October 3, 2023 • 49 mins
In this episode, Heather Holst-Knudsen, CEO of H2K Labs, and Chad Rose, CEO of InsightOut, dive deep into the world of predictive analytics for B2B media companies. They cut through the hype to explore realistic opportunities, potential pitfalls, and best practices for implementing predictive analytics to drive revenue growth and mitigate risks. Whether you're just starting your data journey or looking to refine your existing analytics strategy, this episode offers valuable insights and practical advice for leveraging the power of predictive analytics in your business.
Key Topics Covered:
1. Introduction to predictive analytics and its relevance for B2B media companies
2. Unpacking the hype: Realistic expectations and opportunities
3. How predictive analytics can drive revenue growth and risk mitigation
4. Use cases for predictive analytics in media and events businesses:
- Customer churn risk prediction
- Sales forecasting
- Content optimization
- Pricing optimization
- New revenue stream identification
5. The process of implementing predictive analytics:
- Defining objectives and assembling the right team
- Data collection, preparation, and tool selection
- Model development, training, and validation
- Iterative improvement and scaling
6. Common pitfalls and "landmines" to avoid
7. The importance of data democratization and change management
Key Takeaways:
- Predictive analytics can significantly improve decision-making, operational efficiency, and revenue growth
- Start with small, focused projects to prove value before scaling
- Data quality and quantity are crucial for successful predictive models
- Proper communication and education are essential for adoption across the organization
- Predictive analytics is an iterative process that requires ongoing refinement
The podcast emphasizes the importance of having a clear objective, assembling the right team, and focusing on data preparation before diving into advanced analytics. It also stresses the need for a cultural shift towards data-driven decision-making and the value of democratizing data access across the organization.
Interested in joining Revenue Room™ Connect, the first C-Suite network for CEOs and their revenue-critical C-Suite teams in media, events, data/information, and marketplace sectors? We focus on turbo-charging enterprise value using data, digital, and AI.
About Heather Holst-Knudsen
Heather Holst-Knudsen is the founder and CEO of H2K Labs and Revenue Room™ Connect. She is a seasoned executive with extensive experience in digital transformation, data, and revenue growth. She is a recognized leader and operator in media, marketplaces, events, and adjacent technologies. Heather has a proven track record of leading organizations to achieve customer-centric innovation, revenue growth, and enterprise value creation. As a thought leader, Heather shares her insights on multisided business models under The Revenue Room™. Connect with Heather on LinkedIn.
ABOUT H2K LABS
H2K Labs is a tech-enabled value creation specialist that helps media, data/information, event, and marketplace businesses accelerate revenue, drive profitability, and fuel enterprise value using data, digital, and AI. We host The Revenue Room™ Podcast, curate Revenue Room™ Connect, a professional network for CEO and their revenue-critical C-Suite teams, and produce events including RevvedUP 2025. For more information, please visit https://www.h2klabs.com
To join Revenue Room™ Connect, please visi
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
Welcome to the Revenue Room,presented by H2K Labs.
(00:05):
Here's your host, HeatherHolst-Knudsen.
well, welcome everybody to theRevenue Room Bootcamp, this is
an event that is hosted by H2KLabs and, my pal over here, Chad
Rose from Inside Out.
What we try to do is provide acombination of data and business
(00:27):
insights to help you drive datafueled revenue growth.
So today, we'll kick off and gothrough, a few areas.
One is unpacking the hype.
Every, everyone's talking aboutpredictive analytics, but you
know, there's a lot that goesinto it.
What's realistic, what's not theopportunities, both in term of
(00:47):
identifying risk and alsorevenue opportunity, the process
that you need to go by and alsolandmines to avoid.
So I'm Heather Holtz Knudsen.
I'm the CEO of H2K Labs.
I am a veteran in the B2B mediaand events and digital
information space.
(01:07):
Chad is the CEO of InsideOut.
And Chad, want to give a realquick background?
Yeah, thank you, Heather.
So my background is primarily indata engineering and analytics.
So I've been doing that sincefor my entire career and helping
companies specifically withinthe middle market and enterprise
space to, make the most out ofdata, get the most out of the
(01:28):
data.
Great.
And what we do at H2K labs is wehelp in three ways.
We help through consultingacross data, revenue and product
strategy.
We also help with planning andexecution through workshops and
capability audits.
And we have two platforms, oneof which we've partnered with
chat on called insightify, whichis a modern data management and
(01:50):
predictive analytics platformwith very industry specific
capabilities for media events,digital information and another
solution called channel metrics,which helps on the customer
delivery side.
So to get the most out of theboot camp, if you feel up to it,
put your camera on.
If you have questions please usethe comment section and we'll
(02:11):
stop.
It, this is meant to be adiscussion forum and it says
polling.
We don't have polling for thisone but after the boot camp,
you're also going to get somegreat information, including a
playbook, the deck, and a linkto the recording, which you can
share with your peers.
So the focus is really on data.
If you'll revenue andprofitability when we talk about
data, it really is on therevenue side.
(02:34):
There are many people out therealready handling audience.
We connect with audience when wetalk about things, but we're
really about the commercial sideof the equation and our content
while free.
It is highly specialized andthere's a lot of depth to it.
So we hope that you do walkaway.
Thank you.
Either finding out how to figuresomething you didn't know, or
(02:56):
you didn't know something andnow you do.
And then after the bootcamp, asyou a series of content items
that can help you.
And also we're offering officehours an hour to dive deeper
into some of the things wetalked about to help you as you
build out your strategy.
So let's talk about the hype.
(03:17):
So what are predictiveanalytics?
It's the way you're taking yourdata past and current and
blending that together with newinsights, utilizing structured
and unstructured formats andapplying statistical and machine
learning to predict futureoutcomes.
It's not just to predict, it'sactually to activate.
Like, how do we change thecourse of something if it's
(03:40):
Looking not so great or how dowe activate and leverage an
opportunity if it has if thathas come to surface and
predictive analytics our meansit's a way to get to a better
business outcome, a leaner, moreproactive company, better and
greater transparency andaccountability into processes
and outcomes and deeper insightsinto unmet customer needs to
(04:02):
gain market share.
But Chad, did I miss anything onthis slide?
Do you think?
No, I think you got it.
I think on the hype side, it'sjust one thing to add there.
Oftentimes folks think that theycan apply predictive analytics
across the entire organizationin every domain, every
department.
As you'll see, as we go throughthis, there are some precursors,
there's some requirements, so tospeak, that make it more
(04:25):
feasible in some areas thanothers.
And so You know, to your pointhere, Heather, it's really there
are means that it's not the endall, be all at for every part of
the company and every person inthe company agree 100%.
And the other point I like tomake is because this is a real.
Real big issue, especially forlike CEOs and CFOs is, we're
(04:48):
investing in data, but We don'tsee it.
What's the ROI you're investingin your business and data is the
enabler of those investments soyou need to look at it across
things like Building data drivendecision making across the
organization How do I make myteam?
And every function that requiresthis able to use data to improve
(05:08):
how they perform and how thebusiness performs.
How am I using data to growrevenue?
I always like to say there's nota CRO, CFO, or CEO I've ever
spoken to that says, we'rereally satisfied with the
revenue right now and we don'tneed to do anything.
No it's an investment infiguring out growth, innovative
growth, and it's a way to gainoperational efficiency in smart
(05:29):
ways.
Like where can we look at areas?
to recalibrate resources sothey're applied to the most
profitable high growth areasversus the not so high growth
areas.
And frankly, from, an investmentstandpoint or you want to be
acquired, it's a way to increaseyour multiple your valuation,
(05:49):
your acquisition attractivenessand the speed.
to which this, an acquisitiontakes place.
And it is definitely an improvedboard and investor.
It improves the way youcommunicate with your board and
investors.
I was speaking to another CRO inB2B media.
And one of the big things theysaid is, they're owned by
private equity is theirinvestors are asking for data.
(06:09):
More frequently, more granularand more like on things that
literally it's it takes themweeks to pull it together.
By taking this approach, it cangreatly enhance that
communication, which had on thisarea.
I know you deal a lot withprivate equity, for example.
Are there other areas that youthink that impact the business
when you invest in data?
(06:30):
I think you've covered some ofthe key ones here.
Heather double down on theimpact it has on the team.
And so as you as if you do thiscorrectly and you're able to
embed predictive analytics intothe business the team will
largely or oftentimes kind oftake take that on.
And what I mean by that isthey'll be more interested in
(06:50):
pursuing.
Other areas of their day to daywhere they might see a value in
adding additional predictiveanalytics.
So they might even, research howto do it themselves or create
models themselves.
So as soon as you start to embedthis within the business, the
team members do take a biginterest and take ownership of
it.
And that's in the best casescenario.
(07:11):
And so I think that's a reallyimportant point.
Small wins spark the appetite,right?
Another area that we see whenworking with compliance is that,
looking, taking a huge bite ortrying to eat the whole cake.
It's really start small, savorand really learn what the
ingredients are before you moveon to the next one.
(07:31):
But small wins will not onlyenable a much faster, better
path to R Y.
It actually will prove the valueand which will enable it to
scale.
And there is, there's definitelywe always see low hanging fruit
that you can go after first,whether it's in a very specific
division like the financedepartment or it could be sales.
(07:52):
But again, small, quick wins arethe way to go.
And then it's a journey, right?
So you have to start somewhere.
And I guess the question that Iwould ask everyone is, Is what
I'm doing today.
Okay.
Can I continue to make mybusiness grow the way I needed
to grow just doing what we'redoing?
Or do I have to take this on andI need to run my business
(08:14):
better.
I have to become data driven.
And so where do I start?
And then finally, it takes avillage.
It's, and I'm going to actually,Chad, because you deal with this
so much is there so much,there's a process which we'll
dive into later into thisdiscussion.
Bye.
Bye.
Bye.
And all of these elements areabsolutely critical.
(08:37):
What can you talk, tell, talk tous about what you see when you
look at the, how important thisprocess is.
Yeah.
I mean, in terms of, thismessage here, it really starts
at the executive level, andhaving sponsorship there and
taking on the challenge ofreally becoming more data
focused and oriented.
And then, you have.
(08:57):
Impacts on the organization inthe source systems, like your
CRM, where you're going to, ifyou have predictive models that
are built on top of the datathat you're collecting, there's
a dependency on the people whoare putting the data in there.
And so in the way that thesystems are configured, and so
there are elements across thebusiness that start to impact
the outcomes of the analytics.
(09:18):
And so it does hit a lot ofpeople in a lot of different
areas as you start to reallyoperationalize it.
Absolutely.
Okay.
Does anyone want to, talk aboutwhat your, what are your main
business objectives when youthink about predictive
analytics?
Maybe post it in the commentsor, and this will help us make
(09:38):
sure that we're talking aboutthings that and hitting squarely
on some of your top priorities.
Feel free to throw that in hereand Chad will be watching as I
go through so let's talk aboutcapabilities.
I took this course and Irecommend anyone who's
interested in data monetizationor how to improve your business
and predictive analytics fallsunder data monetization.
(09:59):
It's by MIT Sizer.
It's fabulous.
It is it really helps with,organizing the thought process
and how you approach it.
But there are two really bigoutcomes that I took from this.
And that is you need to havedata capabilities that map to
this to where you want to be interms of a data maturity model
and how you're activating that.
(10:20):
And there are guiding principlesthat, that underscore.
So in terms of predictiveanalytics, these are the five
areas.
If you're pursuing datamonetization, As it relates to
predictive analytics, whichfalls under the internal, right?
You're using data to add cash tothe bottom line by improving
what you're doing internally.
(10:41):
You need to have very sophist onyour data assets, single source
of truth.
It needs to be able to beblended and you need to apply
data science and the machinelearning side.
You need to be able to provideaccess to that data to your team
members, because it needs to betransparent.
So you need a platform.
I mentioned the machinelearning.
And that's not just about one ofthe big myths about predictive
(11:04):
analytics is like, Oh, I put apredictive analytics platform in
and like, let me push a button.
Predictive analytics, actually,the data needs to mature and
learn for a period of time.
So having this capability isimportant.
Thank you.
You don't need to be superadvanced on the acceptable data
you side, but you do need tohave internal oversight and
permissioning who sees what, etcetera.
(11:25):
And on customer understandingagain, not you don't need to be
out in what I call the outwardcommercialization area.
But you do need to understandwhat the data is telling you
about.
your customers.
And then the guiding principlesabout that underscore this is
that, data monetization is thedirect or indirect conversion of
data into financial games dataliquidity.
(11:48):
And this is the area where wespecialize in our businesses
that have very high dataliquidity.
But this is the ease of whichyour data can be monetized.
And in the area of events,media, business information,
marketing services you'redealing with huge amounts of
data that changes every singleday, every minute, actually
every action on your website isa new data event.
(12:10):
Being a data drivenorganization, it's an
organization that's constantlyinnovating and scaling using
data to improve businessoutcomes and this data
democratization, which anyonewho's on this call and has
spoken to me before, I'm verypassionate about is, Your whole
team needs to have access todata.
It needs to be built forbusiness users.
You are not going to be able toacquire data skills.
You have to actually build theminternally.
(12:32):
So one way to do that isdemocratizing the data.
So another question, if you wantto throw it into the chat is
areas we're going to talk aboutare like gaps that are stopping
you and based on what we justreviewed.
All right.
So this is the risk andopportunity side.
(12:54):
You have more data than you everhad before.
And again, if you're in themedia events or any type of
business that has a two sidedbusiness model, or is dealing
with multi channel marketing,you've got more data than you
ever had beyond what I call atraditional business.
It's where do you start?
And it's not just enough toacquire it.
(13:14):
You have to understand it, haveto be able to operationalize it,
right?
How am I activating behaviorsinternally?
Based on what we're learning toimpact, and I should have a
fourth hour here.
And that is how am I measuringit?
Okay.
So what Chad and I did is webroke down three ways that
(13:36):
predictive analytics can helpyour business.
There's one is growing sales.
Identifying expansionopportunities, personalizing
campaigns to have better impacton pipeline conversion,
maximizing your lifetime value,whether it's your audience or
your your exhibitor oradvertiser side reducing churn
and optimizing pricing, you canimprove or optimize performance
(13:59):
by.
Bringing the right products tomarket or sunsetting less
profitable ones.
Optimizing resource allocationbased on customer spend, not all
customers are alike and you needto, measure that.
How can I look through the data,what's working from a sales
standpoint and scale that acrossall of my sales team players so
(14:22):
that I can improve my A game andreduce my C game.
And how do I optimize contentand how we're investing in it
and how we're producing it,audience and the conversion side
to programs.
And then clearly there's therisk, deal risk, customer churn
risk.
And again, in the world of, thatwe operate in it's churn is
(14:44):
actually a very complicatedmatter because there could be a
lot of revenue reductionhappening along the way.
Well, before the churn takesplace.
So how can you utilizepredictive analytics to identify
that erosion before the actuallogo goes away payment risk?
It's another area attrition onactually, which is Repeat of the
(15:06):
employee fight flight risk andthen product adoption risk.
So we put together some usecases.
There are about six right nowthat we have in here.
And this first one is thecustomer churn risk.
This is and you could slice anddice this one in many different
ways, but in this use case,let's, it's a marketing services
company that's serving diversecustomer segments across
(15:26):
multiple brands in differentcountries.
And the channels include events,digital advertising and legion.
You could add.
More the point is that there areall these different factors that
create data complexity.
And what they're trying to do isthey want to identify red flag
alerts in time to action andfix.
Right?
Because as I mentioned, thechurn is a painful, slow trickle
(15:48):
sometimes.
But unfortunately, as I'veexperienced personally, and I
have seen with customers by thetime that red flag really gets
raised, it's too late, right?
And we're all running aroundwith our, like chickens with
their head cut off trying toimprove, but the damage has been
done.
So how this customer wants toidentify this further.
So you build a stakeholder team,you put sales, customer success,
(16:10):
operations and finance together,and you map things that are in
your business that would saythis constitutes a red flag
alert, right?
And in which in this case, CRM,we're going to map CRM data with
operational data and program.
Program performance data.
We actually also want to maporder management because we want
(16:32):
to ensure, one of the biggestissues with churn risk is the
the program doesn't deliver.
So I need to know what waspromised and what they paid for
and when it was supposed to bedelivered.
And I also want to put in someother things like third party
data.
So LinkedIn, for example andanother churn risk is if someone
departs The actual customercompany and LinkedIn, for
(16:54):
example, has a, an alert systemthat tells you if someone moved
or there's a change in LinkedInprofile.
So you then create insightsthrough the data and it could be
the program performancedashboard.
Is there if you're doing surveysat events?
the customer turnover, the thirdparty data that I just
mentioned, and also obviouslythe salesperson traction and
(17:15):
pipeline traction.
And you create dashboards basedon what you believe are these
red flag alerts.
And you also, on top of it, havean actioning process, right?
What are we going to do whenthese alerts are surfaced.
So and it's a process.
Everything that impacts thesales team, the ops team, the
customer success team and thefinance team.
(17:35):
Another data point you may wantto put in here is if the
customer typically pays on time.
If payment is delayed, thatcould be an indicator of
financial troubles.
And then you assign some ROImetrics, churn rate, net
retained revenue, customersatisfaction.
And these all go into thedashboard that you're looking at
so that you can actually notonly see what's happening and
(17:57):
action it, but is it making animpact?
Are there any questions on thisuse case or does anyone have an
example where they're doing thisright now?
Okay.
The next use case, and this isfrom an outlier, right?
This is Chad has customers withthe product we sell.
And this one I love because itis I actually can liken this to
(18:18):
acquiring attendees for an eventor selling audience
subscriptions because there'sseasonality to it.
But Chad, you want to talkthrough this use case?
Sure.
Absolutely.
So in this use case, it's reallyabout trying to get ahead of Any
sort of downturn or reduction insale and future sales.
So the idea here was to build amodel to predict out a couple of
(18:41):
weeks, a couple of months inadvance based on current year
data and prior year data, howmuch.
The business was going to do innet new sales and how much they
were going to do in same storesales year over year, which is
their primary metric of success.
Right?
And building a predictive modelhere allowed them to see out
(19:02):
into the future very accuratelywithin the first within the next
couple of weeks.
And even further out, To say ifthey see a downtick in bookings
or in potential sales, they wereable to respond with increased
promotional activity, increasedmarketing activity and other
triaging to ensure that theyactually address that shortfall
(19:24):
ahead of time.
So great example of wherepredictive analytics is really
giving them, visibility into thefuture and giving them,
actionable things that they needto do to address problems before
they, surface.
Anyone have a question on thator how that how that could apply
or translate into your business?
(19:45):
Okay, so the next use case iscontent optimization and one of
the things for that will dobetter next boot camps is we
actually got this from one ofthe participants on the call who
spoke to me about like, I'mhaving an issue with this.
Can you.
Address it.
And it's how do we and thisreally applies if you're also
(20:08):
selling content as part of asubscription, or you need to
produce content that appeals toan audience that you are
attracting for an event is ifyou have content that's on your
website.
And you're producing that allthe time and there is user
engagement and you can slice anddice that.
(20:29):
How what can we use that datafor in terms of helping us plan
for better engagement, bettercommercialization.
And better investment in thiscontent creation.
So it's a B2C media company thathas multiple brands, very large.
And they have a paid contentsubscription model.
And their objective is to reducethe hours they spend each week,
creating a scalable educationcontent packages for their paid
(20:52):
subscribers.
Every week they get down and sitdown at a table.
They're sitting there trying to,well, I think this, well, let's
do that.
So it's very anecdotal.
It's got, again, having beenthere many times that you're
wrong.
So how can you actually use datato do a few things?
One data to reinforce and thedecisions, right?
No, here's the data that'stelling me that this is or us.
(21:13):
This is what we should be doingto reduce the time that's being
spent agonizing over what tocreate.
That time could be spentcreating value.
And three being able to actuallymap this content really better
use the investment better.
Okay.
Towards what your audience istelling you they want today and
some trends of what they maywant tomorrow.
(21:33):
It's also, by the way,underlying this there, there is
actually a a new productideation component as well.
But for this particular usecase, you get the stakeholder
teams of marketing content,audience and data together.
You're you need to map websiteanalytics, own social analytics,
your C.
D.
P.
Data, C.
(21:54):
M.
S.
Data.
And really break down intoaudience core cohorts to content
segmentation and really identifythat to demographics, location,
cohorts and content type.
Also, by the way, it's not Youknow, is it product information?
Is it thought leadership?
Is it, do they like short andsweet?
Or are they, is this type ofaudience looking for more
(22:17):
detailed, bullet point type ofcontent?
So it really helps youunderstand not just the topic,
but the type and the deliverymodel.
And then the actions you cantake is that not only can you
should you have dashboards thatare telling you that, this very
high, this is a very high valueaudience cohort, right, that
we've identified.
(22:37):
And this particular cohort,we're seeing this trend data in
terms of content types.
We're actually now going tobuild out and be able to plan
for the next three to sixmonths.
Based on this the contentprogram.
And the ROI metrics you're goingto use are, the one engagement.
And it's not just engagement ofthey're looking at it and
reading it, but are they sharingit?
(22:57):
Are they liking it?
Are they coming back to it?
Are they going to another, ifyou have carrots, IE, like this
content is supposed to lead tothis, are those carrots being
followed?
The retention, the referralrate, if it's, if It may also be
something that you're doing todo new subscriber acquisition.
These are all the things youwould measure in this use case.
Did this use case resonate withanybody on the call here?
(23:21):
And we have some stuff fromKathleen.
This is similar case for me too.
Content development based onseasonality, type, title,
author, popularity, trendingtopics and profitability.
So Kathleen, out of curiosity,is this like what we described
this use case?
Are you able to deploy somethinglike this in your business, or
if not, what would need to takeplace?
(23:43):
I would say that only factorthat is being incorporated
currently is profitability, butreally it's all of those
multifaceted things that are alittle bit more touchy feely,
like the popularity of theseasonality.
Is it Christmastime and we'reoffering a content on like new
year, new you, all of that is alittle bit more subjective and
(24:05):
it's hard to incorporate, butI'm sure you probably have a way
to do this, but I feel like theonly things that are being taken
into account now is just sheerprofit, which is, subtracting
the marketing dollars from thesales generated from the
subscription, but I don't thinkit's taking into account a lot
of other things that should beadded to that.
Analyzed in the decision makingprocess, if that makes any sense
(24:26):
at all.
Yeah, no, absolutely.
That actually is goes back to myI'm the Sierra that says the
revenue we have right now isenough.
Right.
Exactly.
Yeah.
And Kathleen, just foreverybody, Kathleen's with
everyday health.
Which is owned by Spice ZiffDavis.
Yes, and it's a subscriptionbased.
(24:47):
Almost like, I would say theypurchase courses, but it's not
really a subscription.
It's purchasing individualcourses.
Exactly.
So another use case we have is amedia company with deep and
expansive reach into highlycoveted audience segments who
wants to improve pricing.
And this is a really interestingtopic that I am personally very
(25:08):
fascinated with is pricing,right?
And it is in this particular usecase, they want to identify ways
to improve pricing power fordigital advertising campaigns.
And so one of the biggest assetsyou have is a media or an events
company or digital informationis the audience and not just the
number, the quantity it's thedepth of which, the audience and
(25:28):
what they're doing and thesignals, the purchase signals.
In this case, we put togetherthe sales marketing, digital ad
ops and audience data andfinance teams.
And what we want to map isadvertiser ICP and buyer
personas.
So that's actually a new dataset.
So what would have to happen isin their CRM, they would have to
(25:50):
actually add these in there.
And in this case, I would say,do a test of their top 50
advertisers and start reallyunderstanding, from, sector.
Title region, who are theyreally wanting to look at, look
look to to to target and thenhow are we on the audience side
looking at what data we'recurrently collecting.
(26:12):
Versus data that we can accrueor gather from behavior without
asking someone to fill in a formthat would make that audience
segment map to the advertiserside very highly attractive.
So the insights you're trying togather here's my top 50
accounts.
Here's who they want to reach inthis particular brand.
(26:35):
Here's the audience we have.
And of that audience here arevery high value prospects,
right?
These are the ones that theywould covet the most and who are
the most engaged with us.
And then you trickle that downand then you could price
accordingly.
based on like dynamic pricing.
So the more highly coveted, themore engaged, the more valuable,
(26:56):
the higher the prices.
And you could do this acrossadvertising.
You could do this across leadgeneration, but that is one way
to optimize pricing based on oneof your most valuable assets,
which is your audience.
And the ROI you would look at isrevenue per audience member.
As well as.
On the advertiser side is, spendtime to acquire the 1st or the
(27:21):
renewal deal.
The program performance metrics,customer satisfaction is anyone
here on that?
And the call doing this rightnow?
Steven hasn't.
I think Steven, this might havebeen towards the to the earlier
use case.
Correct?
The good morning.
I texted in on the on thecontent monetization part, but.
(27:44):
I mean, it, it applies to both.
I mean, we chatted about this alittle bit a couple weeks ago,
but the, we're not we've gotpartial subscription, paid
subscription model, and we'remigrating to more paid products.
And the pricing of individualproducts, the pricing of
subscription products, and thensort of bundles.
Yeah.
That that applies generally.
But we don't have, we so far wedid some industry sort of
(28:07):
scanning about what people arecharging for what.
Content they have, and we'vedone a little bit of internal
work, but it was not done insort of a dashboard form.
It's done more informally thanthat.
But my interest applies to bothjust from a broad standpoint.
Well, actually, just in yourcase, another area that actually
comes to mind of a use cases isthe packaging of the
(28:29):
subscription, right?
Certain.
And especially if you're inmultiple countries there's
different ways people want topurchase content, whether, for
example, in, let's say the UK,they want to buy from a company
standpoint, whereas in the U Sit's individual.
That's my gut, right.
But where's the data to supportit and need that.
Yeah.
Because the UK has successfullyforged ahead with corporate
(28:52):
subscriptions and completelygotten nowhere with individual
subscriptions, but they, that'sfine.
We're using that as a model forthe U.
S., but the U.
S.
can do more like smaller groups.
I mean, we're going to go to ourbig customers.
And say, here's the 5, 000people in your company that we
have on file.
Here's 3000 of them that areactive users, and here's the
(29:15):
price, so to speak, to, to havethem have, the full gamut, and
it's going to be a lot lessexpensive than doing it
individually or departmentally.
So we have to have thatconversation, but that's 20
companies, right?
Well, we'll make money withanother a hundred companies that
are smaller individual groupsand libraries and other types
of, sort of non traditional.
(29:36):
Subscribers.
Right.
And actually even taking thedata analytics side to that, if
literally be being able toactivate your marketing based on
the insights you're getting fromthe different types of cohorts
you're going after so that, it'snot so manual.
That would be anotherapplication.
So another use case is revenuestream identification.
(30:00):
And this is how do I look atwhat's happening within my
current.
asset base to find new revenuestreams.
And again, we'll just use themedia company as a use case.
One that has audience segmentsacross multiple adjacent
markets.
So essentially, you would getyour stakeholder team together
and you're mapping data acrossyour C.
(30:21):
D.
P.
Your serum and then unstructureddata sources.
Both internally and externally.
Again, unstructured being,website usage, your social
media, your own social media andmaybe even third party.
And what you're finding is,where is that intersection
across perhaps these brands thatwe're not addressing, but that
there is.
There's need that we'reidentifying.
(30:43):
And how do we then take that?
And it could be that it's asimple as it's a new topic.
So we have a, spoke to somebodywho is talking to, has all
different markets.
Some of them not even adjacentto one another, but obviously
this artificial intelligencetopic is of huge demand.
(31:03):
So they're launching anartificial intelligence paid
newsletter that they're going tobe able to market to, 500, 000
plus database independent of themarkets.
So again, you would create youraction plan.
This one's more about testing ahypothesis first and rolling out
and testing.
And then with a longer termview, but again, your data can
(31:24):
tell you that if it's organizedcorrectly.
Chad, I've been kind of takingover here.
Do you have on the use case sidefor new revenue stream
identification?
Any examples from your side ofthe fence?
Yeah, absolutely.
We worked in the past with anumber on a number of these
types of examples, but one inparticular where a company was
(31:45):
looking to see in the market newbrands that were entering into a
relatively new industry.
And see which ones early on.
We're showing signs of highpotential growth.
So that was by looking at theirsocial media, by looking at
their sales data, website,traffic, et cetera.
And then identifying thosebrands and those companies that
(32:06):
were new entrance as potentialacquisition targets.
then acquire and roll up undertheir own brand.
So that was very, a veryinteresting kind of project
there to look out at, thirdparty data, trying to get a
sense of the market, trying toget a sense of where the
opportunities are early onenough that they could, acquire
these organizations before theygrew.
(32:28):
Before they grew too big.
Yep.
All right.
So just from a time check, Iwanted to, we've got a few other
ones, which we can go over in a,if you'd like to contact us for
a one off, but there's revenueexpansion.
And again, there's a gazilliondifferent ways you can use
predictive analytics to helpidentify risk and capture
opportunity.
But I want to hand this over toChad at this point.
(32:51):
And this is the process.
The process is More on thetechnical data side, but it is
very important because one ofthe things that I hear and so
does Chad is they're missing ahuge part of what has to happen
in order for them, to arrive atpredictive analytics.
It's like they want to go frombeing a baby to a 21 year old
(33:13):
and skip all the years inbetween.
I'm going to, we did, we puttogether this process diagram
and I'll have Chad Chad, I'll bedriving the slide.
So you'll just, let me know whento click next, but the floor is
yours.
Thanks Heather.
So yeah, this is a prettydetailed overview of kind of the
steps you need to take.
To Heather's point, this isreally speaking more to fully
(33:33):
operationalizing analytics andpredictive analytics within a
business.
The same process could be usedif you're doing a one time
exploration of data and tryingto get a one time kind of output
of a read on certain a certainpart of the business.
So you could do, and I'll speakto it as we go through, it could
be used for a one time projectas well, but really we're kind
of focused on how do weoperationalize this in a a very
(33:53):
efficient and robust mannerwithin a business.
So go ahead, Heather.
So first one, obviously you kindof want to understand what your
objectives are.
It's not enough to say, Hey, wewant predictive analytics.
We want to be like everyone elsewho's super cool.
And that area you have to havesome sort of objective to start.
Right.
And then, honestly to Heather'spoint earlier on in the
(34:15):
conversation, improving sales isalways an objective and it's a
great place to start.
So there are a number of otherexamples here, but you really
want to have something in mindspecifically, otherwise you're
going to be wasting time andmoney.
So next is kind of defining theteam.
So the, you need, kind of a.
Ownership on the project or onthe initiative.
(34:35):
Ideally, the executive level.
So someone who's going tosponsor it and make sure that
people who are on the team havetime to do, to partake in the
work.
You have to identify, as youkind of, as we'll go through,
you'll see, but we have toidentify folks who are able to,
who have the skills necessary tomake this happen and identify
where you might not have thatinternally.
So you might need to go out andget external help.
(34:56):
So this third one here, this isalso very important early on.
So you have your objective,you're trying to, improve
revenue forecasting as anexample, in order to do anything
here, you have to have data andyou have to have enough of it.
So there, in the example Iprovided before with regards to
finding acquisitionopportunities within the market.
(35:17):
Scarcity of data was a realproblem there.
It was tough to say, we knowexactly how much these new
brands are selling, or we knowexactly how much activity they
have in their business.
And so a big part of the projectas in that example was spending
time on this step here, lookingat third parties who.
provided data within certainareas that we could grab and use
(35:38):
to help fuel the model for,predicting the outcomes.
And not only that, you obviouslyhave to, on the, if you want to
improve revenue forecasting orsales, you have to look at the
CRM.
You have to look at, if you'vedone acquisition across a number
of different companies, are theyall on the same platform?
Do they have different standardsof how they're managing the data
and managing a sale?
(35:59):
And you have to start toidentify where those data points
reside and how much of it do wehave.
The amount of data, becomespretty important depending on
what you want to achieve here.
But this is, that exercise ofsaying, okay, here are the data
points that we have internally,here are the systems where they
reside.
And here's the kind ofnomenclature and the methodology
we use to collect them.
(36:20):
So here on the tools you,that's, also a kind of a big
step here, depending on whereyou are in the process.
So if you're starting fromcomplete greenfield and meaning
you don't have any technology inhouse to manage information or
manage data you could be lookingat a decent sized list of tools
that you need to go out into themarket and find and install but
(36:43):
some of the key ones that youwould absolutely need,
especially if you're doing thison an ongoing basis is the data
warehouse side.
You need some someplace to storethe single source of truth to
bring the data in from a dataintegration standpoint.
So what do we mean is just, amethod to automatically extract
your source data from your CRMand other systems and push it
into a centralized warehouse.
(37:03):
That way you can have it in amore flexible.
environment that you can modelthe data.
You can, run certain algorithmson it.
And you can, clean the data aswell.
The data visualization,algorithm development.
I'll get into that a little moredetail, but those are also going
to be, pretty importantdepending on the approach you're
taking.
So data preparation And if wethink about one, steps one
(37:25):
through five here, a lot of thisso far would be kind of the
precursor that Heather wasdescribing.
You kind of, you have to get toa point where you have your data
collected and cleaned up inorder to do much of anything,
even if you're not trying to dopredictive analytics, this is
something that you would do justto get General business
(37:45):
intelligence on theorganization, right?
So it's a step that's well worthit to go through or these steps
are really well worth it Even ifwithout the predictive analytics
part as the goal and the datapreparation side is really
coming up with a structure thatunifies the data from the
different systems So if you havecustomers in the customer
success System that havedifferent names than the
(38:06):
customers in the sales systemand the crm You have to have
some way of preparing that andblending those datasets
together.
So you can see from a termperspective, potentially, what's
their customer, what's thecustomer support activity and
success activity matched againstthe sales and the pipeline that
we have in the CRM.
And actually, Chad, I want tobring one thing up here.
Just again, knowing the types ofbusinesses that we work with.
(38:30):
There are so many tech tools andplatforms that are being used to
support events and media wherethis data is sitting for the
same customer understanding howto create that taxonomy, that
single source of truth.
While still moving your businessforward with all of those
platforms, it is a real bigfactor in this data prep part.
(38:51):
And there are definitely ways todo it without interrupting the
business flow or the sourcedata.
But that is something that I seestopping a lot of people because
of the enormity of it all, butthere are definitely are
solutions.
So step seven, six and sevenhere.
This is probably where you endup without internal expertise.
(39:13):
You might stall out a littlebit.
So this is going back to thefirst couple of steps.
What are our objectives?
What data do we have and how dowe map that to our objectives?
Now you have to figure out,like, how do we go about
actually creating the algorithmor creating the process, right?
Without having a lot ofexpertise or experience in this,
(39:33):
in doing this from scratch,you're going to, again, you
might need to go out externallyto get expertise to set it up as
a one time process that you canthen continue to run over time.
But I would also say that,depending on the nature of the
objective There are tools in themarket that can do this and have
determined the right model,determine the right features or
have a mechanism to determinethe features automatically and
(39:55):
by feature.
We mean thinking again about theeasy one that everyone knows on
the sales side.
I go back to a story from ourown business way back when we
were starting.
We had a little smaller set ofdata, obviously, as we were just
But we decided to do a kind ofan experiment and review of our
CRM and the deals that we hadclosed, the deals that we had
(40:16):
lost.
And in doing it just as a onetime evaluation, we were able to
identify three key attributes onour sales.
So on the opportunities datapoints, we collected that led to
a greater than 90 percent closerate.
So those three attributes, thenwe're pretty much our sales
method and process goingforward.
And we use those to close,business at a higher rate.
(40:38):
And continue to kind of keep aneye on those attributes.
And that's kind of what you meanby the feature selection.
It's like what data points arereally impactful.
Give it to that drive theoutcomes statistically.
So again, there are tools thatif you're looking for certain
outcomes or certain predictivemodels, there are tools that are
built in the market that willhave some of this all ready to
(41:00):
find.
If you have something that'svery unique and very niche in
your space, or you have datathat's very complex then you
might need to go and build moreof it from scratch.
And that's where you have to getinto more of the technical
expertise.
So here on the next step, afteryou've kind of defined your
model, you're really going totrain it and kind of validate
(41:21):
the results, right?
You might not have picked theright approach, might not have
picked the right data sets, thedate, the attributes, this step.
Is a means to kind of figurethat out and to over time
determine what is the mostoptimal model and most optimal
data set to use to generate theoutcomes.
And so you know, just simply,you kind of, you can think of it
(41:42):
where you're, you're splittingyour data sets.
into some that are fed throughthe model and you validate
against the model and then someyou are fed and you train the
model on and then you ideallywant to see the model work well
with new data that it hasn'tseen before.
And that's the process thatyou're overseeing here.
So that's, it's going, going toan example that most people
probably know with, which chapwith chap, she, Bt, they trained
(42:06):
their models internally.
You had a ton of people doingit.
Obviously, and then theyreleased it in the wild, and,
they're getting feedback basedon its performance that
incorporates itself back in tothe model to get better over
time.
And that's kind of touched onhere as well.
So the evaluation, youobviously, you want to see it
work well against the trainingdata, but also in the new data
(42:26):
as it comes in.
So there are methods fordetermining, how accurate is
this or how reliable is thisgiven the different nuances we
see in the data.
The step 10 iterativeimprovement again, you
oftentimes won't get itcompletely right on the first
time or in the first try.
And so you kind of, you need totake this as a, known factor
(42:46):
going into the project, into theinitiative, you're going to have
to, adjust over time continue tolook at the performance and see
if it's.
Continue to perform well overtime and make changes if needed.
So on step 11, educate andcommunicate really important
here.
If you want actual adoptionwithin the business, a good
example from our experiences, insome cases, if you define the
(43:10):
model and define the outputs andthey're very obscure or very
difficult to comprehend.
In a dashboard or in a report oranything like that and you try
to distribute that to peopleThey would just they just won't
use it.
They won't believe it.
They won't understand it andYou'll have a bunch of wasted
effort.
So you really have to make surethat the outputs are something
(43:30):
that are easily digestible justin the visual and report
presentation form, but you alsoneed to educate them on how is
it, how was it constructed?
What factors do we use todetermine the future sales or
the churn risk?
As soon as they understand that,then they can have more faith in
the model and have more faith inthe outcomes in the outputs, and
(43:50):
they'll be able to adopt it.
But if you don't do that it'srare that folks will really
trust it from the outset.
And I'll just pipe in here.
What I've seen is if that's notdone well, then all of a sudden,
well, that's that's data iswrong, so it just gives someone
if it doesn't say what theylike.
So this part is, I think, supercritical and then after this,
(44:12):
you get to go on and do more.
So if you pick something narrowand very, contained, very Again,
the idea is that you can scalethis to other areas.
And we've seen that within ourcustomers time and time again,
we set up something for them.
They love the output and thenthey might start to do things on
their own, or at the very leastsuggest other ideas.
(44:33):
And areas within the businessthat they can apply these
techniques to.
And so it's, from start smalland then iteratively kind of
scale up and have the teamwithin the business, take on
some of that some of thatexploration and model
development.
Any questions about the process?
(44:55):
All right.
Again, if you want to talk to usoffline, we're here.
So my favorite part is alwaysthe landmines.
We, the landmines are andlandmines happen before, during
and after.
So they're really important tounderstand.
But Chad, you want to kick thisoff?
Sure.
Just a couple of that.
I'll call out.
(45:15):
So not enough data.
Thank you.
If you have and, ambitious ornot, maybe not even ambitious,
objective you need to make sureyou have enough data to actually
make it statistically possibleto produce predictive outputs.
We, there It's a little bit of achicken and egg, but.
You do want to make sure youhave sufficient historical clean
(45:36):
data, in your single source oftruth that you can run the
models through.
And that's oftentimes forsmaller organizations.
That's where that's really wherethey aren't able to leverage.
Predictive analytics as much asbecause there's just not enough
data to make use or definepatterns.
But, that's one I would saydiscarding models too early as
well.
You might see outputs early onthat don't make any sense.
(45:57):
It doesn't mean that everythingis wrong.
It just means that you mighthave it off.
You might have it configuredslightly off.
You might have, a couple inputsthat are not clean, more couple
inputs and weights within themodel that aren't working.
And so it's.
It's don't try to don't quit tooearly on.
When you're starting theprocess.
Yeah.
And I think the other one is andwe touched on all of these, but
(46:18):
I think that the big one is theactivation plan and process to
activate and how you're going totrack and measure.
it's very important to show yourboard, your CFO the value that's
being added to the business forthese investments.
But there's a lot of culturechange and process change that's
required.
So thinking that through aheadof time is critical and not
(46:40):
doing it is definitely a way tohave that landmine.
Go off.
I also think that not proceedinguntil the data is perfect.
No data is perfect.
And especially in the businesseswe're talking about it is
there's so much data.
But you have to start somewhere.
So figure out where that is, andyou'll perfect as you go.
Capabilities are evolutionary.
(47:00):
And I would say the the assumingeveryone is on board, or I like
to call the saboteur in theroom.
You've got to identify theperson or people who are...
are not willing or really tryingto stop progress from happening
in your business through the useof data.
I personally see it as theemperor's clothes.
(47:20):
They don't want to, they don'twant to expose what's really
happening.
But really finding that out andunderstanding why any resistance
and why, democratized.
data environment and everyone isbetter at their jobs when they
have data at the ready.
Those are just a few of thelandmines.
(47:41):
I don't know if any of you areencountering any that you'd be
willing to share, but we have acouple left but that's our
presentation.
Any landmines that anyone wantsto dive into we've had a few
people have to leave, but therearen't any questions.
I will share these last twoslides.
There's our, the chat on myemail are there to take
(48:03):
advantage of the office hours.
You can email us.
I'm happy to sit down and talkabout where you are in your
journey.
And again, these are the waysthat we can help at H2K.
And then we have another bootcamp coming up on October the
19th, and we have Denise Medved,the Chief Commercial Officer.
She's new in the role atInformer Markets, and Matt York,
(48:25):
who's the CRO of Foundry, and itwill be actually a really
interesting discussion becauseInforma is very heavy on the
event side.
Foundry is 80 percent digitallead gen media.
And they're both doing somereally exciting things in terms
of a single source of truth andpursuing revenue excellence
through data.
(48:46):
So with that, I would say ifthere's no more comments or
questions, you'll get fourminutes back to your day and
we'll send everyone therecording.
Great.
Thank you so much, everyone.
Thanks, Heather.
Heather Holst-Knudsen (48:59):
You can find us@2klabs.com.
Thank you.
Thank you for listening to TheRevenue Room by H2K Labs.
Subscribe to our channel today.
Welcome to the Revenue Room,presented by H2K Labs.
(00:05):
Here's your host, HeatherHolst-Knudsen.
well, welcome everybody to theRevenue Room Bootcamp, this is
an event that is hosted by H2KLabs and, my pal over here, Chad
Rose from Inside Out.
What we try to do is provide acombination of data and business
(00:27):
insights to help you drive datafueled revenue growth.
So today, we'll kick off and gothrough, a few areas.
One is unpacking the hype.
Every, everyone's talking aboutpredictive analytics, but you
know, there's a lot that goesinto it.
What's realistic, what's not theopportunities, both in term of
(00:47):
identifying risk and alsorevenue opportunity, the process
that you need to go by and alsolandmines to avoid.
So I'm Heather Holtz Knudsen.
I'm the CEO of H2K Labs.
I am a veteran in the B2B mediaand events and digital
information space.
(01:07):
Chad is the CEO of InsideOut.
And Chad, want to give a realquick background?
Yeah, thank you, Heather.
So my background is primarily indata engineering and analytics.
So I've been doing that sincefor my entire career and helping
companies specifically withinthe middle market and enterprise
space to, make the most out ofdata, get the most out of the
(01:28):
data.
Great.
And what we do at H2K labs is wehelp in three ways.
We help through consultingacross data, revenue and product
strategy.
We also help with planning andexecution through workshops and
capability audits.
And we have two platforms, oneof which we've partnered with
chat on called insightify, whichis a modern data management and
(01:50):
predictive analytics platformwith very industry specific
capabilities for media events,digital information and another
solution called channel metrics,which helps on the customer
delivery side.
So to get the most out of theboot camp, if you feel up to it,
put your camera on.
If you have questions please usethe comment section and we'll
(02:11):
stop.
It, this is meant to be adiscussion forum and it says
polling.
We don't have polling for thisone but after the boot camp,
you're also going to get somegreat information, including a
playbook, the deck, and a linkto the recording, which you can
share with your peers.
So the focus is really on data.
If you'll revenue andprofitability when we talk about
data, it really is on therevenue side.
(02:34):
There are many people out therealready handling audience.
We connect with audience when wetalk about things, but we're
really about the commercial sideof the equation and our content
while free.
It is highly specialized andthere's a lot of depth to it.
So we hope that you do walkaway.
Thank you.
Either finding out how to figuresomething you didn't know, or
(02:56):
you didn't know something andnow you do.
And then after the bootcamp, asyou a series of content items
that can help you.
And also we're offering officehours an hour to dive deeper
into some of the things wetalked about to help you as you
build out your strategy.
So let's talk about the hype.
(03:17):
So what are predictiveanalytics?
It's the way you're taking yourdata past and current and
blending that together with newinsights, utilizing structured
and unstructured formats andapplying statistical and machine
learning to predict futureoutcomes.
It's not just to predict, it'sactually to activate.
Like, how do we change thecourse of something if it's
(03:40):
Looking not so great or how dowe activate and leverage an
opportunity if it has if thathas come to surface and
predictive analytics our meansit's a way to get to a better
business outcome, a leaner, moreproactive company, better and
greater transparency andaccountability into processes
and outcomes and deeper insightsinto unmet customer needs to
(04:02):
gain market share.
But Chad, did I miss anything onthis slide?
Do you think?
No, I think you got it.
I think on the hype side, it'sjust one thing to add there.
Oftentimes folks think that theycan apply predictive analytics
across the entire organizationin every domain, every
department.
As you'll see, as we go throughthis, there are some precursors,
there's some requirements, so tospeak, that make it more
(04:25):
feasible in some areas thanothers.
And so You know, to your pointhere, Heather, it's really there
are means that it's not the endall, be all at for every part of
the company and every person inthe company agree 100%.
And the other point I like tomake is because this is a real.
Real big issue, especially forlike CEOs and CFOs is, we're
(04:48):
investing in data, but We don'tsee it.
What's the ROI you're investingin your business and data is the
enabler of those investments soyou need to look at it across
things like Building data drivendecision making across the
organization How do I make myteam?
And every function that requiresthis able to use data to improve
(05:08):
how they perform and how thebusiness performs.
How am I using data to growrevenue?
I always like to say there's nota CRO, CFO, or CEO I've ever
spoken to that says, we'rereally satisfied with the
revenue right now and we don'tneed to do anything.
No it's an investment infiguring out growth, innovative
growth, and it's a way to gainoperational efficiency in smart
(05:29):
ways.
Like where can we look at areas?
to recalibrate resources sothey're applied to the most
profitable high growth areasversus the not so high growth
areas.
And frankly, from, an investmentstandpoint or you want to be
acquired, it's a way to increaseyour multiple your valuation,
(05:49):
your acquisition attractivenessand the speed.
to which this, an acquisitiontakes place.
And it is definitely an improvedboard and investor.
It improves the way youcommunicate with your board and
investors.
I was speaking to another CRO inB2B media.
And one of the big things theysaid is, they're owned by
private equity is theirinvestors are asking for data.
(06:09):
More frequently, more granularand more like on things that
literally it's it takes themweeks to pull it together.
By taking this approach, it cangreatly enhance that
communication, which had on thisarea.
I know you deal a lot withprivate equity, for example.
Are there other areas that youthink that impact the business
when you invest in data?
(06:30):
I think you've covered some ofthe key ones here.
Heather double down on theimpact it has on the team.
And so as you as if you do thiscorrectly and you're able to
embed predictive analytics intothe business the team will
largely or oftentimes kind oftake take that on.
And what I mean by that isthey'll be more interested in
(06:50):
pursuing.
Other areas of their day to daywhere they might see a value in
adding additional predictiveanalytics.
So they might even, research howto do it themselves or create
models themselves.
So as soon as you start to embedthis within the business, the
team members do take a biginterest and take ownership of
it.
And that's in the best casescenario.
(07:11):
And so I think that's a reallyimportant point.
Small wins spark the appetite,right?
Another area that we see whenworking with compliance is that,
looking, taking a huge bite ortrying to eat the whole cake.
It's really start small, savorand really learn what the
ingredients are before you moveon to the next one.
(07:31):
But small wins will not onlyenable a much faster, better
path to R Y.
It actually will prove the valueand which will enable it to
scale.
And there is, there's definitelywe always see low hanging fruit
that you can go after first,whether it's in a very specific
division like the financedepartment or it could be sales.
(07:52):
But again, small, quick wins arethe way to go.
And then it's a journey, right?
So you have to start somewhere.
And I guess the question that Iwould ask everyone is, Is what
I'm doing today.
Okay.
Can I continue to make mybusiness grow the way I needed
to grow just doing what we'redoing?
Or do I have to take this on andI need to run my business
(08:14):
better.
I have to become data driven.
And so where do I start?
And then finally, it takes avillage.
It's, and I'm going to actually,Chad, because you deal with this
so much is there so much,there's a process which we'll
dive into later into thisdiscussion.
Bye.
Bye.
Bye.
And all of these elements areabsolutely critical.
(08:37):
What can you talk, tell, talk tous about what you see when you
look at the, how important thisprocess is.
Yeah.
I mean, in terms of, thismessage here, it really starts
at the executive level, andhaving sponsorship there and
taking on the challenge ofreally becoming more data
focused and oriented.
And then, you have.
(08:57):
Impacts on the organization inthe source systems, like your
CRM, where you're going to, ifyou have predictive models that
are built on top of the datathat you're collecting, there's
a dependency on the people whoare putting the data in there.
And so in the way that thesystems are configured, and so
there are elements across thebusiness that start to impact
the outcomes of the analytics.
(09:18):
And so it does hit a lot ofpeople in a lot of different
areas as you start to reallyoperationalize it.
Absolutely.
Okay.
Does anyone want to, talk aboutwhat your, what are your main
business objectives when youthink about predictive
analytics?
Maybe post it in the commentsor, and this will help us make
(09:38):
sure that we're talking aboutthings that and hitting squarely
on some of your top priorities.
Feel free to throw that in hereand Chad will be watching as I
go through so let's talk aboutcapabilities.
I took this course and Irecommend anyone who's
interested in data monetizationor how to improve your business
and predictive analytics fallsunder data monetization.
(09:59):
It's by MIT Sizer.
It's fabulous.
It is it really helps with,organizing the thought process
and how you approach it.
But there are two really bigoutcomes that I took from this.
And that is you need to havedata capabilities that map to
this to where you want to be interms of a data maturity model
and how you're activating that.
(10:20):
And there are guiding principlesthat, that underscore.
So in terms of predictiveanalytics, these are the five
areas.
If you're pursuing datamonetization, As it relates to
predictive analytics, whichfalls under the internal, right?
You're using data to add cash tothe bottom line by improving
what you're doing internally.
(10:41):
You need to have very sophist onyour data assets, single source
of truth.
It needs to be able to beblended and you need to apply
data science and the machinelearning side.
You need to be able to provideaccess to that data to your team
members, because it needs to betransparent.
So you need a platform.
I mentioned the machinelearning.
And that's not just about one ofthe big myths about predictive
(11:04):
analytics is like, Oh, I put apredictive analytics platform in
and like, let me push a button.
Predictive analytics, actually,the data needs to mature and
learn for a period of time.
So having this capability isimportant.
Thank you.
You don't need to be superadvanced on the acceptable data
you side, but you do need tohave internal oversight and
permissioning who sees what, etcetera.
(11:25):
And on customer understandingagain, not you don't need to be
out in what I call the outwardcommercialization area.
But you do need to understandwhat the data is telling you
about.
your customers.
And then the guiding principlesabout that underscore this is
that, data monetization is thedirect or indirect conversion of
data into financial games dataliquidity.
(11:48):
And this is the area where wespecialize in our businesses
that have very high dataliquidity.
But this is the ease of whichyour data can be monetized.
And in the area of events,media, business information,
marketing services you'redealing with huge amounts of
data that changes every singleday, every minute, actually
every action on your website isa new data event.
(12:10):
Being a data drivenorganization, it's an
organization that's constantlyinnovating and scaling using
data to improve businessoutcomes and this data
democratization, which anyonewho's on this call and has
spoken to me before, I'm verypassionate about is, Your whole
team needs to have access todata.
It needs to be built forbusiness users.
You are not going to be able toacquire data skills.
You have to actually build theminternally.
(12:32):
So one way to do that isdemocratizing the data.
So another question, if you wantto throw it into the chat is
areas we're going to talk aboutare like gaps that are stopping
you and based on what we justreviewed.
All right.
So this is the risk andopportunity side.
(12:54):
You have more data than you everhad before.
And again, if you're in themedia events or any type of
business that has a two sidedbusiness model, or is dealing
with multi channel marketing,you've got more data than you
ever had beyond what I call atraditional business.
It's where do you start?
And it's not just enough toacquire it.
(13:14):
You have to understand it, haveto be able to operationalize it,
right?
How am I activating behaviorsinternally?
Based on what we're learning toimpact, and I should have a
fourth hour here.
And that is how am I measuringit?
Okay.
So what Chad and I did is webroke down three ways that
(13:36):
predictive analytics can helpyour business.
There's one is growing sales.
Identifying expansionopportunities, personalizing
campaigns to have better impacton pipeline conversion,
maximizing your lifetime value,whether it's your audience or
your your exhibitor oradvertiser side reducing churn
and optimizing pricing, you canimprove or optimize performance
(13:59):
by.
Bringing the right products tomarket or sunsetting less
profitable ones.
Optimizing resource allocationbased on customer spend, not all
customers are alike and you needto, measure that.
How can I look through the data,what's working from a sales
standpoint and scale that acrossall of my sales team players so
(14:22):
that I can improve my A game andreduce my C game.
And how do I optimize contentand how we're investing in it
and how we're producing it,audience and the conversion side
to programs.
And then clearly there's therisk, deal risk, customer churn
risk.
And again, in the world of, thatwe operate in it's churn is
(14:44):
actually a very complicatedmatter because there could be a
lot of revenue reductionhappening along the way.
Well, before the churn takesplace.
So how can you utilizepredictive analytics to identify
that erosion before the actuallogo goes away payment risk?
It's another area attrition onactually, which is Repeat of the
(15:06):
employee fight flight risk andthen product adoption risk.
So we put together some usecases.
There are about six right nowthat we have in here.
And this first one is thecustomer churn risk.
This is and you could slice anddice this one in many different
ways, but in this use case,let's, it's a marketing services
company that's serving diversecustomer segments across
(15:26):
multiple brands in differentcountries.
And the channels include events,digital advertising and legion.
You could add.
More the point is that there areall these different factors that
create data complexity.
And what they're trying to do isthey want to identify red flag
alerts in time to action andfix.
Right?
Because as I mentioned, thechurn is a painful, slow trickle
(15:48):
sometimes.
But unfortunately, as I'veexperienced personally, and I
have seen with customers by thetime that red flag really gets
raised, it's too late, right?
And we're all running aroundwith our, like chickens with
their head cut off trying toimprove, but the damage has been
done.
So how this customer wants toidentify this further.
So you build a stakeholder team,you put sales, customer success,
(16:10):
operations and finance together,and you map things that are in
your business that would saythis constitutes a red flag
alert, right?
And in which in this case, CRM,we're going to map CRM data with
operational data and program.
Program performance data.
We actually also want to maporder management because we want
(16:32):
to ensure, one of the biggestissues with churn risk is the
the program doesn't deliver.
So I need to know what waspromised and what they paid for
and when it was supposed to bedelivered.
And I also want to put in someother things like third party
data.
So LinkedIn, for example andanother churn risk is if someone
departs The actual customercompany and LinkedIn, for
(16:54):
example, has a, an alert systemthat tells you if someone moved
or there's a change in LinkedInprofile.
So you then create insightsthrough the data and it could be
the program performancedashboard.
Is there if you're doing surveysat events?
the customer turnover, the thirdparty data that I just
mentioned, and also obviouslythe salesperson traction and
(17:15):
pipeline traction.
And you create dashboards basedon what you believe are these
red flag alerts.
And you also, on top of it, havean actioning process, right?
What are we going to do whenthese alerts are surfaced.
So and it's a process.
Everything that impacts thesales team, the ops team, the
customer success team and thefinance team.
(17:35):
Another data point you may wantto put in here is if the
customer typically pays on time.
If payment is delayed, thatcould be an indicator of
financial troubles.
And then you assign some ROImetrics, churn rate, net
retained revenue, customersatisfaction.
And these all go into thedashboard that you're looking at
so that you can actually notonly see what's happening and
(17:57):
action it, but is it making animpact?
Are there any questions on thisuse case or does anyone have an
example where they're doing thisright now?
Okay.
The next use case, and this isfrom an outlier, right?
This is Chad has customers withthe product we sell.
And this one I love because itis I actually can liken this to
(18:18):
acquiring attendees for an eventor selling audience
subscriptions because there'sseasonality to it.
But Chad, you want to talkthrough this use case?
Sure.
Absolutely.
So in this use case, it's reallyabout trying to get ahead of Any
sort of downturn or reduction insale and future sales.
So the idea here was to build amodel to predict out a couple of
(18:41):
weeks, a couple of months inadvance based on current year
data and prior year data, howmuch.
The business was going to do innet new sales and how much they
were going to do in same storesales year over year, which is
their primary metric of success.
Right?
And building a predictive modelhere allowed them to see out
(19:02):
into the future very accuratelywithin the first within the next
couple of weeks.
And even further out, To say ifthey see a downtick in bookings
or in potential sales, they wereable to respond with increased
promotional activity, increasedmarketing activity and other
triaging to ensure that theyactually address that shortfall
(19:24):
ahead of time.
So great example of wherepredictive analytics is really
giving them, visibility into thefuture and giving them,
actionable things that they needto do to address problems before
they, surface.
Anyone have a question on thator how that how that could apply
or translate into your business?
(19:45):
Okay, so the next use case iscontent optimization and one of
the things for that will dobetter next boot camps is we
actually got this from one ofthe participants on the call who
spoke to me about like, I'mhaving an issue with this.
Can you.
Address it.
And it's how do we and thisreally applies if you're also
(20:08):
selling content as part of asubscription, or you need to
produce content that appeals toan audience that you are
attracting for an event is ifyou have content that's on your
website.
And you're producing that allthe time and there is user
engagement and you can slice anddice that.
(20:29):
How what can we use that datafor in terms of helping us plan
for better engagement, bettercommercialization.
And better investment in thiscontent creation.
So it's a B2C media company thathas multiple brands, very large.
And they have a paid contentsubscription model.
And their objective is to reducethe hours they spend each week,
creating a scalable educationcontent packages for their paid
(20:52):
subscribers.
Every week they get down and sitdown at a table.
They're sitting there trying to,well, I think this, well, let's
do that.
So it's very anecdotal.
It's got, again, having beenthere many times that you're
wrong.
So how can you actually use datato do a few things?
One data to reinforce and thedecisions, right?
No, here's the data that'stelling me that this is or us.
(21:13):
This is what we should be doingto reduce the time that's being
spent agonizing over what tocreate.
That time could be spentcreating value.
And three being able to actuallymap this content really better
use the investment better.
Okay.
Towards what your audience istelling you they want today and
some trends of what they maywant tomorrow.
(21:33):
It's also, by the way,underlying this there, there is
actually a a new productideation component as well.
But for this particular usecase, you get the stakeholder
teams of marketing content,audience and data together.
You're you need to map websiteanalytics, own social analytics,
your C.
D.
P.
Data, C.
(21:54):
M.
S.
Data.
And really break down intoaudience core cohorts to content
segmentation and really identifythat to demographics, location,
cohorts and content type.
Also, by the way, it's not Youknow, is it product information?
Is it thought leadership?
Is it, do they like short andsweet?
Or are they, is this type ofaudience looking for more
(22:17):
detailed, bullet point type ofcontent?
So it really helps youunderstand not just the topic,
but the type and the deliverymodel.
And then the actions you cantake is that not only can you
should you have dashboards thatare telling you that, this very
high, this is a very high valueaudience cohort, right, that
we've identified.
(22:37):
And this particular cohort,we're seeing this trend data in
terms of content types.
We're actually now going tobuild out and be able to plan
for the next three to sixmonths.
Based on this the contentprogram.
And the ROI metrics you're goingto use are, the one engagement.
And it's not just engagement ofthey're looking at it and
reading it, but are they sharingit?
(22:57):
Are they liking it?
Are they coming back to it?
Are they going to another, ifyou have carrots, IE, like this
content is supposed to lead tothis, are those carrots being
followed?
The retention, the referralrate, if it's, if It may also be
something that you're doing todo new subscriber acquisition.
These are all the things youwould measure in this use case.
Did this use case resonate withanybody on the call here?
(23:21):
And we have some stuff fromKathleen.
This is similar case for me too.
Content development based onseasonality, type, title,
author, popularity, trendingtopics and profitability.
So Kathleen, out of curiosity,is this like what we described
this use case?
Are you able to deploy somethinglike this in your business, or
if not, what would need to takeplace?
(23:43):
I would say that only factorthat is being incorporated
currently is profitability, butreally it's all of those
multifaceted things that are alittle bit more touchy feely,
like the popularity of theseasonality.
Is it Christmastime and we'reoffering a content on like new
year, new you, all of that is alittle bit more subjective and
(24:05):
it's hard to incorporate, butI'm sure you probably have a way
to do this, but I feel like theonly things that are being taken
into account now is just sheerprofit, which is, subtracting
the marketing dollars from thesales generated from the
subscription, but I don't thinkit's taking into account a lot
of other things that should beadded to that.
Analyzed in the decision makingprocess, if that makes any sense
(24:26):
at all.
Yeah, no, absolutely.
That actually is goes back to myI'm the Sierra that says the
revenue we have right now isenough.
Right.
Exactly.
Yeah.
And Kathleen, just foreverybody, Kathleen's with
everyday health.
Which is owned by Spice ZiffDavis.
Yes, and it's a subscriptionbased.
(24:47):
Almost like, I would say theypurchase courses, but it's not
really a subscription.
It's purchasing individualcourses.
Exactly.
So another use case we have is amedia company with deep and
expansive reach into highlycoveted audience segments who
wants to improve pricing.
And this is a really interestingtopic that I am personally very
(25:08):
fascinated with is pricing,right?
And it is in this particular usecase, they want to identify ways
to improve pricing power fordigital advertising campaigns.
And so one of the biggest assetsyou have is a media or an events
company or digital informationis the audience and not just the
number, the quantity it's thedepth of which, the audience and
(25:28):
what they're doing and thesignals, the purchase signals.
In this case, we put togetherthe sales marketing, digital ad
ops and audience data andfinance teams.
And what we want to map isadvertiser ICP and buyer
personas.
So that's actually a new dataset.
So what would have to happen isin their CRM, they would have to
(25:50):
actually add these in there.
And in this case, I would say,do a test of their top 50
advertisers and start reallyunderstanding, from, sector.
Title region, who are theyreally wanting to look at, look
look to to to target and thenhow are we on the audience side
looking at what data we'recurrently collecting.
(26:12):
Versus data that we can accrueor gather from behavior without
asking someone to fill in a formthat would make that audience
segment map to the advertiserside very highly attractive.
So the insights you're trying togather here's my top 50
accounts.
Here's who they want to reach inthis particular brand.
(26:35):
Here's the audience we have.
And of that audience here arevery high value prospects,
right?
These are the ones that theywould covet the most and who are
the most engaged with us.
And then you trickle that downand then you could price
accordingly.
based on like dynamic pricing.
So the more highly coveted, themore engaged, the more valuable,
(26:56):
the higher the prices.
And you could do this acrossadvertising.
You could do this across leadgeneration, but that is one way
to optimize pricing based on oneof your most valuable assets,
which is your audience.
And the ROI you would look at isrevenue per audience member.
As well as.
On the advertiser side is, spendtime to acquire the 1st or the
(27:21):
renewal deal.
The program performance metrics,customer satisfaction is anyone
here on that?
And the call doing this rightnow?
Steven hasn't.
I think Steven, this might havebeen towards the to the earlier
use case.
Correct?
The good morning.
I texted in on the on thecontent monetization part, but.
(27:44):
I mean, it, it applies to both.
I mean, we chatted about this alittle bit a couple weeks ago,
but the, we're not we've gotpartial subscription, paid
subscription model, and we'remigrating to more paid products.
And the pricing of individualproducts, the pricing of
subscription products, and thensort of bundles.
Yeah.
That that applies generally.
But we don't have, we so far wedid some industry sort of
(28:07):
scanning about what people arecharging for what.
Content they have, and we'vedone a little bit of internal
work, but it was not done insort of a dashboard form.
It's done more informally thanthat.
But my interest applies to bothjust from a broad standpoint.
Well, actually, just in yourcase, another area that actually
comes to mind of a use cases isthe packaging of the
(28:29):
subscription, right?
Certain.
And especially if you're inmultiple countries there's
different ways people want topurchase content, whether, for
example, in, let's say the UK,they want to buy from a company
standpoint, whereas in the U Sit's individual.
That's my gut, right.
But where's the data to supportit and need that.
Yeah.
Because the UK has successfullyforged ahead with corporate
(28:52):
subscriptions and completelygotten nowhere with individual
subscriptions, but they, that'sfine.
We're using that as a model forthe U.
S., but the U.
S.
can do more like smaller groups.
I mean, we're going to go to ourbig customers.
And say, here's the 5, 000people in your company that we
have on file.
Here's 3000 of them that areactive users, and here's the
(29:15):
price, so to speak, to, to havethem have, the full gamut, and
it's going to be a lot lessexpensive than doing it
individually or departmentally.
So we have to have thatconversation, but that's 20
companies, right?
Well, we'll make money withanother a hundred companies that
are smaller individual groupsand libraries and other types
of, sort of non traditional.
(29:36):
Subscribers.
Right.
And actually even taking thedata analytics side to that, if
literally be being able toactivate your marketing based on
the insights you're getting fromthe different types of cohorts
you're going after so that, it'snot so manual.
That would be anotherapplication.
So another use case is revenuestream identification.
(30:00):
And this is how do I look atwhat's happening within my
current.
asset base to find new revenuestreams.
And again, we'll just use themedia company as a use case.
One that has audience segmentsacross multiple adjacent
markets.
So essentially, you would getyour stakeholder team together
and you're mapping data acrossyour C.
(30:21):
D.
P.
Your serum and then unstructureddata sources.
Both internally and externally.
Again, unstructured being,website usage, your social
media, your own social media andmaybe even third party.
And what you're finding is,where is that intersection
across perhaps these brands thatwe're not addressing, but that
there is.
There's need that we'reidentifying.
(30:43):
And how do we then take that?
And it could be that it's asimple as it's a new topic.
So we have a, spoke to somebodywho is talking to, has all
different markets.
Some of them not even adjacentto one another, but obviously
this artificial intelligencetopic is of huge demand.
(31:03):
So they're launching anartificial intelligence paid
newsletter that they're going tobe able to market to, 500, 000
plus database independent of themarkets.
So again, you would create youraction plan.
This one's more about testing ahypothesis first and rolling out
and testing.
And then with a longer termview, but again, your data can
(31:24):
tell you that if it's organizedcorrectly.
Chad, I've been kind of takingover here.
Do you have on the use case sidefor new revenue stream
identification?
Any examples from your side ofthe fence?
Yeah, absolutely.
We worked in the past with anumber on a number of these
types of examples, but one inparticular where a company was
(31:45):
looking to see in the market newbrands that were entering into a
relatively new industry.
And see which ones early on.
We're showing signs of highpotential growth.
So that was by looking at theirsocial media, by looking at
their sales data, website,traffic, et cetera.
And then identifying thosebrands and those companies that
(32:06):
were new entrance as potentialacquisition targets.
then acquire and roll up undertheir own brand.
So that was very, a veryinteresting kind of project
there to look out at, thirdparty data, trying to get a
sense of the market, trying toget a sense of where the
opportunities are early onenough that they could, acquire
these organizations before theygrew.
(32:28):
Before they grew too big.
Yep.
All right.
So just from a time check, Iwanted to, we've got a few other
ones, which we can go over in a,if you'd like to contact us for
a one off, but there's revenueexpansion.
And again, there's a gazilliondifferent ways you can use
predictive analytics to helpidentify risk and capture
opportunity.
But I want to hand this over toChad at this point.
(32:51):
And this is the process.
The process is More on thetechnical data side, but it is
very important because one ofthe things that I hear and so
does Chad is they're missing ahuge part of what has to happen
in order for them, to arrive atpredictive analytics.
It's like they want to go frombeing a baby to a 21 year old
(33:13):
and skip all the years inbetween.
I'm going to, we did, we puttogether this process diagram
and I'll have Chad Chad, I'll bedriving the slide.
So you'll just, let me know whento click next, but the floor is
yours.
Thanks Heather.
So yeah, this is a prettydetailed overview of kind of the
steps you need to take.
To Heather's point, this isreally speaking more to fully
(33:33):
operationalizing analytics andpredictive analytics within a
business.
The same process could be usedif you're doing a one time
exploration of data and tryingto get a one time kind of output
of a read on certain a certainpart of the business.
So you could do, and I'll speakto it as we go through, it could
be used for a one time projectas well, but really we're kind
of focused on how do weoperationalize this in a a very
(33:53):
efficient and robust mannerwithin a business.
So go ahead, Heather.
So first one, obviously you kindof want to understand what your
objectives are.
It's not enough to say, Hey, wewant predictive analytics.
We want to be like everyone elsewho's super cool.
And that area you have to havesome sort of objective to start.
Right.
And then, honestly to Heather'spoint earlier on in the
(34:15):
conversation, improving sales isalways an objective and it's a
great place to start.
So there are a number of otherexamples here, but you really
want to have something in mindspecifically, otherwise you're
going to be wasting time andmoney.
So next is kind of defining theteam.
So the, you need, kind of a.
Ownership on the project or onthe initiative.
(34:35):
Ideally, the executive level.
So someone who's going tosponsor it and make sure that
people who are on the team havetime to do, to partake in the
work.
You have to identify, as youkind of, as we'll go through,
you'll see, but we have toidentify folks who are able to,
who have the skills necessary tomake this happen and identify
where you might not have thatinternally.
So you might need to go out andget external help.
(34:56):
So this third one here, this isalso very important early on.
So you have your objective,you're trying to, improve
revenue forecasting as anexample, in order to do anything
here, you have to have data andyou have to have enough of it.
So there, in the example Iprovided before with regards to
finding acquisitionopportunities within the market.
(35:17):
Scarcity of data was a realproblem there.
It was tough to say, we knowexactly how much these new
brands are selling, or we knowexactly how much activity they
have in their business.
And so a big part of the projectas in that example was spending
time on this step here, lookingat third parties who.
provided data within certainareas that we could grab and use
(35:38):
to help fuel the model for,predicting the outcomes.
And not only that, you obviouslyhave to, on the, if you want to
improve revenue forecasting orsales, you have to look at the
CRM.
You have to look at, if you'vedone acquisition across a number
of different companies, are theyall on the same platform?
Do they have different standardsof how they're managing the data
and managing a sale?
(35:59):
And you have to start toidentify where those data points
reside and how much of it do wehave.
The amount of data, becomespretty important depending on
what you want to achieve here.
But this is, that exercise ofsaying, okay, here are the data
points that we have internally,here are the systems where they
reside.
And here's the kind ofnomenclature and the methodology
we use to collect them.
(36:20):
So here on the tools you,that's, also a kind of a big
step here, depending on whereyou are in the process.
So if you're starting fromcomplete greenfield and meaning
you don't have any technology inhouse to manage information or
manage data you could be lookingat a decent sized list of tools
that you need to go out into themarket and find and install but
(36:43):
some of the key ones that youwould absolutely need,
especially if you're doing thison an ongoing basis is the data
warehouse side.
You need some someplace to storethe single source of truth to
bring the data in from a dataintegration standpoint.
So what do we mean is just, amethod to automatically extract
your source data from your CRMand other systems and push it
into a centralized warehouse.
(37:03):
That way you can have it in amore flexible.
environment that you can modelthe data.
You can, run certain algorithmson it.
And you can, clean the data aswell.
The data visualization,algorithm development.
I'll get into that a little moredetail, but those are also going
to be, pretty importantdepending on the approach you're
taking.
So data preparation And if wethink about one, steps one
(37:25):
through five here, a lot of thisso far would be kind of the
precursor that Heather wasdescribing.
You kind of, you have to get toa point where you have your data
collected and cleaned up inorder to do much of anything,
even if you're not trying to dopredictive analytics, this is
something that you would do justto get General business
(37:45):
intelligence on theorganization, right?
So it's a step that's well worthit to go through or these steps
are really well worth it Even ifwithout the predictive analytics
part as the goal and the datapreparation side is really
coming up with a structure thatunifies the data from the
different systems So if you havecustomers in the customer
success System that havedifferent names than the
(38:06):
customers in the sales systemand the crm You have to have
some way of preparing that andblending those datasets
together.
So you can see from a termperspective, potentially, what's
their customer, what's thecustomer support activity and
success activity matched againstthe sales and the pipeline that
we have in the CRM.
And actually, Chad, I want tobring one thing up here.
Just again, knowing the types ofbusinesses that we work with.
(38:30):
There are so many tech tools andplatforms that are being used to
support events and media wherethis data is sitting for the
same customer understanding howto create that taxonomy, that
single source of truth.
While still moving your businessforward with all of those
platforms, it is a real bigfactor in this data prep part.
(38:51):
And there are definitely ways todo it without interrupting the
business flow or the sourcedata.
But that is something that I seestopping a lot of people because
of the enormity of it all, butthere are definitely are
solutions.
So step seven, six and sevenhere.
This is probably where you endup without internal expertise.
(39:13):
You might stall out a littlebit.
So this is going back to thefirst couple of steps.
What are our objectives?
What data do we have and how dowe map that to our objectives?
Now you have to figure out,like, how do we go about
actually creating the algorithmor creating the process, right?
Without having a lot ofexpertise or experience in this,
(39:33):
in doing this from scratch,you're going to, again, you
might need to go out externallyto get expertise to set it up as
a one time process that you canthen continue to run over time.
But I would also say that,depending on the nature of the
objective There are tools in themarket that can do this and have
determined the right model,determine the right features or
have a mechanism to determinethe features automatically and
(39:55):
by feature.
We mean thinking again about theeasy one that everyone knows on
the sales side.
I go back to a story from ourown business way back when we
were starting.
We had a little smaller set ofdata, obviously, as we were just
But we decided to do a kind ofan experiment and review of our
CRM and the deals that we hadclosed, the deals that we had
(40:16):
lost.
And in doing it just as a onetime evaluation, we were able to
identify three key attributes onour sales.
So on the opportunities datapoints, we collected that led to
a greater than 90 percent closerate.
So those three attributes, thenwe're pretty much our sales
method and process goingforward.
And we use those to close,business at a higher rate.
(40:38):
And continue to kind of keep aneye on those attributes.
And that's kind of what you meanby the feature selection.
It's like what data points arereally impactful.
Give it to that drive theoutcomes statistically.
So again, there are tools thatif you're looking for certain
outcomes or certain predictivemodels, there are tools that are
built in the market that willhave some of this all ready to
(41:00):
find.
If you have something that'svery unique and very niche in
your space, or you have datathat's very complex then you
might need to go and build moreof it from scratch.
And that's where you have to getinto more of the technical
expertise.
So here on the next step, afteryou've kind of defined your
model, you're really going totrain it and kind of validate
(41:21):
the results, right?
You might not have picked theright approach, might not have
picked the right data sets, thedate, the attributes, this step.
Is a means to kind of figurethat out and to over time
determine what is the mostoptimal model and most optimal
data set to use to generate theoutcomes.
And so you know, just simply,you kind of, you can think of it
(41:42):
where you're, you're splittingyour data sets.
into some that are fed throughthe model and you validate
against the model and then someyou are fed and you train the
model on and then you ideallywant to see the model work well
with new data that it hasn'tseen before.
And that's the process thatyou're overseeing here.
So that's, it's going, going toan example that most people
probably know with, which chapwith chap, she, Bt, they trained
(42:06):
their models internally.
You had a ton of people doingit.
Obviously, and then theyreleased it in the wild, and,
they're getting feedback basedon its performance that
incorporates itself back in tothe model to get better over
time.
And that's kind of touched onhere as well.
So the evaluation, youobviously, you want to see it
work well against the trainingdata, but also in the new data
(42:26):
as it comes in.
So there are methods fordetermining, how accurate is
this or how reliable is thisgiven the different nuances we
see in the data.
The step 10 iterativeimprovement again, you
oftentimes won't get itcompletely right on the first
time or in the first try.
And so you kind of, you need totake this as a, known factor
(42:46):
going into the project, into theinitiative, you're going to have
to, adjust over time continue tolook at the performance and see
if it's.
Continue to perform well overtime and make changes if needed.
So on step 11, educate andcommunicate really important
here.
If you want actual adoptionwithin the business, a good
example from our experiences, insome cases, if you define the
(43:10):
model and define the outputs andthey're very obscure or very
difficult to comprehend.
In a dashboard or in a report oranything like that and you try
to distribute that to peopleThey would just they just won't
use it.
They won't believe it.
They won't understand it andYou'll have a bunch of wasted
effort.
So you really have to make surethat the outputs are something
(43:30):
that are easily digestible justin the visual and report
presentation form, but you alsoneed to educate them on how is
it, how was it constructed?
What factors do we use todetermine the future sales or
the churn risk?
As soon as they understand that,then they can have more faith in
the model and have more faith inthe outcomes in the outputs, and
(43:50):
they'll be able to adopt it.
But if you don't do that it'srare that folks will really
trust it from the outset.
And I'll just pipe in here.
What I've seen is if that's notdone well, then all of a sudden,
well, that's that's data iswrong, so it just gives someone
if it doesn't say what theylike.
So this part is, I think, supercritical and then after this,
(44:12):
you get to go on and do more.
So if you pick something narrowand very, contained, very Again,
the idea is that you can scalethis to other areas.
And we've seen that within ourcustomers time and time again,
we set up something for them.
They love the output and thenthey might start to do things on
their own, or at the very leastsuggest other ideas.
(44:33):
And areas within the businessthat they can apply these
techniques to.
And so it's, from start smalland then iteratively kind of
scale up and have the teamwithin the business, take on
some of that some of thatexploration and model
development.
Any questions about the process?
(44:55):
All right.
Again, if you want to talk to usoffline, we're here.
So my favorite part is alwaysthe landmines.
We, the landmines are andlandmines happen before, during
and after.
So they're really important tounderstand.
But Chad, you want to kick thisoff?
Sure.
Just a couple of that.
I'll call out.
(45:15):
So not enough data.
Thank you.
If you have and, ambitious ornot, maybe not even ambitious,
objective you need to make sureyou have enough data to actually
make it statistically possibleto produce predictive outputs.
We, there It's a little bit of achicken and egg, but.
You do want to make sure youhave sufficient historical clean
(45:36):
data, in your single source oftruth that you can run the
models through.
And that's oftentimes forsmaller organizations.
That's where that's really wherethey aren't able to leverage.
Predictive analytics as much asbecause there's just not enough
data to make use or definepatterns.
But, that's one I would saydiscarding models too early as
well.
You might see outputs early onthat don't make any sense.
(45:57):
It doesn't mean that everythingis wrong.
It just means that you mighthave it off.
You might have it configuredslightly off.
You might have, a couple inputsthat are not clean, more couple
inputs and weights within themodel that aren't working.
And so it's.
It's don't try to don't quit tooearly on.
When you're starting theprocess.
Yeah.
And I think the other one is andwe touched on all of these, but
(46:18):
I think that the big one is theactivation plan and process to
activate and how you're going totrack and measure.
it's very important to show yourboard, your CFO the value that's
being added to the business forthese investments.
But there's a lot of culturechange and process change that's
required.
So thinking that through aheadof time is critical and not
(46:40):
doing it is definitely a way tohave that landmine.
Go off.
I also think that not proceedinguntil the data is perfect.
No data is perfect.
And especially in the businesseswe're talking about it is
there's so much data.
But you have to start somewhere.
So figure out where that is, andyou'll perfect as you go.
Capabilities are evolutionary.
(47:00):
And I would say the the assumingeveryone is on board, or I like
to call the saboteur in theroom.
You've got to identify theperson or people who are...
are not willing or really tryingto stop progress from happening
in your business through the useof data.
I personally see it as theemperor's clothes.
(47:20):
They don't want to, they don'twant to expose what's really
happening.
But really finding that out andunderstanding why any resistance
and why, democratized.
data environment and everyone isbetter at their jobs when they
have data at the ready.
Those are just a few of thelandmines.
(47:41):
I don't know if any of you areencountering any that you'd be
willing to share, but we have acouple left but that's our
presentation.
Any landmines that anyone wantsto dive into we've had a few
people have to leave, but therearen't any questions.
I will share these last twoslides.
There's our, the chat on myemail are there to take
(48:03):
advantage of the office hours.
You can email us.
I'm happy to sit down and talkabout where you are in your
journey.
And again, these are the waysthat we can help at H2K.
And then we have another bootcamp coming up on October the
19th, and we have Denise Medved,the Chief Commercial Officer.
She's new in the role atInformer Markets, and Matt York,
(48:25):
who's the CRO of Foundry, and itwill be actually a really
interesting discussion becauseInforma is very heavy on the
event side.
Foundry is 80 percent digitallead gen media.
And they're both doing somereally exciting things in terms
of a single source of truth andpursuing revenue excellence
through data.
(48:46):
So with that, I would say ifthere's no more comments or
questions, you'll get fourminutes back to your day and
we'll send everyone therecording.
Great.
Thank you so much, everyone.
Thanks, Heather.
Heather Holst-Knudsen (48:59):
You can find us@2klabs.com.
Thank you.
Thank you for listening to TheRevenue Room by H2K Labs.
Subscribe to our channel today.
Popular Podcasts
United States of Kennedy
United States of Kennedy is a podcast about our cultural fascination with the Kennedy dynasty. Every week, hosts Lyra Smith and George Civeris go into one aspect of the Kennedy story.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com