July 3, 2025 • 41 mins
I am always looking for ways to improve. Send me a text and let me know your thoughts! - Kevin
What if 90% of your marketing budget is being stolen right in front of your eyes? In this eye-opening conversation, Dr. Augustine Fou—MIT PhD and leading expert in digital ad fraud—reveals the shocking truth about how bots are quietly draining millions from marketing budgets, including those in B2B tech.
Having worked in digital advertising since the mid-90s, Dr. Fou explains how programmatic advertising created the perfect storm for fraud to flourish. Bad actors create thousands of websites with plagiarized or AI-generated content, purchase bot traffic, and sell ad impressions to advertisers at a profit. The fraud takes many forms: fake impressions, fraudulent clicks, form-filling bots that generate leads, and even attribution manipulation that makes ineffective channels look successful.
Most alarming for tech marketers is learning that performance-based campaigns aren't immune. "The bots will do exactly what you pay for," explains Dr. Fou. If you're paying for leads, sophisticated bots will complete your forms and solve CAPTCHAs, leaving you with seemingly legitimate but worthless leads. Even worse, traditional metrics like viewability and attention are easily manipulated by fraudsters.
Dr. Fou introduces a powerful alternative metric: "attentiveness," which measures actual human interaction after someone clicks through to your landing page. Unlike easily faked metrics, attentiveness captures real engagement through mouse movements, scrolling, and clicking—behaviors that bots typically don't replicate convincingly.
The conversation offers practical advice for protecting your budget: watch for suspiciously high click-through rates, monitor for leads completed in humanly impossible timeframes, and add invisible form fields that only bots will complete. The potential upside is enormous—companies could save up to 90% of their digital ad budgets by eliminating fraudulent spend.
Ready to stop throwing money at bots? Follow Dr. Augustine Fou on LinkedIn where he's published over 800 articles on digital ad fraud or at his company, Fou Analytics.
🎧 Tech Marketing Rewired is hosted by Kevin Kerner, founder of Mighty & True.
New episodes drop regularly with unfiltered conversations from the frontlines of B2B and tech marketing.
Catch the video version of the podcast on our YouTube channel.
👉 Subscribe, share, and connect with us at www.mightyandtrue.com
📩 Got a guest idea or question? Email us at kevin@mightyandtrue.com
🔗 Follow Kevin on LinkedIn for more insights and behind-the-scenes takes.
📰 Subscribe to Kevin's Substack at: Subscribe to Kevin's Substack at: https://kevinkerner.substack.com/
📰 Want more tech marketing insights? Subscribe to the Drag & Drop newsletter for fresh thinking, real talk, and tools that make your job easier.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Kevin Kerner (00:00):
Hey again everyone .
This is Kevin Kerner with TechMarketing Rewired.
What if I told you that 90% ofyour ad impressions might never
be seen by a human and no onewants to talk about it?
In this episode, I spoke withDr Augustine Fu, a leading
expert in digital ad fraud andmedia transparency.
We dug into how ad fraudactually works, why many brands
are flying blind, the differencebetween attention and
(00:23):
attentiveness, and how bots arequietly draining millions from
marketing budgets, even in B2B.
Let's get to it.
This is Tech Marketing Rewired.
Hello everyone, and welcomeback to Tech Marketing Rewired.
Today's guest is Dr AugustineFu.
(00:43):
Dr Fu is one of the mostrespected experts in digital ad
fraud and media transparency.
He holds a PhD from MIT andpreviously served as chief
digital officer at bothOmnicom's Healthcare Consultancy
Group and McCann Worldwide.
Today, he runs Fu Analytics andworks as an independent
cybersecurity advisor, helpingbrands uncover and fight fraud
(01:04):
in their digital media.
Dr Augustine, welcome to theshow.
Thank you, kevin, glad to behere with you.
Yeah, I am really excited tohave you here.
Just personally, I wanted tolearn more about this space,
particularly as it relates toB2B and tech.
I've been reading your researchand following you on LinkedIn
and I really would love to getinto the various forms of
digital ad fraud.
(01:25):
But first I wonder if you couldgive us a quick introduction of
yourself and then, fooAnalytics, tell us what you do
All right.
Dr. Augustine Fou (01:33):
Well, I've been in digital advertising
since the very beginning of mid90s, so to speak.
I was at McKinsey Company andwe built a little lab in the New
York office with two Applecomputers, two IBM PCs and two
Sun workstations.
So that was right at the verybeginning, when web pages were
simply pages of text with a fewwords underlined that you could
(01:55):
actually click on.
It would bring you to the nextpage, called hypertext.
So it was since that time welooked at the potential impact
of the internet on entireindustries and obviously we've
seen that come to fruition.
It's changed so many industries, including ours, the digital
advertising industry.
So I've been in it for a longtime.
(02:15):
I worked on both the clientside at American Express as well
as the agency side at both IPGHolding Company and Omnicom and
ultimately left about 13, almost14 years ago to get back to my
own consulting practice, andthat's when we started building
what is now known as FooAnalytics.
So back then we were seeing,you know, somewhat of a in a B2B
(02:38):
context as well.
We were serving pharmaceuticalclients and they were doing a
lot of search marketing to tryto get doctors to understand
their prescription drugs andthings like that, and we're
seeing a lot of clicks,sometimes even 200%
click-through rates right, justcompletely insane CTRs and
sometimes those clicks wouldkeep coming even after the
campaign was over.
(02:58):
So clearly something was wrong,but at the time no one could
really explain what was going onlike what was causing that.
So some of the tools I startedbuilding was to really start
auditing these campaigns andfigure out what was going on.
So in the very beginning wewere doing on-site measurements
and we would see the clickscoming from search campaigns and
(03:19):
other types of marketing and wecould tell these were obviously
bots clicking.
They were leaving right away.
They weren't doing anything onthe site.
So we were able to detect thebots and fast forward.
Now, 15 years, we've addedin-ad detection.
So Foo Analytics has bothon-site measurement as well as
in-ad measurement, which meanswe measure the ads themselves,
(03:43):
and it's all centered aroundfinding the bots and also other
forms of fraud that areimpacting these ads.
So that's been like 30 years ofmy life doing digital marketing
and the last 15 focused on thefraud problem.
That's pretty rampant indigital advertising, affecting
both the B2C side, which isconsumer marketing, and B2C side
(04:04):
, which is consumer marketing,and B2B side, which is marketing
to businesses.
Kevin Kerner (04:09):
Yeah, fascinating If you could get into what are
some of the forms of digital adfraud, just as the primer on
this stuff.
What type of fraud are youseeing?
Dr. Augustine Fou (04:18):
Yeah, so I'm gonna stick with the IAB
framework.
So they have four main types ofdigital ad revenue, so CPM, cpc
, cpl and CPA.
So I'll explain each of thesein order.
So CPM means cost per mil orcost per thousand, and these are
typically the way advertiserspay for display ads or video ads
(04:39):
.
Right, every thousand displayads you pay a certain amount, a
dollar amount.
So in this case the type offraud is simply CPM fraud, where
all the bots have to do is loadthe ad impressions right, and
if they load a thousand timesthey get paid.
Cpc is cost per click, so youactually have to click on the ad
, like a search ad, in order toget paid the CPC.
(05:01):
So the bots not only cause thead to load right, they type in a
search keyword, they see allthe search ads that load and
then they click on certain onesand so then they get paid for
the CPC.
The third kind is a cost perlead, cpl, and those are
typically like B2B marketers ormaybe, say, university marketers
(05:22):
, where they're paying on a costper lead basis, right?
So every time they get acompleted form fill, they pay
out.
You know, could be $5, could be50, could even be $500 per lead
, depending on the industry.
So in those cases we've seenthe bots complete those lead
forms and submit them.
Because the way you think aboutthe bots activity is they will
(05:43):
do the exact thing you pay for,right?
So if it's a cost per lead thatyou pay, they'll do exactly
that, nothing more, nothing less.
Right.
They're very, very efficient interms of their use.
And then, finally, there's CPA,which is cost per acquisition,
and you can usually think aboutthat in terms of affiliate fraud
, or now attribution fraud,right, that's after the sale
(06:05):
occurs, a revenue share getspaid on the purchase value, and
so the bots can also fake thatright.
So there's various forms offraud in every single revenue
model, right, every single adtype that the IAB has.
So that's kind of the contextof everything.
Kevin Kerner (06:21):
And then we can kind of dig into the details of
each.
Incredible what's in it.
I each Incredible what's in it.
I'm just curious what's in itfor the bad actors?
Dr. Augustine Fou (06:29):
Like why Just money, pure profit.
So I'll use a very simplescenario.
Right 15 years ago, when westarted adding many, many
websites to run ads in the veryearly days, you remember Yahoo
and that's like a big, hugeportal that had all sorts of
content on there from weather tosports scores, stock quotes to
news and all that kind of stuff.
(06:50):
So it aggregated large humanaudiences.
And if you buy ads on the Yahooportal, you got mass awareness
because lots of people saw it.
So since that time we'vestarted evolving and adding what
people call long tail websites.
So now we have millions uponmillions of long tail websites
(07:10):
that by themselves they don'tget a ton of traffic, but if you
put them all together, that'sagain the theory Right?
So if you have lots and lots ofthese small sites, the
collective traffic is enough tobe sizable.
According to that theory, younow have millions of long tail
sites that are now selling adsthrough programmatic exchanges.
So, just like Wall Street is astock exchange where buyers and
(07:33):
sellers of shares of stock cometogether at auction auction each
individual stock share, in theprogrammatic ad exchanges we do
the same with each ad impression.
So now every single adopportunity is auctioned off
Whoever wants to bid on that andwhoever bids the highest will
typically win, and then theyhave the right to serve the ad
(07:54):
into that.
Now what's happened is that inthe last 15 years, when
programmatic media buying becamethe dominant form of buying ads
, that's when the fraud exploded, because, for the very first
time, these long tail websitesthat nobody has ever heard of
before have the opportunity tosell ads to the largest of
(08:16):
advertisers, right.
Previously.
Just imagine a no-name companywalking into a P&G to say oh
yeah, we have a hundred billionimpressions to sell you.
P&g will kick them out the doorLike who the heck are you guys,
right, we've never heard of you, we're never going to buy from
you.
So back then, when the mediabuying took place between the
advertiser and a real publisher,like a Condé Nast, a Hearst, a
(08:39):
Meredith right, a Time Inc thatwas still relatively clean
because you could actually seewho you're buying from.
But now, when you're buyingfrom a programmatic exchange,
there's millions of sites thatyou've never, ever heard of that
are selling ads right, andthey're selling millions and
millions of impressions.
So, again because ofprogrammatic advertising, that's
(08:59):
when fraud exploded.
And now getting back to how thebad guys make money.
Bad guys would basically set up10,000 sites in a month, all
using WordPress templates andall using plagiarized content,
right?
Whether it's text or images.
They just go, scrape it off ofsomeone else's site and just
assemble it into their ownWordPress sites.
Now, these days, with AI, youdon't even need to steal content
(09:22):
, right?
You can just generate it.
Okay, so it's all beenhappening for 15 years.
But those long tail websitesthat nobody's ever heard of
would obviously have no humanvisitors.
So what do they need to do?
They literally have to go outand buy all their traffic.
So if you just Google, buytraffic from my website, there's
going to be tons of sellers,right?
(09:44):
Tons of vendors happily sellingyou traffic.
So it's the exact samephenomenon, as you can go on.
Buy views for your YouTubechannel, right.
Buy likes for your Facebookpage.
Buy followers for yourInstagram account right.
All of those things can bepurchased.
But obviously those are nothumans.
Those are bots, even if thelegacy verification vendors
(10:05):
can't detect them as bots.
But just use a little bit ofcommon sense and just think
about.
There's not a whole bunch ofhumans sitting around with
nothing to do but to go to yourwebsite when you tell them to.
That just doesn't happen inreal life life.
So when you can say, oh, I need10 million page views this
month, how much does it cost?
It is 100% manufactured by bots, and so that's how the bad guys
(10:34):
make money.
So they're buying traffic fortheir website, they're selling
the ads at a higher CPM thantheir cost of traffic and
they've made pure profit.
Right, it's just arbitrage.
And then there's all thevendors in between, but we won't
get into all those other things.
But like, there's a wholeecosystem to support the fraud.
Kevin Kerner (10:49):
Wow, just amazing.
I mean, what is wrong withpeople?
It's incredible the amount ofcomplexity and all that.
There's simplicity.
It's a little simple, butthere's a lot of complexity.
Dr. Augustine Fou (10:57):
Yeah, I think it boils down to you know, in
the physical world it's actuallymuch harder to fake stuff.
Like you know, in the physicalworld it's actually much harder
to fake stuff Like imagine ifyou had to make a counterfeit
Hermes handbag, ok, youliterally have to go make one
Right In digital.
It's all bits and bytes.
So you can literally register adomain for $7.99 on GoDaddy or
something Right.
(11:18):
So the ways of mimicking ormaking the illusion of something
is so much easier in digital.
And also you can fake all themetrics, by the way.
So you know, making acounterfeit things in digital is
so much easier and that's whyliterally everybody and their
brother-in-law are gettinginvolved.
Kevin Kerner (11:37):
Wow, Amazing.
I want to get into some of theschemes that are happening here,
but I wanted to ask you aquestion that came to mind
what's the?
Have you found a way toquantify the impact of this
across the industry?
And then, what's the averagelike effect if I'm a brand let's
say I'm a mid-sized tech brandand I'm running ads like what's
the monetary effect of this?
Do you have any?
Dr. Augustine Fou (11:57):
idea.
Yeah, I mean, if I say this outloud, everyone will think I'm
crazy, right?
Not that they don't alreadythink I'm crazy, but it's the
vast majority of this stuff,right?
Because the problem is, 15years ago the big advertisers
were saying, okay, there's suchlimited quantity of ads we can
buy from a legit publisher likeNew York Times.
Why is it so limited?
It's because humans' behaviordoesn't change that much, right?
(12:19):
When you go read an article inNew York Times, you probably
literally look at that onearticle and maybe one more, you
know, generate three to fivepage views and then you're out
of there, right, you leave.
So humans, visitation patternson websites just don't throw off
enough page views to generate,you know, the hockey stick of
growth that VCs wanted.
(12:40):
So they're always looking forwhere can I get more traffic?
Where can I get more adimpressions?
Right?
So then they started looking atprogrammatic exchanges and then
, magically, for the last 15years, we've had enough traffic
to satisfy as much demand as wewant.
So this also even gets to thefundamental economics principles
that you learned in school,right?
(13:01):
If there's a lot of demand,right, huge increase in demand
and there's only finite supply,the prices should go up.
Basic economic principles.
But what we've seen for thelast 15 years is that, even
though there's been massivedemand coming into digital, big
advertisers are taking dollarsaway from print, from newspapers
(13:22):
, from TV, and shifting it intodigital.
Despite this massive increasein demand, the prices went down.
That means the amount of supplywent up even faster than the
demand and you can basically sayall of that supply was
manufactured out of thin air,right?
Because you can literally tella bot okay, this week I need a
(13:45):
million page views, next week Ineed 10.
The following week, I need 100million page views.
It can all be manufactured outof thin air, right?
It's just bits and bytes, so inthat sense, you can think of it
as the majority.
Now, it's not evenly spread,because there are certain
advertisers that are extremelystrict and extremely vigilant in
(14:08):
their buying practices and theywill definitely be affected
less by the fraud.
But when you have a super largeadvertiser, say, for example, a
CPG company, that has way, waytoo much money to spend, right,
okay, they were given $2 billionto spend in digital, okay, how
do I go spend all of this?
They basically tasked theiragencies with an impossible task
(14:30):
.
Go spend it all for me, right,and the agencies happily do that
.
And when they go to exchangeafter exchange, publisher after
publisher, and ask how muchinventory can you sell me?
They literally will run out.
There's just not enoughinventory to be bought.
So then they start digging intoplaces they shouldn't be
digging into, like, oh, who elsecan sell me stuff?
(14:51):
All the bad guys will raisetheir hand oh yeah, we got a
whole bunch of impressions overhere to sell.
You, come on over.
So that's where stuff goessouth.
And even the middlemen, like theagencies, the exchanges
themselves or whatever, eventhough they're not the ones
making the bots, they'recertainly benefiting from the
existence of those bots becausethey need the volume to satisfy
(15:14):
all the demand.
So again, you can see how a lotof these things are misaligned
incentives, so to speak.
Right, so I'll kind of put thisout there.
Ad fraud is not a tech issuethat you can solve by throwing
more tech at it.
Ad fraud is an incentives issue.
If you don't want it solved,you can just go about your day
and just buy unlimited more adsthat you want.
(15:37):
But if you actually want itsolved, right, there's ways to
solve it and it's very, verysimple and straightforward,
right.
It's not about buying moreverification tech and throwing
more tech at it to try to solveit, because the bots will always
be better.
Kevin Kerner (15:51):
Wow, amazing.
Yeah, it's a really simpleexplanation you gave.
You're right, like if you're onthe New York Times or some
Wired or whatever.
There's only so much to goaround.
Dr. Augustine Fou (16:02):
Very finite, very limited inventory.
Yeah.
Kevin Kerner (16:05):
That is crazy.
I can't believe I had neverthought of that, but that is
really the truth.
Now in B2B let's say midsize tolarger tech brands- Even more
limited.
Yeah yeah, it's super limited.
Yeah, really incredible.
So now I want to get into someof the schemes that are being
used out there and I read someof your research on this.
(16:26):
It's really fascinating.
So can you take us through,like, some of the things that
they actually do and then maybeyour advice on how to catch
these things and what you shoulddo if you're trying, if you see
one of these things happening?
Dr. Augustine Fou (16:38):
Yeah.
So I'm going to start withsomething very simple and you
can kind of understand how itevolved, right, and that's
basically click fraud.
And then I'll end with the formfills, right, the cost per lead
and all that kind of stuff andtalk about how performance
marketers should not assume thatthey're immune from fraud.
Right, a lot of B2B performancemarketers say, oh, we don't
care because we're only payingfor the lead, okay.
(16:59):
Well, is that lead real?
Okay, so I'll get to that in asecond.
Seriously, let me go back toclick fraud, because that's the
most basic form and we saw this15 years ago and that's kind of
what got me into the fraudresearch, right.
So when you put up a banner adon Yahoo, you know you're
basically crossing your fingersand hoping that someone's going
(17:19):
to click on it, because thatwould indicate that there was
some interest in it.
Right, and when you clickthrough on a banner ad, you come
to the site and that kind ofstuff.
So in the very early days,clicks and click rates were used
as a proxy for success.
Right, the more clicks you got,okay, the campaign's working
really great.
And in the early days that wasa good indicator, right Before
the bots got involved, right?
(17:40):
So when a human deliberatelyclicked on an ad, that means
they actually were interested,and when they get to your
landing page, they'll probablylook around at your content and
maybe even buy something fromyou.
Okay, so that was perfectlyfine in the good old days.
But because a lot of peoplestarted focusing on the clicks,
the bots then said, oh well, whydon't we just click on stuff to
make it look like it'sperforming?
(18:01):
And that's how we can trick theadvertisers into giving us more
money, right, allocating morebudget to those sites.
So, even with search ads, inthe very beginning, there are
ads that ran on Google's mainproperty search ads, and there's
also something called thesearch partner network, which
were all of the outside websitesthat also ran Google search ads
(18:21):
for revenue.
Right, so they would get partof the revenue.
Google would get the other part.
Right, they'd do a revenueshare.
Now, because of that, it createdthe incentive for those sites
to cheat.
We could make more money whenthere's more clicks.
So if there's not enough humansclicking right, because, again,
humans don't click on searchads more than say 1% and humans
(18:42):
don't click on display ads morethan say 0.1% right, one in a
thousand.
So how do we make that numberlarger?
Okay, we start cheating.
We use bots because we can makethe bots click on 100% of the
stuff, because we need to clickin order for us to make the
money.
So it created the incentive.
So now the bad guys would setup all these fake websites.
(19:04):
So very early on in our research, we saw countless sites that
were solely based on searchkeywords.
So not only the domain wouldcontain those, but all the
content on the pages wouldcontain search keywords and
things like that.
So when people got there andthen they clicked on something,
that's how those sites makemoney because they're part of
(19:24):
the Google search partnernetwork.
So that created a lot of clickfraud in the very beginning and
that was success.
And when we saw more clicks,the advertisers said, oh, while
these campaigns are working,awesome, let's now move more
money into digital and pay forthese things.
Now, when you actually lookedunder the hood a little bit more
, right?
So one of the earliest examplesfor a pharma company is we saw
(19:47):
a blended average click rate of9%, greater than 9%.
Okay, that seems odd, becauseeven humans don't click on
search ads that much.
Right, it's higher than displayads, but not at 9%.
Okay, that seems odd, becauseeven humans don't click on
search ads that much.
Right, it's higher than displayads, but not at 9%.
And then, when you startedunpacking that again, this goes
back to just common sense wesaid, okay, well, can you show
(20:07):
us a report on a domain bydomain basis?
Right, we don't want to see thebig average, because that's
average across everything.
Once we broke out domain bydomain, we saw a whole bunch of
domains pegged at 100%click-through rates.
That doesn't make any sense,right, and that's tens of
thousands of ad impressions andclicks.
Well, broke out with sufficientdetail.
(20:30):
Right, the average is kind ofhide those details from you,
right, but when you actuallybreak it out and see line by
line, how the heck are some ofthese websites getting 100%
click-through rates?
Okay, oh, by the way, these alllook like fraudulent sites when
you actually visit them.
That's how we starteduncovering the fraud, even
without any specializedtechnologies, right, it's really
(20:51):
about asking for enough detailsso that you can kind of break
through those averages, becausethe averages hide the fraud very
easily for you or from you.
So when you look at the details, you can already say, okay, we
got to turn off all of thesesites because they're completely
fraudulent.
Now, that was very simple andthat was rampant amounts of
click fraud more than 20 yearsago.
(21:12):
And then now fast forward totoday and let me kind of get
into the opposite end of thespectrum, which is the
performance marketers and a lotof B2B ones are performance
marketers right, they'll say weonly pay for the lead, we're not
going to pay for the impression, we're not going to pay for the
click, we only pay when we getthe outcome that we want, which
is a lead.
Okay, so, like I said earlierin this podcast, the bots will
(21:34):
do exactly what you pay for,right, what you want to pay for,
in this case, the leads.
So, when you have a lead form,they will fill it out completely
, submit it, right, they'll evensolve the CAPTCHA and then
submit it to you.
So now you're paying, you know,$5 per lead, $50 per lead,
sometimes $500 per lead.
Now, in those cases, let meagain tell you how we use very
(21:58):
simple common sense to detectthis.
If you see a lead form submittedin one second, okay, is it even
humanly possible for a human tofill out all the forms fields
in your form and submit it inone second?
Okay, obviously not.
But the bot has all of the dataright From a database somewhere
(22:19):
of stolen information.
By the way, to complete yourform and solve the captain,
submit it in one second.
Just by having basic code onpage we can say, okay, this
doesn't make any sense.
So therefore this is a fakelead.
Even if every field in there,right, the zip code matches the
city, the phone number area codematches the city, all that kind
(22:40):
of stuff, Everything matches up, everything looks right.
But it's not real.
It wasn't submitted by a human.
Another very simple thing thatadvertisers are doing is just
adding invisible fields to theform.
A human can't see it, so theycouldn't possibly have filled it
out and submitted it.
A bot will see every form,every field on that form, and
submit it.
So when you get these formfills that have those fields
(23:03):
completed, you know definitely ahuman didn't do it, so then you
can just discard those outright.
But those are varioustechniques.
I'm sure the bots will getbetter.
We talk about AI agents buyingstuff for you or filling out
forms.
Even the purchase can be doneeasily by bots, right?
You've seen these cases whereTaylor Swift's tickets were
basically all purchased upwithin 15 minutes of the launch
(23:26):
by scalper bots right, the botswill buy all the tickets so that
scalpers can sell them for fivetimes the money on the
aftermarket.
Right.
So bots can do anything that ahuman can do in a browser.
It's just whether they have afinancial motive to do that
thing.
If they're paid for that, theywill do exactly that.
Kevin Kerner (23:45):
Golly, you know what I love about what you're
saying here.
It's so what you're asking, thequestions you're asking are so
simple.
Like could it be?
Would it be possible to havethat like in a second, or would
it be possible?
It's really.
I guess there's so muchbelievability behind the ad
industry.
You're just like what.
Dr. Augustine Fou (24:03):
They said it was someone looked at this.
They just didn't ask thosequestions.
Kevin Kerner (24:07):
But, to me we're not asking the right questions.
You know, one thing I lovedabout the stuff that I read
before our call was that youwere asking the question, or
you're making the point thatthere's a difference between
someone actually doing somethingwith an ad versus just seeing
an ad.
It's not just the attentionside, it's the attentiveness
side.
Can you break that down a bit,because I think it was really
(24:27):
that's the point.
You know what you need todetermine if someone has
attentiveness with the ad.
Is that correct?
Dr. Augustine Fou (24:34):
Yeah.
So let me kind of put it incontext, because a lot of people
have heard of the termviewability.
They've also recently heard theterm attention and I think a
lot of advertisers areoptimizing for attention.
I'm going to put a third terminto this mix and I'm going to
call it attentiveness.
But before I get toattentiveness, let me kind of
set the context by talking aboutthose first two.
(24:54):
So viewability is a very basicconcept and it kind of makes
common sense as well.
It's the viewability of the aditself, right, so whether the ad
was viewable.
So the main concept and there'ssome standards around this now
which is 50% of the pixels ofthe ad in view for one second,
because the common logic willsay if the ad itself wasn't even
(25:14):
viewable, then people couldn'thave seen it, it couldn't have
had any kind of impact whatever.
It makes total sense.
So it basically means thatadvertisers want to buy viewable
ads versus non-viewable ads.
Now, that's a good idea intheory, but what has happened in
the real world is that the badguys have easily faked the
(25:35):
viewability measurements orthey've basically stacked all
their ads above the fold so theyregister as 100% viewable even
if they're not right.
When you stack ads.
Say, 50 ads on top of eachother, only the top one is
visible or viewable right, theother 49 are behind it.
So obviously, a human can't seeit, even though the detection
(25:56):
tech will all register it asviewable okay, because it was
above the fold in the viewportand all that kind of stuff.
So, again in digital, there'seasy technical ways for the bots
or bad guys to fake any metricthat you want to see.
And in 2018, there was a casewhere Newsweek a mainstream
publisher right, supposedlypublisher was using very simple
(26:16):
JavaScript code to just falsifythe viewability measurement to
make all of their ads look likethey were 100% viewable.
So, when the advertisers thinkthey're buying 100% viewable ads
, they got to ask the nextquestion okay, were they
actually viewable?
Okay, so, that being said, wenow move to what you just
mentioned attention, which isokay.
(26:37):
Here's the next evolution ofviewability.
Right, the viewability has todo with the ad itself.
Attention, which means theydrive more business outcomes,
(27:07):
makes total sense, because alarger ad that gets your
attention will obviously performbetter and drive more business
outcomes than a small ad thatyou may not have even seen.
Like you skipped it, or it wastoo.
It was on the bottom.
So you didn't even look down atthe bottom of the screen or
something, right?
So higher attention meansbetter outcomes, makes total
sense.
The problem that a lot ofpeople don't really understand
(27:31):
is that you can't actuallymeasure attention in real life,
in real campaigns.
Right, you can measure it in aneye tracking study in a
laboratory environment, but youcan't measure it in real life
because the JavaScript is notallowed to turn on your camera
to see if you were actuallylooking at the screen.
Right, it would throw a prompt,a user prompt, security issue
(27:52):
if it tried to access yourcamera.
Yeah, no one's going to do thatno one's going to do that.
And that's what I meant byattention.
Vendors can't actually measureattention because they can't
show you that the person wasactually looking at the screen
the moment the ad showed up.
So then the optimizations thatend up happening right when
you're optimizing for higherattention means you're basically
(28:13):
optimizing for larger ads,because an ad that covers up the
entire screen obviously getsmore attention than an ad that
doesn't cover up the entirescreen, that's hidden on the
very bottom right, a mobile ador whatever.
So basically that's still fine,right Directionally.
More attention gets you moreoutcomes, that's totally fine.
I'm not saying don't use it,but there's a better way and I'm
(28:34):
a scientist so I want to makesure I can measure it.
So that's where attentivenesscomes in.
And attentiveness doesn'thappen on the ad and doesn't
happen on the person's device,because we can't turn on the
camera.
Attention, attentiveness happenson the advertiser's landing
page.
That's the page to which theuser clicks after they click the
(28:55):
ad.
And, like I said earlier, if ahuman deliberately clicked an ad
, that means they were kind ofinspired by it and they were
curious about something.
That means they were kind ofinspired by it and they were
curious about something.
So when they clicked on it,when they get to your website
they're probably going to dosomething like look at the
content, move around on the page.
So there are certain humansignals, like mouse movement,
(29:16):
page scrolling, click events,maybe touch events if it's on a
smartphone that we can actuallysee and directly measure on the
landing page of the advertiser.
So attentiveness is a verysimple concept.
It's basically the collectionof these human interaction
events mouse movement, pagescrolling, clicks, touch events
and the percentage of users thatactually did that when they got
there.
So if the majority of the usersthat clicked on your ad and got
(29:40):
to your landing page actuallydid something on your landing
page, we call that highattentiveness, and when maybe 1%
of the users that got to yourlanding page only did something,
we call that very lowattentiveness.
So you can now use attentivenessto gauge the relative
(30:00):
effectiveness of your ads, andattentiveness actually takes
into account both viewabilityand attention already, because
the ad would have had to havebeen viewable and then the
person would have had to payattention to it and then click
on it and attentiveness on thelanding page.
That's a good thing, because ifthey came to your site or
(30:29):
active on your site and didsomething.
That's a step towards aconversion.
So higher attentiveness on theadvertiser's landing page is
better than lower attentivenessor no attentiveness on the
landing page.
And in fact, zero attentivenessis the same as what you might
see in your own Google Analyticsas 100% bounce rates or zero
(30:50):
time on site.
Right, they left right awaywithout doing anything.
So I think in that case, justto kind of reiterate viewability
, attention and attentiveness, Iwould prefer to measure
attentiveness on the landingpage because it already takes
into account viewability andattention.
Kevin Kerner (31:08):
So simple and so logical.
These are really easy questions.
Dr. Augustine Fou (31:12):
Yeah, I've been thinking about this for a
long time, you know there aresimple things, non-technical
things that advertisers can doto make their B2B campaigns way
better.
Kevin Kerner (31:22):
Yeah, it's really a blind spot.
You know it's a blind spot for,I'm sure, a lot of CMOs, a lot
of people in performancemarketing, because you're just
trusting that the system works.
It has to work.
You know, it's been around forso long.
The one thing I wanted to askyou very quickly about is
attribution.
We get a lot of skepticism onattribution from technology
marketers and I think it's youknow, part of it.
(31:43):
They're trying to say to theirexecutive leadership team here's
the stuff that's working, solet's spend more.
And then part of it isperformance based.
It's like what is working andhow am I going to?
You know what am I going toinvest in more?
How should CMOs or performancemarketers and B2Bs start
thinking about attribution?
Like how should they trust itat all?
(32:03):
How?
Dr. Augustine Fou (32:03):
do they validate performance?
I would say they can trust it,but they should be skeptical.
Right, there should be ahealthy, healthy amount of
skepticism.
And let me kind of explain why.
Right, so the concept ofattribution is attributing sales
and outcomes to the ad thatcaused it right, or the ad
impressions that caused it right.
That's great, that's fine.
(32:24):
But what has happened is that alot of these attribution models
are literally just that they'remodels, right, they're
mathematical models,methodologies to attribute sales
.
So let me use a very simple wayto explain this.
Right, you've all heard ofattribution windows, right?
And those can be set to oneweek to 30 days, 60 days,
whatever.
The attribution window is thetime that you would basically
(32:48):
look at after the ad exposure tosay, ok, if any sales occur in
that window of time 30 days, 60days, whatever then we can
attribute that sale or thatoutcome event to the ad exposure
, the ad exposure.
Now what has happened is thatpeople can very easily game the
(33:10):
attribution reporting to make itlook like they got a lot of
outcomes, simply by changing theattribution window.
So say, for example, you setthe window to one week, you have
ad exposure, you wait for sevendays and see if any sales
occurred.
If it did, then you wouldattribute.
And then you're a marketer andsay, oh, there's actually too
few sales that occurred in thatfirst week.
Why don't we wait for longer?
Why don't we wait two weeks?
(33:31):
Why don't we wait for 30 days?
See if any sales occurred.
So it kind of depends on theproduct, right?
So some products take longer tosell, especially B2B.
Those are very long saleswindow.
By adjusting the attributionwindow, they can now say we just
got a lot more sales attributedto these ads simply because
they waited longer.
Now what happens if you reallywant to cheat and say, okay, why
(33:52):
don't we make it 60 days?
Why don't we make it 90 days?
Okay, how about we make it halfa year?
So any sale that happened inthe half a year we can now
attribute back to the adimpression.
Yay, okay.
So you see how easily the ROASor ROI reporting, the
attribution reporting, can betricked to be made to appear
(34:14):
better, right?
So that's where the marketershave to really understand their
particular vertical, theirindustry, their product category
, whatever.
So certain products, like smallticket items, attribution
window should be very, veryshort.
Can of soda, very short.
B2b it could be six months.
Kevin Kerner (34:30):
That could be a legit attribution window, and
it's a very complex journey too.
Dr. Augustine Fou (34:35):
Yeah, exactly A lot of people involved, and
they're interacting with allkinds of stuff, so the marketers
need to be honest withthemselves in setting the
attribution window, but that'swhat I meant by you know, if you
let someone else set theattribution window, or you don't
even know what the attributionwindow is because the platform
just gave you a ROAS report, youcould be very easily tricked
into thinking something'sworking when it actually is not.
(34:58):
It's simply claiming credit forsales that either had occurred
in the past or would haveoccurred anyway, right?
It's simply a figment of howthey did the reporting, the
attribution reporting, right.
So it's not to say it's a badthing.
You just have to understand thelimitations and be very
skeptical yourself to make sureyou're getting the right numbers
, so that you understand whereto allocate more budget because
(35:21):
it actually is working, notbecause it looks like it's
working.
Kevin Kerner (35:24):
Yeah, really good advice, Really good advice.
The last question here, andthen I want to get into AI
roulette.
There's going to be a lot morecontent created and potentially
even like look what Meta isdoing with trying to make
advertising easy for the smallbusiness.
Is AI the?
Will AI be the problem, or isthere something in AI that will
be the solution in all this as afraud?
Dr. Augustine Fou (35:47):
researcher.
I'll tell you AI is going tocause more problems than it will
ever be the solution for.
And, very simply, you know,it's actually a useful thing for
small businesses, like if theyhave certain product pictures
right of the stuff that they'reselling.
It is trivial and it makestotal sense to let AI generate
the banner ads for you.
(36:07):
You don't need humans to dothat, so there are certain ways
you can use AI to your advantage, obviously.
So even automating some ofthose processes would not be a
bad thing for small and mediumbusinesses, who may not have a
creative agency to do that forthem.
They don't need to pay acreative agency to do stuff
that's going to be templatizedand easy to automate anyway.
However, like we said earlier,for 15 years, bad guys have
(36:31):
simply been plagiarizing contentfrom other places pictures and
texts and just putting it ontheir WordPress templates to
make fake websites.
With AI, they don't need toplagiarize it anymore.
Ai has already done it for them.
They'll just generate thecontent.
So the proliferation of morefake sites is just going to be
easier and easier going forward.
That just means that we'redemocratizing crime.
(36:52):
We're democratizing ad fraud,so more people can just sit in
their armchair at home and do adfraud.
But, that being said, if youreally understand that fraud has
been at the 90% level for thelast 15 years, a little bit more
of it due to AI is not going toreally matter.
Right, you're still going towhat you're going to get to like
92%.
Kevin Kerner (37:11):
That's going to be the problem, yeah just went up
by 1%, okay.
Dr. Augustine Fou (37:14):
So I think a lot of people think that ad
fraud was 1% for the last 15years.
So it's probably going to beshocking to them when they
realize 90% of their stuff isnot shown to humans.
And I think you asked thequestion earlier in this podcast
to say what can advertisers do?
They can probably save 90% oftheir digital budgets right now
(37:38):
and do better Talk about ROAS.
If you're saving that money,every dollar saved goes straight
to your bottom line, yourprofit.
So that's enormous ROAS thatmarketers can do.
That's the cheat code righthere.
Right, they can actually showway more ROAS simply by not
spending the dollar thanspending it on showing ads to
bots.
Kevin Kerner (37:58):
Yeah, incredible, absolutely incredible.
Do people really think you'recrazy?
Dr. Augustine Fou (38:01):
Absolutely right.
I've been a pariah for the last15 years and I'm totally okay
with that, because I think youknow, when we're talking about
this, there are going to be ahandful of marketers will say,
oh shit, we better do somethingdifferent, because all of our
assumptions may not have beencorrect for the last 15 years,
but there's going to be a wholebunch of them will say well, I
prefer doing what I'm doingright now.
(38:23):
I don't want to lose my job overthis.
So they're just going to notlisten to it.
But I'm fine with that too,right?
Kevin Kerner (38:28):
So and you and you have the weight of the whole
advertising industry.
Dr. Augustine Fou (38:32):
really, they're all in con right now.
Right they're.
They're all drinking rosé andpatting each other on the back
so.
I'm fine, I'm still here in NewYork, but.
Kevin Kerner (38:46):
I love it, I love my.
The rebel inside me loves thefight.
I just love the fight.
You're out, so I'm I'm a fan,all right, sure, okay, I warned
you about this.
I do an ai roulette question atthe end.
Dr. Augustine Fou (38:51):
So now we're gonna get to the hardest
question of all just the bestquestion, okay, and so let me.
Kevin Kerner (38:59):
so what I did is I loaded, uh, perplexity up with
your a couple of the researchreports, the questions that I
was going to think about askinghere, and your LinkedIn profile,
and so I'm going to push, sendon Perplexity and let's see what
it gives us here.
It's going to give you aquestion that you'll have to
answer on the spot.
Okay, here it goes.
Dr. Augustine Fou (39:17):
Oh man, it's really roulette.
You just actually submitted inreal time.
Yes, all right, okay.
Kevin Kerner (39:22):
All right, okay, it hasn't blown up yet.
Okay, but this could be theinaugural one.
So here we go.
If you had a magic button that,when pressed, would instantly
reveal to every CMO and brandexactly how much their digital
ad spend is wasted on fraud andwho in their organization or
agency has been knowinglyignoring it, do you think the
(39:43):
industry would actually changefor the better, or would it just
lead to mass chaos, fingerpointing and maybe even the
collapse of some of the biggestplayers?
Would you press it?
Why or why not?
Dr. Augustine Fou (39:55):
I wouldn't press it because from 15 years
of experience it won't doanything.
They will literally schedulenext year's con and just all go
to con and ignore it.
I've seen this for 15 years.
The story is not going tochange.
Kevin Kerner (40:09):
That is so awesome .
So the nuclear option.
There's no mutualself-destruction in any of this.
Dr. Augustine Fou (40:14):
It's like, okay, whenever they come around,
I'm going to be here.
Kevin Kerner (40:18):
That is so great.
That is so great.
Well, this has been amazing.
I mean, I've just been sittingback soaking it all in.
I'm so delighted to have you onthe podcast and to get your
information out, particularly tothe tech and B2B space.
This has really been fantastic.
Thanks, Kevin.
If people want to take the nextstep and get a hold of you or
(40:39):
get more information, what's thebest way to?
Dr. Augustine Fou (40:45):
information.
What's the best way to?
Most of my content is actuallyon LinkedIn, so there's 800 plus
articles from the past 15 years.
It's all published on LinkedIn.
So just find me on LinkedIn,follow me if you'd like and just
reach out directly.
Right?
It's just augustinfu atfuanalyticscom, f-o-u
analyticscom, happy to answerany questions.
I'm always happy to support anadvertiser who really is brave
enough to take a look moreclosely under the hood, because
(41:08):
that's really the first step.
Right, we can't solve fraudovernight, but we need these
baby steps and we need moreadvertisers to have the courage
to take that first step, and I'mhere to help.
Kevin Kerner (41:17):
Yeah, I love it.
Well, thank you.
Thank you so much, dr Fu.
We'll stay connected and keepfighting the good fight.
Thank you, kevin, good to seeyou.
All right, talk soon.
Hey again everyone .
This is Kevin Kerner with TechMarketing Rewired.
What if I told you that 90% ofyour ad impressions might never
be seen by a human and no onewants to talk about it?
In this episode, I spoke withDr Augustine Fu, a leading
expert in digital ad fraud andmedia transparency.
We dug into how ad fraudactually works, why many brands
are flying blind, the differencebetween attention and
(00:23):
attentiveness, and how bots arequietly draining millions from
marketing budgets, even in B2B.
Let's get to it.
This is Tech Marketing Rewired.
Hello everyone, and welcomeback to Tech Marketing Rewired.
Today's guest is Dr AugustineFu.
(00:43):
Dr Fu is one of the mostrespected experts in digital ad
fraud and media transparency.
He holds a PhD from MIT andpreviously served as chief
digital officer at bothOmnicom's Healthcare Consultancy
Group and McCann Worldwide.
Today, he runs Fu Analytics andworks as an independent
cybersecurity advisor, helpingbrands uncover and fight fraud
(01:04):
in their digital media.
Dr Augustine, welcome to theshow.
Thank you, kevin, glad to behere with you.
Yeah, I am really excited tohave you here.
Just personally, I wanted tolearn more about this space,
particularly as it relates toB2B and tech.
I've been reading your researchand following you on LinkedIn
and I really would love to getinto the various forms of
digital ad fraud.
(01:25):
But first I wonder if you couldgive us a quick introduction of
yourself and then, fooAnalytics, tell us what you do
All right.
Dr. Augustine Fou (01:33):
Well, I've been in digital advertising
since the very beginning of mid90s, so to speak.
I was at McKinsey Company andwe built a little lab in the New
York office with two Applecomputers, two IBM PCs and two
Sun workstations.
So that was right at the verybeginning, when web pages were
simply pages of text with a fewwords underlined that you could
(01:55):
actually click on.
It would bring you to the nextpage, called hypertext.
So it was since that time welooked at the potential impact
of the internet on entireindustries and obviously we've
seen that come to fruition.
It's changed so many industries, including ours, the digital
advertising industry.
So I've been in it for a longtime.
(02:15):
I worked on both the clientside at American Express as well
as the agency side at both IPGHolding Company and Omnicom and
ultimately left about 13, almost14 years ago to get back to my
own consulting practice, andthat's when we started building
what is now known as FooAnalytics.
So back then we were seeing,you know, somewhat of a in a B2B
(02:38):
context as well.
We were serving pharmaceuticalclients and they were doing a
lot of search marketing to tryto get doctors to understand
their prescription drugs andthings like that, and we're
seeing a lot of clicks,sometimes even 200%
click-through rates right, justcompletely insane CTRs and
sometimes those clicks wouldkeep coming even after the
campaign was over.
(02:58):
So clearly something was wrong,but at the time no one could
really explain what was going onlike what was causing that.
So some of the tools I startedbuilding was to really start
auditing these campaigns andfigure out what was going on.
So in the very beginning wewere doing on-site measurements
and we would see the clickscoming from search campaigns and
(03:19):
other types of marketing and wecould tell these were obviously
bots clicking.
They were leaving right away.
They weren't doing anything onthe site.
So we were able to detect thebots and fast forward.
Now, 15 years, we've addedin-ad detection.
So Foo Analytics has bothon-site measurement as well as
in-ad measurement, which meanswe measure the ads themselves,
(03:43):
and it's all centered aroundfinding the bots and also other
forms of fraud that areimpacting these ads.
So that's been like 30 years ofmy life doing digital marketing
and the last 15 focused on thefraud problem.
That's pretty rampant indigital advertising, affecting
both the B2C side, which isconsumer marketing, and B2C side
(04:04):
, which is consumer marketing,and B2B side, which is marketing
to businesses.
Kevin Kerner (04:09):
Yeah, fascinating If you could get into what are
some of the forms of digital adfraud, just as the primer on
this stuff.
What type of fraud are youseeing?
Dr. Augustine Fou (04:18):
Yeah, so I'm gonna stick with the IAB
framework.
So they have four main types ofdigital ad revenue, so CPM, cpc
, cpl and CPA.
So I'll explain each of thesein order.
So CPM means cost per mil orcost per thousand, and these are
typically the way advertiserspay for display ads or video ads
(04:39):
.
Right, every thousand displayads you pay a certain amount, a
dollar amount.
So in this case the type offraud is simply CPM fraud, where
all the bots have to do is loadthe ad impressions right, and
if they load a thousand timesthey get paid.
Cpc is cost per click, so youactually have to click on the ad
, like a search ad, in order toget paid the CPC.
(05:01):
So the bots not only cause thead to load right, they type in a
search keyword, they see allthe search ads that load and
then they click on certain onesand so then they get paid for
the CPC.
The third kind is a cost perlead, cpl, and those are
typically like B2B marketers ormaybe, say, university marketers
(05:22):
, where they're paying on a costper lead basis, right?
So every time they get acompleted form fill, they pay
out.
You know, could be $5, could be50, could even be $500 per lead
, depending on the industry.
So in those cases we've seenthe bots complete those lead
forms and submit them.
Because the way you think aboutthe bots activity is they will
(05:43):
do the exact thing you pay for,right?
So if it's a cost per lead thatyou pay, they'll do exactly
that, nothing more, nothing less.
Right.
They're very, very efficient interms of their use.
And then, finally, there's CPA,which is cost per acquisition,
and you can usually think aboutthat in terms of affiliate fraud
, or now attribution fraud,right, that's after the sale
(06:05):
occurs, a revenue share getspaid on the purchase value, and
so the bots can also fake thatright.
So there's various forms offraud in every single revenue
model, right, every single adtype that the IAB has.
So that's kind of the contextof everything.
Kevin Kerner (06:21):
And then we can kind of dig into the details of
each.
Incredible what's in it.
I each Incredible what's in it.
I'm just curious what's in itfor the bad actors?
Dr. Augustine Fou (06:29):
Like why Just money, pure profit.
So I'll use a very simplescenario.
Right 15 years ago, when westarted adding many, many
websites to run ads in the veryearly days, you remember Yahoo
and that's like a big, hugeportal that had all sorts of
content on there from weather tosports scores, stock quotes to
news and all that kind of stuff.
(06:50):
So it aggregated large humanaudiences.
And if you buy ads on the Yahooportal, you got mass awareness
because lots of people saw it.
So since that time we'vestarted evolving and adding what
people call long tail websites.
So now we have millions uponmillions of long tail websites
(07:10):
that by themselves they don'tget a ton of traffic, but if you
put them all together, that'sagain the theory Right?
So if you have lots and lots ofthese small sites, the
collective traffic is enough tobe sizable.
According to that theory, younow have millions of long tail
sites that are now selling adsthrough programmatic exchanges.
So, just like Wall Street is astock exchange where buyers and
(07:33):
sellers of shares of stock cometogether at auction auction each
individual stock share, in theprogrammatic ad exchanges we do
the same with each ad impression.
So now every single adopportunity is auctioned off
Whoever wants to bid on that andwhoever bids the highest will
typically win, and then theyhave the right to serve the ad
(07:54):
into that.
Now what's happened is that inthe last 15 years, when
programmatic media buying becamethe dominant form of buying ads
, that's when the fraud exploded, because, for the very first
time, these long tail websitesthat nobody has ever heard of
before have the opportunity tosell ads to the largest of
(08:16):
advertisers, right.
Previously.
Just imagine a no-name companywalking into a P&G to say oh
yeah, we have a hundred billionimpressions to sell you.
P&g will kick them out the doorLike who the heck are you guys,
right, we've never heard of you, we're never going to buy from
you.
So back then, when the mediabuying took place between the
advertiser and a real publisher,like a Condé Nast, a Hearst, a
(08:39):
Meredith right, a Time Inc thatwas still relatively clean
because you could actually seewho you're buying from.
But now, when you're buyingfrom a programmatic exchange,
there's millions of sites thatyou've never, ever heard of that
are selling ads right, andthey're selling millions and
millions of impressions.
So, again because ofprogrammatic advertising, that's
(08:59):
when fraud exploded.
And now getting back to how thebad guys make money.
Bad guys would basically set up10,000 sites in a month, all
using WordPress templates andall using plagiarized content,
right?
Whether it's text or images.
They just go, scrape it off ofsomeone else's site and just
assemble it into their ownWordPress sites.
Now, these days, with AI, youdon't even need to steal content
(09:22):
, right?
You can just generate it.
Okay, so it's all beenhappening for 15 years.
But those long tail websitesthat nobody's ever heard of
would obviously have no humanvisitors.
So what do they need to do?
They literally have to go outand buy all their traffic.
So if you just Google, buytraffic from my website, there's
going to be tons of sellers,right?
(09:44):
Tons of vendors happily sellingyou traffic.
So it's the exact samephenomenon, as you can go on.
Buy views for your YouTubechannel, right.
Buy likes for your Facebookpage.
Buy followers for yourInstagram account right.
All of those things can bepurchased.
But obviously those are nothumans.
Those are bots, even if thelegacy verification vendors
(10:05):
can't detect them as bots.
But just use a little bit ofcommon sense and just think
about.
There's not a whole bunch ofhumans sitting around with
nothing to do but to go to yourwebsite when you tell them to.
That just doesn't happen inreal life life.
So when you can say, oh, I need10 million page views this
month, how much does it cost?
It is 100% manufactured by bots, and so that's how the bad guys
(10:34):
make money.
So they're buying traffic fortheir website, they're selling
the ads at a higher CPM thantheir cost of traffic and
they've made pure profit.
Right, it's just arbitrage.
And then there's all thevendors in between, but we won't
get into all those other things.
But like, there's a wholeecosystem to support the fraud.
Kevin Kerner (10:49):
Wow, just amazing.
I mean, what is wrong withpeople?
It's incredible the amount ofcomplexity and all that.
There's simplicity.
It's a little simple, butthere's a lot of complexity.
Dr. Augustine Fou (10:57):
Yeah, I think it boils down to you know, in
the physical world it's actuallymuch harder to fake stuff.
Like you know, in the physicalworld it's actually much harder
to fake stuff Like imagine ifyou had to make a counterfeit
Hermes handbag, ok, youliterally have to go make one
Right In digital.
It's all bits and bytes.
So you can literally register adomain for $7.99 on GoDaddy or
something Right.
(11:18):
So the ways of mimicking ormaking the illusion of something
is so much easier in digital.
And also you can fake all themetrics, by the way.
So you know, making acounterfeit things in digital is
so much easier and that's whyliterally everybody and their
brother-in-law are gettinginvolved.
Kevin Kerner (11:37):
Wow, Amazing.
I want to get into some of theschemes that are happening here,
but I wanted to ask you aquestion that came to mind
what's the?
Have you found a way toquantify the impact of this
across the industry?
And then, what's the averagelike effect if I'm a brand let's
say I'm a mid-sized tech brandand I'm running ads like what's
the monetary effect of this?
Do you have any?
Dr. Augustine Fou (11:57):
idea.
Yeah, I mean, if I say this outloud, everyone will think I'm
crazy, right?
Not that they don't alreadythink I'm crazy, but it's the
vast majority of this stuff,right?
Because the problem is, 15years ago the big advertisers
were saying, okay, there's suchlimited quantity of ads we can
buy from a legit publisher likeNew York Times.
Why is it so limited?
It's because humans' behaviordoesn't change that much, right?
(12:19):
When you go read an article inNew York Times, you probably
literally look at that onearticle and maybe one more, you
know, generate three to fivepage views and then you're out
of there, right, you leave.
So humans, visitation patternson websites just don't throw off
enough page views to generate,you know, the hockey stick of
growth that VCs wanted.
(12:40):
So they're always looking forwhere can I get more traffic?
Where can I get more adimpressions?
Right?
So then they started looking atprogrammatic exchanges and then
, magically, for the last 15years, we've had enough traffic
to satisfy as much demand as wewant.
So this also even gets to thefundamental economics principles
that you learned in school,right?
(13:01):
If there's a lot of demand,right, huge increase in demand
and there's only finite supply,the prices should go up.
Basic economic principles.
But what we've seen for thelast 15 years is that, even
though there's been massivedemand coming into digital, big
advertisers are taking dollarsaway from print, from newspapers
(13:22):
, from TV, and shifting it intodigital.
Despite this massive increasein demand, the prices went down.
That means the amount of supplywent up even faster than the
demand and you can basically sayall of that supply was
manufactured out of thin air,right?
Because you can literally tella bot okay, this week I need a
(13:45):
million page views, next week Ineed 10.
The following week, I need 100million page views.
It can all be manufactured outof thin air, right?
It's just bits and bytes, so inthat sense, you can think of it
as the majority.
Now, it's not evenly spread,because there are certain
advertisers that are extremelystrict and extremely vigilant in
(14:08):
their buying practices and theywill definitely be affected
less by the fraud.
But when you have a super largeadvertiser, say, for example, a
CPG company, that has way, waytoo much money to spend, right,
okay, they were given $2 billionto spend in digital, okay, how
do I go spend all of this?
They basically tasked theiragencies with an impossible task
(14:30):
.
Go spend it all for me, right,and the agencies happily do that
.
And when they go to exchangeafter exchange, publisher after
publisher, and ask how muchinventory can you sell me?
They literally will run out.
There's just not enoughinventory to be bought.
So then they start digging intoplaces they shouldn't be
digging into, like, oh, who elsecan sell me stuff?
(14:51):
All the bad guys will raisetheir hand oh yeah, we got a
whole bunch of impressions overhere to sell.
You, come on over.
So that's where stuff goessouth.
And even the middlemen, like theagencies, the exchanges
themselves or whatever, eventhough they're not the ones
making the bots, they'recertainly benefiting from the
existence of those bots becausethey need the volume to satisfy
(15:14):
all the demand.
So again, you can see how a lotof these things are misaligned
incentives, so to speak.
Right, so I'll kind of put thisout there.
Ad fraud is not a tech issuethat you can solve by throwing
more tech at it.
Ad fraud is an incentives issue.
If you don't want it solved,you can just go about your day
and just buy unlimited more adsthat you want.
(15:37):
But if you actually want itsolved, right, there's ways to
solve it and it's very, verysimple and straightforward,
right.
It's not about buying moreverification tech and throwing
more tech at it to try to solveit, because the bots will always
be better.
Kevin Kerner (15:51):
Wow, amazing.
Yeah, it's a really simpleexplanation you gave.
You're right, like if you're onthe New York Times or some
Wired or whatever.
There's only so much to goaround.
Dr. Augustine Fou (16:02):
Very finite, very limited inventory.
Yeah.
Kevin Kerner (16:05):
That is crazy.
I can't believe I had neverthought of that, but that is
really the truth.
Now in B2B let's say midsize tolarger tech brands- Even more
limited.
Yeah yeah, it's super limited.
Yeah, really incredible.
So now I want to get into someof the schemes that are being
used out there and I read someof your research on this.
(16:26):
It's really fascinating.
So can you take us through,like, some of the things that
they actually do and then maybeyour advice on how to catch
these things and what you shoulddo if you're trying, if you see
one of these things happening?
Dr. Augustine Fou (16:38):
Yeah.
So I'm going to start withsomething very simple and you
can kind of understand how itevolved, right, and that's
basically click fraud.
And then I'll end with the formfills, right, the cost per lead
and all that kind of stuff andtalk about how performance
marketers should not assume thatthey're immune from fraud.
Right, a lot of B2B performancemarketers say, oh, we don't
care because we're only payingfor the lead, okay.
(16:59):
Well, is that lead real?
Okay, so I'll get to that in asecond.
Seriously, let me go back toclick fraud, because that's the
most basic form and we saw this15 years ago and that's kind of
what got me into the fraudresearch, right.
So when you put up a banner adon Yahoo, you know you're
basically crossing your fingersand hoping that someone's going
(17:19):
to click on it, because thatwould indicate that there was
some interest in it.
Right, and when you clickthrough on a banner ad, you come
to the site and that kind ofstuff.
So in the very early days,clicks and click rates were used
as a proxy for success.
Right, the more clicks you got,okay, the campaign's working
really great.
And in the early days that wasa good indicator, right Before
the bots got involved, right?
(17:40):
So when a human deliberatelyclicked on an ad, that means
they actually were interested,and when they get to your
landing page, they'll probablylook around at your content and
maybe even buy something fromyou.
Okay, so that was perfectlyfine in the good old days.
But because a lot of peoplestarted focusing on the clicks,
the bots then said, oh well, whydon't we just click on stuff to
make it look like it'sperforming?
(18:01):
And that's how we can trick theadvertisers into giving us more
money, right, allocating morebudget to those sites.
So, even with search ads, inthe very beginning, there are
ads that ran on Google's mainproperty search ads, and there's
also something called thesearch partner network, which
were all of the outside websitesthat also ran Google search ads
(18:21):
for revenue.
Right, so they would get partof the revenue.
Google would get the other part.
Right, they'd do a revenueshare.
Now, because of that, it createdthe incentive for those sites
to cheat.
We could make more money whenthere's more clicks.
So if there's not enough humansclicking right, because, again,
humans don't click on searchads more than say 1% and humans
(18:42):
don't click on display ads morethan say 0.1% right, one in a
thousand.
So how do we make that numberlarger?
Okay, we start cheating.
We use bots because we can makethe bots click on 100% of the
stuff, because we need to clickin order for us to make the
money.
So it created the incentive.
So now the bad guys would setup all these fake websites.
(19:04):
So very early on in our research, we saw countless sites that
were solely based on searchkeywords.
So not only the domain wouldcontain those, but all the
content on the pages wouldcontain search keywords and
things like that.
So when people got there andthen they clicked on something,
that's how those sites makemoney because they're part of
(19:24):
the Google search partnernetwork.
So that created a lot of clickfraud in the very beginning and
that was success.
And when we saw more clicks,the advertisers said, oh, while
these campaigns are working,awesome, let's now move more
money into digital and pay forthese things.
Now, when you actually lookedunder the hood a little bit more
, right?
So one of the earliest examplesfor a pharma company is we saw
(19:47):
a blended average click rate of9%, greater than 9%.
Okay, that seems odd, becauseeven humans don't click on
search ads that much.
Right, it's higher than displayads, but not at 9%.
Okay, that seems odd, becauseeven humans don't click on
search ads that much.
Right, it's higher than displayads, but not at 9%.
And then, when you startedunpacking that again, this goes
back to just common sense wesaid, okay, well, can you show
(20:07):
us a report on a domain bydomain basis?
Right, we don't want to see thebig average, because that's
average across everything.
Once we broke out domain bydomain, we saw a whole bunch of
domains pegged at 100%click-through rates.
That doesn't make any sense,right, and that's tens of
thousands of ad impressions andclicks.
Well, broke out with sufficientdetail.
(20:30):
Right, the average is kind ofhide those details from you,
right, but when you actuallybreak it out and see line by
line, how the heck are some ofthese websites getting 100%
click-through rates?
Okay, oh, by the way, these alllook like fraudulent sites when
you actually visit them.
That's how we starteduncovering the fraud, even
without any specializedtechnologies, right, it's really
(20:51):
about asking for enough detailsso that you can kind of break
through those averages, becausethe averages hide the fraud very
easily for you or from you.
So when you look at the details, you can already say, okay, we
got to turn off all of thesesites because they're completely
fraudulent.
Now, that was very simple andthat was rampant amounts of
click fraud more than 20 yearsago.
(21:12):
And then now fast forward totoday and let me kind of get
into the opposite end of thespectrum, which is the
performance marketers and a lotof B2B ones are performance
marketers right, they'll say weonly pay for the lead, we're not
going to pay for the impression, we're not going to pay for the
click, we only pay when we getthe outcome that we want, which
is a lead.
Okay, so, like I said earlierin this podcast, the bots will
(21:34):
do exactly what you pay for,right, what you want to pay for,
in this case, the leads.
So, when you have a lead form,they will fill it out completely
, submit it, right, they'll evensolve the CAPTCHA and then
submit it to you.
So now you're paying, you know,$5 per lead, $50 per lead,
sometimes $500 per lead.
Now, in those cases, let meagain tell you how we use very
(21:58):
simple common sense to detectthis.
If you see a lead form submittedin one second, okay, is it even
humanly possible for a human tofill out all the forms fields
in your form and submit it inone second?
Okay, obviously not.
But the bot has all of the dataright From a database somewhere
(22:19):
of stolen information.
By the way, to complete yourform and solve the captain,
submit it in one second.
Just by having basic code onpage we can say, okay, this
doesn't make any sense.
So therefore this is a fakelead.
Even if every field in there,right, the zip code matches the
city, the phone number area codematches the city, all that kind
(22:40):
of stuff, Everything matches up, everything looks right.
But it's not real.
It wasn't submitted by a human.
Another very simple thing thatadvertisers are doing is just
adding invisible fields to theform.
A human can't see it, so theycouldn't possibly have filled it
out and submitted it.
A bot will see every form,every field on that form, and
submit it.
So when you get these formfills that have those fields
(23:03):
completed, you know definitely ahuman didn't do it, so then you
can just discard those outright.
But those are varioustechniques.
I'm sure the bots will getbetter.
We talk about AI agents buyingstuff for you or filling out
forms.
Even the purchase can be doneeasily by bots, right?
You've seen these cases whereTaylor Swift's tickets were
basically all purchased upwithin 15 minutes of the launch
(23:26):
by scalper bots right, the botswill buy all the tickets so that
scalpers can sell them for fivetimes the money on the
aftermarket.
Right.
So bots can do anything that ahuman can do in a browser.
It's just whether they have afinancial motive to do that
thing.
If they're paid for that, theywill do exactly that.
Kevin Kerner (23:45):
Golly, you know what I love about what you're
saying here.
It's so what you're asking, thequestions you're asking are so
simple.
Like could it be?
Would it be possible to havethat like in a second, or would
it be possible?
It's really.
I guess there's so muchbelievability behind the ad
industry.
You're just like what.
Dr. Augustine Fou (24:03):
They said it was someone looked at this.
They just didn't ask thosequestions.
Kevin Kerner (24:07):
But, to me we're not asking the right questions.
You know, one thing I lovedabout the stuff that I read
before our call was that youwere asking the question, or
you're making the point thatthere's a difference between
someone actually doing somethingwith an ad versus just seeing
an ad.
It's not just the attentionside, it's the attentiveness
side.
Can you break that down a bit,because I think it was really
(24:27):
that's the point.
You know what you need todetermine if someone has
attentiveness with the ad.
Is that correct?
Dr. Augustine Fou (24:34):
Yeah.
So let me kind of put it incontext, because a lot of people
have heard of the termviewability.
They've also recently heard theterm attention and I think a
lot of advertisers areoptimizing for attention.
I'm going to put a third terminto this mix and I'm going to
call it attentiveness.
But before I get toattentiveness, let me kind of
set the context by talking aboutthose first two.
(24:54):
So viewability is a very basicconcept and it kind of makes
common sense as well.
It's the viewability of the aditself, right, so whether the ad
was viewable.
So the main concept and there'ssome standards around this now
which is 50% of the pixels ofthe ad in view for one second,
because the common logic willsay if the ad itself wasn't even
(25:14):
viewable, then people couldn'thave seen it, it couldn't have
had any kind of impact whatever.
It makes total sense.
So it basically means thatadvertisers want to buy viewable
ads versus non-viewable ads.
Now, that's a good idea intheory, but what has happened in
the real world is that the badguys have easily faked the
(25:35):
viewability measurements orthey've basically stacked all
their ads above the fold so theyregister as 100% viewable even
if they're not right.
When you stack ads.
Say, 50 ads on top of eachother, only the top one is
visible or viewable right, theother 49 are behind it.
So obviously, a human can't seeit, even though the detection
(25:56):
tech will all register it asviewable okay, because it was
above the fold in the viewportand all that kind of stuff.
So, again in digital, there'seasy technical ways for the bots
or bad guys to fake any metricthat you want to see.
And in 2018, there was a casewhere Newsweek a mainstream
publisher right, supposedlypublisher was using very simple
(26:16):
JavaScript code to just falsifythe viewability measurement to
make all of their ads look likethey were 100% viewable.
So, when the advertisers thinkthey're buying 100% viewable ads
, they got to ask the nextquestion okay, were they
actually viewable?
Okay, so, that being said, wenow move to what you just
mentioned attention, which isokay.
(26:37):
Here's the next evolution ofviewability.
Right, the viewability has todo with the ad itself.
Attention, which means theydrive more business outcomes,
(27:07):
makes total sense, because alarger ad that gets your
attention will obviously performbetter and drive more business
outcomes than a small ad thatyou may not have even seen.
Like you skipped it, or it wastoo.
It was on the bottom.
So you didn't even look down atthe bottom of the screen or
something, right?
So higher attention meansbetter outcomes, makes total
sense.
The problem that a lot ofpeople don't really understand
(27:31):
is that you can't actuallymeasure attention in real life,
in real campaigns.
Right, you can measure it in aneye tracking study in a
laboratory environment, but youcan't measure it in real life
because the JavaScript is notallowed to turn on your camera
to see if you were actuallylooking at the screen.
Right, it would throw a prompt,a user prompt, security issue
(27:52):
if it tried to access yourcamera.
Yeah, no one's going to do thatno one's going to do that.
And that's what I meant byattention.
Vendors can't actually measureattention because they can't
show you that the person wasactually looking at the screen
the moment the ad showed up.
So then the optimizations thatend up happening right when
you're optimizing for higherattention means you're basically
(28:13):
optimizing for larger ads,because an ad that covers up the
entire screen obviously getsmore attention than an ad that
doesn't cover up the entirescreen, that's hidden on the
very bottom right, a mobile ador whatever.
So basically that's still fine,right Directionally.
More attention gets you moreoutcomes, that's totally fine.
I'm not saying don't use it,but there's a better way and I'm
(28:34):
a scientist so I want to makesure I can measure it.
So that's where attentivenesscomes in.
And attentiveness doesn'thappen on the ad and doesn't
happen on the person's device,because we can't turn on the
camera.
Attention, attentiveness happenson the advertiser's landing
page.
That's the page to which theuser clicks after they click the
(28:55):
ad.
And, like I said earlier, if ahuman deliberately clicked an ad
, that means they were kind ofinspired by it and they were
curious about something.
That means they were kind ofinspired by it and they were
curious about something.
So when they clicked on it,when they get to your website
they're probably going to dosomething like look at the
content, move around on the page.
So there are certain humansignals, like mouse movement,
(29:16):
page scrolling, click events,maybe touch events if it's on a
smartphone that we can actuallysee and directly measure on the
landing page of the advertiser.
So attentiveness is a verysimple concept.
It's basically the collectionof these human interaction
events mouse movement, pagescrolling, clicks, touch events
and the percentage of users thatactually did that when they got
there.
So if the majority of the usersthat clicked on your ad and got
(29:40):
to your landing page actuallydid something on your landing
page, we call that highattentiveness, and when maybe 1%
of the users that got to yourlanding page only did something,
we call that very lowattentiveness.
So you can now use attentivenessto gauge the relative
(30:00):
effectiveness of your ads, andattentiveness actually takes
into account both viewabilityand attention already, because
the ad would have had to havebeen viewable and then the
person would have had to payattention to it and then click
on it and attentiveness on thelanding page.
That's a good thing, because ifthey came to your site or
(30:29):
active on your site and didsomething.
That's a step towards aconversion.
So higher attentiveness on theadvertiser's landing page is
better than lower attentivenessor no attentiveness on the
landing page.
And in fact, zero attentivenessis the same as what you might
see in your own Google Analyticsas 100% bounce rates or zero
(30:50):
time on site.
Right, they left right awaywithout doing anything.
So I think in that case, justto kind of reiterate viewability
, attention and attentiveness, Iwould prefer to measure
attentiveness on the landingpage because it already takes
into account viewability andattention.
Kevin Kerner (31:08):
So simple and so logical.
These are really easy questions.
Dr. Augustine Fou (31:12):
Yeah, I've been thinking about this for a
long time, you know there aresimple things, non-technical
things that advertisers can doto make their B2B campaigns way
better.
Kevin Kerner (31:22):
Yeah, it's really a blind spot.
You know it's a blind spot for,I'm sure, a lot of CMOs, a lot
of people in performancemarketing, because you're just
trusting that the system works.
It has to work.
You know, it's been around forso long.
The one thing I wanted to askyou very quickly about is
attribution.
We get a lot of skepticism onattribution from technology
marketers and I think it's youknow, part of it.
(31:43):
They're trying to say to theirexecutive leadership team here's
the stuff that's working, solet's spend more.
And then part of it isperformance based.
It's like what is working andhow am I going to?
You know what am I going toinvest in more?
How should CMOs or performancemarketers and B2Bs start
thinking about attribution?
Like how should they trust itat all?
(32:03):
How?
Dr. Augustine Fou (32:03):
do they validate performance?
I would say they can trust it,but they should be skeptical.
Right, there should be ahealthy, healthy amount of
skepticism.
And let me kind of explain why.
Right, so the concept ofattribution is attributing sales
and outcomes to the ad thatcaused it right, or the ad
impressions that caused it right.
That's great, that's fine.
(32:24):
But what has happened is that alot of these attribution models
are literally just that they'remodels, right, they're
mathematical models,methodologies to attribute sales
.
So let me use a very simple wayto explain this.
Right, you've all heard ofattribution windows, right?
And those can be set to oneweek to 30 days, 60 days,
whatever.
The attribution window is thetime that you would basically
(32:48):
look at after the ad exposure tosay, ok, if any sales occur in
that window of time 30 days, 60days, whatever then we can
attribute that sale or thatoutcome event to the ad exposure
, the ad exposure.
Now what has happened is thatpeople can very easily game the
(33:10):
attribution reporting to make itlook like they got a lot of
outcomes, simply by changing theattribution window.
So say, for example, you setthe window to one week, you have
ad exposure, you wait for sevendays and see if any sales
occurred.
If it did, then you wouldattribute.
And then you're a marketer andsay, oh, there's actually too
few sales that occurred in thatfirst week.
Why don't we wait for longer?
Why don't we wait two weeks?
(33:31):
Why don't we wait for 30 days?
See if any sales occurred.
So it kind of depends on theproduct, right?
So some products take longer tosell, especially B2B.
Those are very long saleswindow.
By adjusting the attributionwindow, they can now say we just
got a lot more sales attributedto these ads simply because
they waited longer.
Now what happens if you reallywant to cheat and say, okay, why
(33:52):
don't we make it 60 days?
Why don't we make it 90 days?
Okay, how about we make it halfa year?
So any sale that happened inthe half a year we can now
attribute back to the adimpression.
Yay, okay.
So you see how easily the ROASor ROI reporting, the
attribution reporting, can betricked to be made to appear
(34:14):
better, right?
So that's where the marketershave to really understand their
particular vertical, theirindustry, their product category
, whatever.
So certain products, like smallticket items, attribution
window should be very, veryshort.
Can of soda, very short.
B2b it could be six months.
Kevin Kerner (34:30):
That could be a legit attribution window, and
it's a very complex journey too.
Dr. Augustine Fou (34:35):
Yeah, exactly A lot of people involved, and
they're interacting with allkinds of stuff, so the marketers
need to be honest withthemselves in setting the
attribution window, but that'swhat I meant by you know, if you
let someone else set theattribution window, or you don't
even know what the attributionwindow is because the platform
just gave you a ROAS report, youcould be very easily tricked
into thinking something'sworking when it actually is not.
(34:58):
It's simply claiming credit forsales that either had occurred
in the past or would haveoccurred anyway, right?
It's simply a figment of howthey did the reporting, the
attribution reporting, right.
So it's not to say it's a badthing.
You just have to understand thelimitations and be very
skeptical yourself to make sureyou're getting the right numbers
, so that you understand whereto allocate more budget because
(35:21):
it actually is working, notbecause it looks like it's
working.
Kevin Kerner (35:24):
Yeah, really good advice, Really good advice.
The last question here, andthen I want to get into AI
roulette.
There's going to be a lot morecontent created and potentially
even like look what Meta isdoing with trying to make
advertising easy for the smallbusiness.
Is AI the?
Will AI be the problem, or isthere something in AI that will
be the solution in all this as afraud?
Dr. Augustine Fou (35:47):
researcher.
I'll tell you AI is going tocause more problems than it will
ever be the solution for.
And, very simply, you know,it's actually a useful thing for
small businesses, like if theyhave certain product pictures
right of the stuff that they'reselling.
It is trivial and it makestotal sense to let AI generate
the banner ads for you.
(36:07):
You don't need humans to dothat, so there are certain ways
you can use AI to your advantage, obviously.
So even automating some ofthose processes would not be a
bad thing for small and mediumbusinesses, who may not have a
creative agency to do that forthem.
They don't need to pay acreative agency to do stuff
that's going to be templatizedand easy to automate anyway.
However, like we said earlier,for 15 years, bad guys have
(36:31):
simply been plagiarizing contentfrom other places pictures and
texts and just putting it ontheir WordPress templates to
make fake websites.
With AI, they don't need toplagiarize it anymore.
Ai has already done it for them.
They'll just generate thecontent.
So the proliferation of morefake sites is just going to be
easier and easier going forward.
That just means that we'redemocratizing crime.
(36:52):
We're democratizing ad fraud,so more people can just sit in
their armchair at home and do adfraud.
But, that being said, if youreally understand that fraud has
been at the 90% level for thelast 15 years, a little bit more
of it due to AI is not going toreally matter.
Right, you're still going towhat you're going to get to like
92%.
Kevin Kerner (37:11):
That's going to be the problem, yeah just went up
by 1%, okay.
Dr. Augustine Fou (37:14):
So I think a lot of people think that ad
fraud was 1% for the last 15years.
So it's probably going to beshocking to them when they
realize 90% of their stuff isnot shown to humans.
And I think you asked thequestion earlier in this podcast
to say what can advertisers do?
They can probably save 90% oftheir digital budgets right now
(37:38):
and do better Talk about ROAS.
If you're saving that money,every dollar saved goes straight
to your bottom line, yourprofit.
So that's enormous ROAS thatmarketers can do.
That's the cheat code righthere.
Right, they can actually showway more ROAS simply by not
spending the dollar thanspending it on showing ads to
bots.
Kevin Kerner (37:58):
Yeah, incredible, absolutely incredible.
Do people really think you'recrazy?
Dr. Augustine Fou (38:01):
Absolutely right.
I've been a pariah for the last15 years and I'm totally okay
with that, because I think youknow, when we're talking about
this, there are going to be ahandful of marketers will say,
oh shit, we better do somethingdifferent, because all of our
assumptions may not have beencorrect for the last 15 years,
but there's going to be a wholebunch of them will say well, I
prefer doing what I'm doingright now.
(38:23):
I don't want to lose my job overthis.
So they're just going to notlisten to it.
But I'm fine with that too,right?
Kevin Kerner (38:28):
So and you and you have the weight of the whole
advertising industry.
Dr. Augustine Fou (38:32):
really, they're all in con right now.
Right they're.
They're all drinking rosé andpatting each other on the back
so.
I'm fine, I'm still here in NewYork, but.
Kevin Kerner (38:46):
I love it, I love my.
The rebel inside me loves thefight.
I just love the fight.
You're out, so I'm I'm a fan,all right, sure, okay, I warned
you about this.
I do an ai roulette question atthe end.
Dr. Augustine Fou (38:51):
So now we're gonna get to the hardest
question of all just the bestquestion, okay, and so let me.
Kevin Kerner (38:59):
so what I did is I loaded, uh, perplexity up with
your a couple of the researchreports, the questions that I
was going to think about askinghere, and your LinkedIn profile,
and so I'm going to push, sendon Perplexity and let's see what
it gives us here.
It's going to give you aquestion that you'll have to
answer on the spot.
Okay, here it goes.
Dr. Augustine Fou (39:17):
Oh man, it's really roulette.
You just actually submitted inreal time.
Yes, all right, okay.
Kevin Kerner (39:22):
All right, okay, it hasn't blown up yet.
Okay, but this could be theinaugural one.
So here we go.
If you had a magic button that,when pressed, would instantly
reveal to every CMO and brandexactly how much their digital
ad spend is wasted on fraud andwho in their organization or
agency has been knowinglyignoring it, do you think the
(39:43):
industry would actually changefor the better, or would it just
lead to mass chaos, fingerpointing and maybe even the
collapse of some of the biggestplayers?
Would you press it?
Why or why not?
Dr. Augustine Fou (39:55):
I wouldn't press it because from 15 years
of experience it won't doanything.
They will literally schedulenext year's con and just all go
to con and ignore it.
I've seen this for 15 years.
The story is not going tochange.
Kevin Kerner (40:09):
That is so awesome .
So the nuclear option.
There's no mutualself-destruction in any of this.
Dr. Augustine Fou (40:14):
It's like, okay, whenever they come around,
I'm going to be here.
Kevin Kerner (40:18):
That is so great.
That is so great.
Well, this has been amazing.
I mean, I've just been sittingback soaking it all in.
I'm so delighted to have you onthe podcast and to get your
information out, particularly tothe tech and B2B space.
This has really been fantastic.
Thanks, Kevin.
If people want to take the nextstep and get a hold of you or
(40:39):
get more information, what's thebest way to?
Dr. Augustine Fou (40:45):
information.
What's the best way to?
Most of my content is actuallyon LinkedIn, so there's 800 plus
articles from the past 15 years.
It's all published on LinkedIn.
So just find me on LinkedIn,follow me if you'd like and just
reach out directly.
Right?
It's just augustinfu atfuanalyticscom, f-o-u
analyticscom, happy to answerany questions.
I'm always happy to support anadvertiser who really is brave
enough to take a look moreclosely under the hood, because
(41:08):
that's really the first step.
Right, we can't solve fraudovernight, but we need these
baby steps and we need moreadvertisers to have the courage
to take that first step, and I'mhere to help.
Kevin Kerner (41:17):
Yeah, I love it.
Well, thank you.
Thank you so much, dr Fu.
We'll stay connected and keepfighting the good fight.
Thank you, kevin, good to seeyou.
All right, talk soon.
Popular Podcasts
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
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.