. And it just, you know,
(43:55):
you may not just know it underthe term AI.
We can see people are also usingAI to collect intelligence on
skills for recruiting,upskilling, career pathing and
or compensation. So there wasnearly quarter of you using it
for that. And then some of youusing AI to benchmark, about
(44:15):
just over 10%, nearly 12% of youusing AI to benchmark and price
jobs or predict pay ranges. Soclearly an area where this is, I
think we're going to see thegrowing influence of AI, Sarah.
Don't know how do you feel whenyou look at those poll results?
Speaker 2 (44:34):
It's interesting, you know, and I think if I pulled my
team, which is all people whobuild AI tools, we would get a
similar distribution of placeswhere people are seeing the
value already and places whereit's still kind of hype and
like, I don't know. And then asyou mentioned, Ruth, places
where the old fashioned ways areactually not very dissimilar
(44:58):
from what we think of as AI. Wetalk about pay equity, for
instance, that's been based onregressions for many years.
Regressions fall into themachine learning and AI umbrella
because it's predicting thingsgiven data. So this is really
interesting to me.
And certainly the collectingintelligence on skills, career
(45:21):
pathing and compensation is anarea where AI, especially the
next generation text processingfrom large language models can
really make a big difference inthe kind of information that is
available to us. So fascinatingto me to see the spread and the
(45:42):
different places where peopleare interested or already using
AI tools.
Speaker 1 (45:49):
So yeah, thank you for sharing all your insights,
everybody, through those threepolls. So closing us out,
really, you know, the keymessages that we wanted to get
across today is, you you need adata strategy. And the aim of
that data strategy is really togo for consistency of
methodology so that you have arepeatable process that makes
(46:11):
sense for your business. Aimingfor transparency becoming
increasingly important, youknow, you're going to have more
demands on you to be able toexplain how you determining each
employee's compensation and howthat number was derived either
from managers who are havingconversations with employees or
(46:31):
directly with employeesthemselves. So you think you
getting comfortable with howtransparent you want to be about
the type of data sources thatyou're using as a journey I
think you need to go through.
Do you want to cover off thelast two, Sarah?
Speaker 2 (46:46):
My goal today was to flag the biases issue and to
show you that it's notnecessarily a bad word, that all
of the data assets that youconsider are biased in some way,
and that there are things thatyou can do to mitigate that bias
and to work with it rather thanagainst it. And then to
(47:07):
highlight a couple of key placeswhere we're seeing AI deliver
value to practitioners like youand have you listen to each
other a little bit through thatpoll to get an idea of where
those tools are moving. Ifyou're interested in hearing
more about AI tools, drop a notein the comments. We could do
(47:27):
another webinar that's morefocused on that, either from a
market landscape or thoughtleadership perspective.
Speaker 1 (47:37):
Okay, all right. Thank you. Right, we're going to
move into Q and A now. There isanother poll running in the
background. If you would like tofind out more about how Payscale
can help you address some of theissues that Sarah and I have
talked today, learn more aboutthe AI innovation journey that
we're on here at Payscale, thenif you click in that poll,
(48:00):
someone will reach out to you toget back to you to help you
understand that.
We're going to jump into the Qand A. I'm going to go back to
one of the first questions thatwas asked of us, which was when
we were really talking abouttransparency. And it says: Do
you recommend employers having awritten codified comp data
(48:20):
strategy versus just somethingyou used in practice? Classic
consulting answer would be itdepends and I think it really
does depend. I'd say thatdepends on the size of your
organization.
If you have many compensationpeople and HR people kind of
working on this together acrossthe organisation, then quite
often having some kind ofwritten data strategy that
(48:44):
people can refer to would beuseful. If you're just like a
sole person on your own andyou're doing this as the comp
analyst on your own and you'reworking with your HR team,
probably useful to train them onit. But do you necessarily need
something in writing? Probablynot, I would say. So it really
kind of depends on how manypeople you're trying to ensure
(49:04):
that consistency applies to.
It's like what is right for youto ensure you're applying the
consistency and the decisionsthat you've made around how
you're going to approach data.Now Sarah, there are a couple of
questions about adjustments inthere.
Speaker 2 (49:21):
Yeah.
Speaker 1 (49:21):
Do you want to take those two and talk to those two?
Speaker 2 (49:24):
I will. So we got two questions that are related. One
was, do you have anyrecommendations for selecting
the appropriate adjustmentpercentage for difference in
industry? And the other was,what advice would you give for
making adjustments where thereisn't applicable industry for
those who are newer to comp? I'mgonna give you two answers.
(49:46):
One is go get a demo of Explorebecause we take the hard part
out of it for you. And we've gotlots of experts that do this at
a job by job level so that youdon't have to be an expert to
understand how to do it. ButI'll tell you how to do it
yourself, too. The biggest pieceof advice I can give you here is
(50:09):
don't think about the industrythat you're working in for a
particular role. Think about theindustry or industries that
you're hiring from.
So if you're hiring, accountantsare a good example, cop analysts
are a good example. If you lookat your books and you say, Okay,
my employees that are in thisrole came from a large number of
(50:30):
different industries. And when Ipost this role, the next few
people I'm hiring could alsocome from education or
construction or retail ortechnology. Then in that case,
an all, all, all data cut, noindustry differential is going
to give you the bestrepresentation of what pay is
(50:52):
for that role, because it's notvery dependent on which industry
you're in. On the other hand, ifyou are pricing jobs or a family
of jobs within your organisationwhere you are consistently
pulling from one industry thatis where you're always hiring
from that industry and you wouldnever consider hiring from a job
(51:15):
outside of that industry.
Then there's kind of twobranches in this decision tree.
One is to think about whetherthe job itself is particular to
the industry that you're hiringin. So if you're working in
agriculture and you're hiring aagricultural laborer, all of the
data from any survey that showsyou what the pay is for
(51:38):
agricultural laborers isbasically going to come from
agriculture. It's not going tocome from other industries.
Speaker 1 (51:53):
I've lost your sound, Sarah, I think. Yes. I'll just
check whether it was me or you.
Speaker 2 (51:59):
Can you hear me now?
Speaker 1 (52:00):
Yes, you're back.
Speaker 2 (52:02):
Okay, great. I have a little bloopity bloop. So in
that case, the job itself isrepresenting the pay in the
industry because it's specificto the industry. The trickiest
part is when you have a jobthat's represented across many
industries and the pay variesacross those industries, like a
(52:23):
project manager that I showedbefore, where they make 40% more
in construction than they do infinance. In that case, what
people generally do, and this iswhy I'm so excited about
Explore, but what peoplegenerally do is they look at
average differentials acrosslots of jobs and say, well,
generally jobs pay a little bitmore in this industry than they
(52:47):
do in this industry.
So if I'm pricing in finance, Imight pay 8% more than in the
construction industry becausemost jobs pay a little bit more
in finance. That doesn't workperfectly, as this example
shows, because some jobs paymore in one industry and less in
another industry. But that'sgenerally what the best practice
(53:11):
has been until the tools havegotten better and caught up with
being able to learn more jobspecific information about how
industry affects pay. So let metry to sum it up. I know that
was a lot of information.
One is in many, many cases, youdon't need to worry about
industry at all because you'repulling talent from across
(53:34):
industries and you're basicallypaying a market rate that is not
dependent on an industry. That'scase one. That's probably 70 to
80% of the jobs that you'repricing. Case two is you're
pricing a job that's only foundin a couple of industries and
all of the market data thatyou're looking at is coming from
those industries. They don'tneed a differential.
(53:56):
That's probably 10%. So nowwe're at like 80 to 90% of the
data, no differential. That last10 is the hard one where the pay
varies across industries and thedata that you have isn't totally
represented by the industry thatyou're pulling from. And then
(54:18):
that's where the art part comesin. And I think AI can make a
really, really big difference.
But the general practice overthe last several years has been
to look at what I call a jobagnostic differential of how
does pay generally differbetween these two industries.
You can get that from a dataasset like PayScale has, or you
(54:39):
can look it up on BLS and youcan see how pay varies across
industries.
Speaker 1 (54:45):
Okay, thank you for that answer to those two
questions, Sarah. I think maybewe've got time for one more
question. Someone was callingout that they are having to
start from scratch at theircompany to figure out pay
ranges, but they don't have themoney to potentially buy pay
scale data sources. What's theirrecommendation on how they can
begin? Well, the good news isthat we have a new freemium
(55:07):
offering that is going to belaunched shortly, that will be
able to give you access to thatnew explorer experience that
Sarah has been talking abouttoday.
If you want to find out moreabout that again, click into the
poll or reach out to us and wecan make sure we can fill you in
about that. So that is one way,depending on the number of jobs
that you've got to price,obviously, you know, there are
(55:29):
freemium options around fromsome vendors. Sarah, you talked
about other free data sourcespotentially like the BLS data.
You know, are other data sourcesthat you can use as well. But
I'd also think about thispotentially, you know, depending
on the number of jobs thatyou've got to price within your
organization from an ROIperspective.
(55:49):
Because quite often people go, Ican't afford to buy pay data.
When you need to think aboutactually what is the impact of
not using pay data and the costto your business potentially of
not doing that. You price fivejobs wrong and that could
actually aggregate up to a valueof $2,030,000 US dollars that
you're making a mistake on andeven more potentially. So, you
(56:12):
know, think about framing itfrom the perspective of the ROI
perspective to yourorganization. And particularly
when we're in the slightlycautious macroeconomic
environment that we're in at themoment, where we know many
organizations are highly costsensitive, You know, the right
pay decision made that isaccurate and near term real term
(56:37):
data, know, that's reallyimportant.
So think about how you framethat from the kind of ROI
perspective to your business.And with that, I think we are at
time. So I'm going to ask Amberto come back on stage and close
us out. But thank you all verymuch for your time today. And
thank you, Sarah, for sharingyour insights.
Speaker 2 (56:54):
Thanks for having me, Ruth.
Speaker 3 (56:57):
Yes, thank you both so much for sharing all of your
insights and expertise on thistopic. It was a wonderful way to
kick off our data series. Youwill we'll be doing a few more
of these events on this topic,covering into different areas
over the next couple of months,so stay tuned. Okay. The session
(57:17):
today comes in around recordingand slides.
We will send both the recordingand slides in one to two
business days, as well as someof the follow-up materials that
Ruth discussed and someadditional ones. So please keep
your eye out for those. Withthat being said, we're out of
time. So I will wish you all awonderful rest of your day, and
(57:41):
we hope to see you at a futureevent soon. Thanks, everyone.
Speaker 1 (57:46):
Thank you.
Popular Podcasts
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Therapy Gecko
An unlicensed lizard psychologist travels the universe talking to strangers about absolutely nothing. TO CALL THE GECKO: follow me on https://www.twitch.tv/lyleforever to get a notification for when I am taking calls. I am usually live Mondays, Wednesdays, and Fridays but lately a lot of other times too. I am a gecko.
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.