Latest / The Yield Doctor with James Deaker / Bad Ads are Good Signals

Transcript

Bad Ads are Good Signals · The Yield Doctor with James Deaker

0:01
Broken ads are rarely the main problem.
0:04
They're usually symptoms. They're
0:06
symptoms of weak decisionmaking,
0:09
poor incentives, or AdOPS teams that are
0:12
constantly fighting rather than
0:14
optimizing. And in this video, I'm going
0:17
to show you the external signs, things
0:20
that anyone can see that point to deeper
0:24
yield and operational issues inside of a
0:27
publisher.
0:29
When I first go to a website or an app,
0:32
I'm looking for signals about their
0:35
business model, I'm looking for the
0:38
quality of the advertisers that are
0:40
showing up and I'm looking for whether
0:42
things are broken because those are
0:44
usually signs of deeper operational
0:47
problems. And this isn't just for
0:50
consultants like me. It's a useful tool
0:53
for investors, for analysts, for anyone
0:57
involved in an acquisition because you
1:00
can tell if a company is well-run
1:03
operationally and whether there
1:05
potentially is upside from better
1:07
advertising or yield management just by
1:10
looking at the pages. In fact, it
1:13
surprises me how often acquisitions are
1:15
done without the potential buyer
1:18
spending meaningful time on the site or
1:21
app in a way that a realworld user
1:24
would. And so today I'm going to walk
1:26
through those external signs, things
1:28
that you can look at in a standard
1:30
browser and what might indicate that
1:33
there are problems. And again, this is
1:35
not about naming and shaming. It's about
1:38
looking for and looking at what you
1:41
should be trying to find. So, as we go
1:44
through these examples, I'm going to
1:46
score each of the problems on two
1:48
dimensions. The first is how easy is it
1:52
to detect? How easy it is to notice the
1:55
problem on the site or the app. And the
1:58
second is the strength of the signal.
2:01
how strongly what we're seeing suggests
2:03
that there might be a problem with the
2:06
operations or the yield management.
2:09
Something can be very easy to spot but
2:11
may not be a big deal. Something else
2:14
can be hard to catch but very concerning
2:17
if you do see it. And the key isn't just
2:20
that you look at one signal, it's the
2:22
pattern across all of the signals. So,
2:26
first things first, example one is ads
2:29
not loading correctly. And I'm looking
2:32
for things like whether the ads are
2:34
breaking or whether they're not loading
2:36
correctly into the frame because this
2:39
can be a huge red flag. And there are a
2:43
number of different flavors of this. One
2:46
is that the ads may not load and they
2:48
may create a blank box or a gray space
2:52
or a white space on the page. Maybe they
2:55
load slowly and take 5 or 10 seconds to
2:58
load. Or maybe they load, but they have
3:01
complete gibberish inside the square.
3:04
You don't see this very often, but I've
3:06
seen examples where it's JavaScript code
3:09
running inside an iframe, and the
3:11
consumer would have no idea what to to
3:13
make of that. And so here I'm I'm
3:16
showing you an example of an ad that
3:18
didn't load, and all you see is the
3:20
whites space. And you might ask
3:22
yourself, James, how do you even know
3:25
that that is supposed to be an ad? Well,
3:26
there is a clue. And the clue is that
3:28
the ad choices icon loaded so that you
3:32
could opt out of the targeting, but in
3:34
this case, the rest of the ad didn't
3:36
load. So, when you see an example like
3:38
this, it means that there is no person
3:41
catching it and there is no machine
3:44
catching it. You would usually want to
3:46
have processes in place to flag when you
3:49
have no ads loading and when you have
3:51
problems. This is an example of direct
3:55
revenue leakage. You don't know if it
3:57
was because of there was a lack of
3:58
demand or whether an advertiser actually
4:01
bought this impression and the ad failed
4:04
to render due to some technical problem.
4:07
But even if the problem originates with
4:09
the advertiser or with the agency, the
4:13
job of a good ad ops team or a data
4:15
quality team is to track this down
4:17
quickly. One downside for looking for a
4:21
signal like this on a publisher page is
4:23
that you may need to spend significant
4:26
time to to find it. Uh I'll actually
4:29
show you examples later in this uh
4:32
presentation of examples where I did not
4:35
have to sp spend spend significant time.
4:38
I caught it pretty quickly. So if we
4:40
summarize this as a signal, how easy is
4:43
it to detect? It can be quite hard. But
4:46
what is the strength of the signal when
4:49
you actually do see it? The strength of
4:51
the signal is high. You know that
4:54
there's going to be a problem. So the
4:56
summary is that broken or non-rendering
4:59
ads are one of the clearest signs of
5:02
operational failure and direct revenue
5:05
loss. All right, example two, high value
5:09
placements showing house ads or PSAs. So
5:13
first we need a couple of definitions.
5:16
A house ad is an ad promoting your own
5:19
products. Think of LinkedIn running an
5:22
ad for their own job services or Yahoo
5:25
running an ad for fantasy Yahoo fantasy
5:28
football or Google running an ad for
5:31
Google Maps. A PSA is short for a public
5:36
service announcement and is also known
5:39
as a charity ad. Many publishers run
5:43
these at low volumes, often behalf of
5:46
the ad council to promote worthy causes.
5:50
So, in and of itself, it isn't
5:52
necessarily a huge red flag. The
5:54
occasional PSA or house ad is fine, but
5:59
no publisher should be running more than
6:01
a few percentage points of their
6:04
impressions in this way, particularly in
6:07
the highest value placements. It's
6:09
usually a sign if it's happening that
6:11
they don't have backfill in place for
6:14
backfill ads or remnant ads and that
6:17
somehow they're just running sponsorship
6:18
ads and they've run out of demand. So,
6:22
this is easy to spot and it's usually a
6:25
sign of trouble or revenue weakness
6:28
because there's really no excuse for
6:30
running high volumes of house or PSA in
6:32
this day and age. So, how would we
6:35
summarize this? The ease of detecting
6:38
this is relatively easy. Spend some time
6:41
on the page and if you haven't seen one
6:43
of these types of ads, hopefully there's
6:45
no problem. The strength of the signal
6:48
relatively high. If you're seeing a lot
6:50
of PSA in house, it's usually a sign
6:53
that there's a problem with the site. So
6:55
summary of this frequent house or PSA
6:58
ads and premium placements usually
7:01
indicates poor yield management logic or
7:04
unde unresolved demand problems.
7:08
All right, example three. The same
7:11
creative appearing in multiple ad slots
7:14
on a page. Another thing that I look at
7:17
when I go to a web page is whether there
7:19
are multiple ad slots showing the same
7:22
creative at the same time. Often it'll
7:24
occur if the placements are right beside
7:27
each other, maybe one above the other.
7:30
And it's actually a tough problem to
7:32
solve and requires more sophistication
7:34
to manage. So I'll show an example here
7:37
where it's the exact same creative. And
7:40
as a diagnostic as a diagnostic
7:43
technique, it's relatively easy to
7:45
force. What you do is you go to a page
7:47
where it has two or more of the exact
7:50
same ad on the page and you just
7:52
constantly hit refresh. And if you've
7:54
done this 20 or 30 times and you haven't
7:57
seen it, there's probably nothing going
8:00
on. Why this occurs is often when a call
8:04
goes out from a page from the buyer
8:07
perspective, they're seeing an
8:09
opportunity to reach that user and
8:11
they're bidding at that user basically
8:14
through multiple placements.
8:17
It tends to occur more frequently than
8:20
broken ads because ad calls often go out
8:23
simultaneously to the same exchange. How
8:26
easy is it to detect? I'm scoring it a
8:29
medium on how hard it is to detect the
8:33
strength of signal medium low. If you're
8:36
seeing it all the time, it's usually a
8:38
sign that the yield management team is
8:41
not on top of their game. But it usually
8:44
also means that at some stage they've
8:46
taken some proactive efforts to put ad
8:49
placements and they've thought about
8:50
their page. So, repeative creative
8:54
signals often can shallow either signal
8:57
either a shallow demand pull or could
9:00
reflect gaps in the optimization logic
9:03
going on on that page. So, another thing
9:07
when I'm actually looking at the ads is
9:10
the diversity of advertisers. So, this
9:12
is example number four. Are the same
9:16
advertisers appearing again and again on
9:19
the page? And here I'm not talking about
9:21
a sponsorship takeover where one
9:24
advertiser may have bought a page or a
9:27
homepage for all of the day. I'm talking
9:30
about if you go to other parts of the
9:31
site, what do you see? And a healthy
9:34
site should see dozens, hundreds, or
9:37
even thousands of advertisers appearing,
9:40
usually well targeted to the audience
9:43
that's showing up. If you keep seeing
9:45
the same brands or categories
9:47
repeatedly, it can be a signal of
9:50
shallow demand pull and over reliance on
9:53
the same group of buyers on that page.
9:56
It usually signals a lack of diversity
9:59
and tends to cap the prices or the CPMs
10:02
that the publisher is going to get even
10:05
if the fill rates look fine. So ease of
10:08
detecting this relatively easy. The
10:12
strength of signal as to whether you
10:14
have a problem or the whether the
10:16
publisher has a problem. This is medium
10:18
strength signal. Uh and the summary here
10:21
is that low advertise account usually
10:24
points to constrain demand and quietly
10:28
limits the the revenue upside on the
10:30
pages. All right. Example number five.
10:34
Aggressive ad refresh. Um, after I've
10:37
looked to see whether the ads have
10:39
actually loaded, you would normally want
10:42
to stay on the page and see whether the
10:44
ads refresh. Are they uh auto refreshing
10:49
every 20, 30, 40 seconds? In the old
10:53
days, an ad would stay on the page and
10:55
it would not rotate or it would not
10:57
change. Now, it's much more common to
11:00
have new ads appear. It can be a sign of
11:03
revenue pressure or it can be a sign of
11:06
a public company maximizing short-term
11:08
revenue. So, the key is to try to work
11:12
out as you're going through this pattern
11:13
matching, is it a tactical response or
11:16
part of a long-term strategy?
11:19
Ease to detect this pretty easy. You go
11:22
on the page and you'll be able to see it
11:24
very quickly. the strength of signal on
11:27
this particular signal in isolation I'm
11:30
calling I'm scoring low and again this
11:33
is one you want to use in conjunction
11:35
with other signals. So the summary here
11:39
is that aggressive refresh alone isn't
11:42
alarming but it can reinforce other
11:45
revenue stresses that you may be seeing
11:47
on the pages. All right, example number
11:50
six, uh a sticky ad. And here what we're
11:53
talking about is when you are scrolling
11:56
through a page through the content are
12:00
the same ad blocks staying on the page
12:04
uh moving with the user. Uh and this
12:07
isn't rocket science but it shows that
12:09
at some stage someone was paying
12:12
attention to monetization.
12:14
The example that I'm showing on the page
12:16
is from Yahoo. When you scroll through
12:18
the content, you'll see ads over on the
12:21
right hand side. You can detect it
12:24
relatively easily. And similar to the
12:27
previous example, the strength of signal
12:29
as to whether there's a problem here in
12:32
isolation is relatively low. But it's
12:35
one of those signals that you can use in
12:37
conjunction with the other signals that
12:39
I'm talking through to understand
12:41
whether the publishers under some
12:43
pressures. And the other thing I'll say
12:45
about sticky ads is it's very different
12:48
based on the size of the screen. When I
12:51
was taking the screenshot of Yahoo on a
12:54
web page, I'm on a 34in monitor. I'm
12:58
going to show some examples here in a
12:59
little while of sticky ads on a phone
13:03
screen and they're extremely annoying
13:05
and they don't go away. And uh to me
13:08
that's a very different experience and
13:10
I'd look at those publishers in a very
13:12
different way. Example seven, popup ads.
13:16
Um in most high quality sites these days
13:21
don't have a lot of pop-up ads. They
13:23
have a lot of the other things that I've
13:25
talked through cuz pop-ups tend to trade
13:28
user trust and brand perception for
13:31
short-term revenue. And so when you see
13:34
them now, it usually signals that that
13:37
company is under significant revenue
13:38
pressure and that there is a weak
13:41
alignment between sort of the
13:43
monetization and the audience strategy.
13:47
The ease to detect this very easy. The
13:50
strength of signal low but low going up
13:54
to medium because it usually indicates
13:57
the pressures that I that I've been
13:58
talking about. All right, example number
14:01
eight. the overall ad quality on the
14:04
page. And here I'm not talking about
14:06
technically are the ads loading. I'm
14:09
looking at what's actually happening
14:11
inside the ads. Who are the brands? And
14:14
what does the ad creative look like? Is
14:17
there a lot of skin? Is it a dating ad?
14:20
Is it penny stocks? Is it junk bonds? Is
14:24
it dating sites? What are the ads that
14:27
are actually running? And I'll show you
14:28
some pretty bad ads here as an example
14:30
in a little bit. Usually, if you've got
14:34
very low quality uh advertisers running
14:38
through your slots, it's a sign of
14:40
revenue desperation, usually an
14:43
underperforming sales team or weak
14:46
controls around the demand quality.
14:49
So good ad ops teams are very deliberate
14:52
about what they allow and who they allow
14:55
to monetize their audience and don't
14:57
just default to if it's revenue will
14:59
take it as a signal. This one's easy to
15:04
detect. And in terms of the strength of
15:07
signal, it's medium. Again, the summary
15:10
here is that poor ad quality usually
15:12
reflects weak demand controls and
15:16
internal misalignment around
15:18
monetization policies. So I've given
15:21
eight examples and if you look at this
15:23
table you again you will see that some
15:25
are relatively easy to test and to
15:28
perform but the signal that they convey
15:30
may be relatively weak. And the thing
15:33
that again I want to impress on you is
15:35
that while some of these may be weak in
15:37
isolation, taken in aggregate across all
15:40
these different signals, they can give
15:42
you a really good understanding of
15:44
what's going on at a publisher. Okay,
15:48
now let's look at some examples I found
15:50
out in the wild with pretty minimal
15:53
effort. Uh some of the techniques I
15:55
talked about, you may need to spend a
15:56
lot of time on a publishers's website.
15:59
Uh, I went to a couple of sites and
16:01
these really jumped out at me. And
16:03
again, I'm not intending to to shame any
16:05
particular publisher, but but these
16:08
examples are pretty egregious. And if
16:09
you go to the sites, you'll see what I'm
16:11
talking about. So, first is from Screen
16:14
Rant. Uh, I was looking at a story on
16:16
the Mighty Nine, a TV show, which I will
16:19
have to say I really enjoy and I do
16:21
recommend. Uh, and this first example
16:24
is, uh, the primary ad is clearly for
16:27
Tai Chi. It's really low quality. I
16:31
mean, I think we I'm not going to go
16:32
into the details of why I think it is.
16:35
Visually, this is a poor ad. Um, the
16:38
second component of this being a a less
16:41
than optimal experience is that the
16:43
second ad showing up on the same page is
16:45
in Spanish. And to some extent, you
16:48
can't blame the advertiser here. Uh,
16:50
because I'll acknowledge also that as a
16:54
publisher, this is a hard thing to
16:56
reconcile. So, if I only showed you this
16:58
screenshot, maybe I have some sympathy
17:01
for the operations team here. However,
17:04
the next example is more egregious. Uh,
17:06
here we see an example of where the
17:08
primary ad hasn't loaded at all. We have
17:12
a secondary ad loading in the bottom and
17:15
a tertiary ad all on the real estate of
17:18
a of a phone screen. Uh, and finally,
17:21
I'm going to show a third experience,
17:22
still on the same page, where one ad is
17:25
covering up another. Uh, and on these
17:27
types of examples where I might have had
17:29
some sympathy for the first one where
17:31
you have Spanish showing up in one ad
17:33
and English in another, in these
17:35
examples, I have very little sympathy
17:36
because it's clearly a te terrible ad
17:38
experience. All right. Now, let's look
17:41
at a very different type of experience.
17:43
And this is a video from AOL. And here
17:48
we see that there is a delay in the ad
17:51
loading. And while we're waiting for the
17:53
ad load, they have a gray box. And then
17:55
the gray box goes away. And we see one
17:58
ad loads. And we wait a few seconds and
18:01
then another ad loads. Again, this is a
18:05
sign that the operations team just
18:07
either isn't monitoring this or or they
18:09
don't care. Uh, and as I as I record
18:13
this, I know for a fact that AOL has a
18:16
lot going on at the corporate level.
18:17
They are moving to a uh a new parent
18:20
acquirer of a of AOL. And it's just one
18:23
of those things where right now they
18:26
probably aren't paying attention to all
18:27
of the stuff going on on the page. Let's
18:30
talk about a different kind of
18:31
experience. This is Forbes. Uh, and
18:34
clearly this ad that I'm showing you is
18:36
deliberate. it is a full screen
18:38
takeover. I believe that the IAB
18:41
guidelines for these types of ads and
18:43
especially for this type of size is that
18:45
there is a way for the user to minimize
18:47
the ad. And uh I didn't see it. If there
18:50
is, it wasn't clear to me. And to be
18:52
clear, this isn't a sign of a bad
18:54
operations team, but it is a sign of a
18:57
website that's really squeezing every
18:59
bit of juice out of its audience. I'll
19:02
leave you to form your own judgments
19:04
about the behavior, but I'd observe
19:06
again that the ad experience tells us a
19:09
lot about what's going on at the
19:10
publisher. All right. So, coming to the
19:13
end of this video, everything that we
19:16
discussed today can be observed using
19:19
nothing more than a standard browser.
19:22
Some of these signals are easy to spot
19:25
and not necessarily a problem on their
19:27
own. What matters is how many of them
19:30
show up together. When you start to see
19:33
multiple signals pointing in the same
19:35
direction, you can form a pretty
19:38
accurate picture of a publishers's ad
19:40
operations and yield maturity even from
19:43
the outside. I'll potentially do a
19:45
follow-up video that still stays the
19:48
browser, but goes deeper into the
19:51
technical signs like privacy handling,
19:54
identity, ads.txt, and header bidding.
19:57
This was tended to be someone who didn't
20:00
have any of those technical uh things
20:02
that they wanted to get into. But
20:04
hopefully you found the content useful.
20:06
So if you are a publisher, investor or
20:09
acquirer and you want a certain set
20:11
second set of eyes on your monetization
20:14
strategy, feel free to reach out. I am
20:17
James Deika. I am the yield doctor.