What you didn't know you didn't know.

When I showed you the following slide previously, it was hard to picture what was going on between the second and third steps -- drop-off from "ads served" to "ads displayed." The middle column shows the ads served. Even if the ad left the ad server, there's no guarantee that it arrived in the device and got rendered on screen. This example shows a drop-off of 19%, making the total drop-off 26% when compared to the number of bids won and paid for by the advertiser.

That's because

1) the real-time bidding process takes so damn long, unnecessarily,

2) mobile bandwidth is lower, so ads, especially video ads, may not have time to fully download into the device, in time to be displayed on screen,

3) ads wrapped by verification vendors' tags are delayed so much that they fail to render on screen at all, and

4) ad blockers prevent the ad from showing on screen even if it were downloaded into the device.

Didn't know about all that? Yeah, well, most advertisers and their media agencies that bought the programmatic media didn't either. That's because the viewability measurement vendors didn't correctly measure viewability for all of the above. In fact, we've documented that vendors didn't measure 9 in 10 ads with a javascript tag, but still marked 90% of the ads as viewable. To re-iterate. The vendor only measured 1 in 10 ads (10%) with a javascript tag but yet marked 90% of them as viewable. How is that possible? It's because they assume what they didn't measure was viewable. Obviously that is incorrect. And you shouldn't be paying for that crap.

See: Revisiting Viewability - Don't Trust, Always Verify for Yourself

Viewability

What are we seeing in terms of viewability across campaigns, using FouAnalytics? Viewability ranges from 2.9% viewable to 47.6% viewable. Nearly 50% viewable is not terrible, but also not the typical 80 - 90% viewability you have seen in legacy verification vendors' reports. Oh, by the way, bad guys falsify viewability measurements to make 100% of their ads appear to be 100% viewable 100% of the time. Security mechanisms in FouAnalytics means bad guys can't tamper with or falsify our detection of viewability in FouAnalytics like they can in the older platforms.

See more details about our viewability methodology here: https://www.dhirubhai.net/pulse/viewability-ivt-out-of-geo-brand-safety-fouanalytics/

Ads didn't run

To be more precise about the ads not arriving in time to be shown on screen, there's a data point in FouAnalytics called "ad-didnt-run." See the example below. You've probably seen this yourself when using your smartphone. When you are scrolling down the page, you notice an ad slot that has not yet been filled by an ad. You keep scrolling. By the time the ad loads into the ad slot, you're already much further down the page and obviously the ad was not seen.

Of course, this data point -- ads-didnt-run -- varies from campaign to campaign. See the 4 examples in yellow highlight above. From experience I can tell you that when there are more mobile apps in the mix, and more mobile devices generally, the percent of ads-didnt-run is higher. Kinda makes sense because we already said above that mobile bandwidth is lower and mobile CPUs are less powerful. So ads may not have downloaded in time or rendered on screen in time before the user scrolls past the ad slot. Even for those sticky ads at the bottom of the screen, sometimes you will notice they are blank -- that means the ad-didnt-run.

Pixel Stuffing

Everyone's heard of this one -- pixel stuffing -- but have any of your legacy verification vendors shown you any details or data about it? Of course they haven't. But I will. In the 2 examples below, you see a data point called pixel-stuffing:1. The first example shows pixel-stuffing at 44.1%. And the supporting data to the right of it shows a window size of 0x0 (0 pixels wide by 0 pixels high). That means humans can't see the ad even if they wanted to.

When a FouAnalytics in-ad tag is used to measure ad impressions, the window size corresponds to the size of the ad slot. In the second example, you see 8.1% pixel-stuffing:1 and on the right you see the 8.9% of impressions that have 0x0 as the size of the ad. You will notice other legitimate ad sizes in the list too, like 300x250, 320x50, 728x90, 300x600, 160x600, 300x50, and 970x250.

There are cases where window=0x0 is NOT pixel-stuffing. Those are called responsive ads. That means they start off at 0x0 but then are resized into normal ad sizes or the size of the screen (screen takeover ads). Pixel-stuffing means the ad remains at 0x0 or 1x1 and there is no proper resize event. From experience pixel-stuffing happens A LOT in mobile apps, especially the crappy ones designed for ad fraud. Be sure to look out for pixel-stuffing in your campaigns. You won't see it in legacy fraud vendors' reports. You can only see "pixel-stuffing:1" in FouAnalytics detailed data.

Ad slot refreshing, video completion rates

Have you heard of the one where the bad guys or MFA sites were refreshing the ad slots at exactly 1,001 milliseconds? Yeah, they do that to meet the definition of a viewable ad according to the MRC standard, but also maximize the number of ads they can load in the ad slot. How do we check for this in FouAnalytics?

In the 4 examples, we have unload_ms (unload milliseconds). This tells us when the user left the page. When we measure this in-ad, it tells is when the user left the ad or when the ad slot was refreshed. You will see 10 decile ranges in the examples below. On the leftmost grid, the first row says 10% - 21 - 218. That means 10% of the impressions were on the page between 21 milliseconds and 218 milliseconds, or 0.2 seconds. If you add up the percentages of the top 5 rows, you will see that 50% have an unload_ms of 1005 milliseconds or less. Those would not quality for the MRC standard of a viewable ad (which has to be available to be seen for 1 second, or 1,000 milliseconds.).

This FouAnalytics data point can also be used to check video completion rates being reported by platforms and vendors. If a 30-second video ad is marked as "viewed to completion" and the user left the page in 2 seconds, that is not possible. If a 15-second video ad is marked "complete" but most of the unload_ms is below 3,000, that is not possible. In video ads, did you know that bad guys simply call the complete.gif tracking pixels whenever they want to make the video ad appear to be viewed to completion in your reporting. Remember all those 80 - 90% completion rates? Bad guys can mark your video ads as 90% complete even if they never ran the ad. Yeah, because they're bad guys. Simple.

Attention

The parameter above -- unload_ms -- can be used to verify attention too. A viewable ad had the opportunity to be seen. But it doesn't mean it was seen. The browser has to be the application in focus (instead of something else like Word or Powerpoint). The tab has to be the active one, because if the user has 15 tabs open and they are not looking at the one that contained the ad, they could not have "paid attention" to the ad, even if the ad was viewable -- i.e. had the OPPORTUNTITY to be seen. Finally, if the human left the page in less than 1 second, or they scrolled past the ad slot in a vertically scrolling page before the ad loaded and displayed on screen, no attention could have been paid. There are other nuances I won't get into here. But suffice it to say you can use FouAnalytics to verify the attention claims that various vendors are claiming.

Frequency capping

You set frequency caps, right? Are you sure? You can use FouAnalytics to check. In the example below on the left, you can see that each fingerprint is getting 150 or more impressions. A fingerprint is an anonymous representation of a unique browser-device combo. That means Chrome on Windows versus Firefox on Windows have different fingerprints. It's our way of identifying unique browser-device combos using javascript parameters only, without needing any PII.

The example on the left means you should probably check or have your agency check whether frequency caps were set up correctly in your campaign. The example on the right shows you single-digit impressions shown to unique users. So that implies frequency caps are set, and working properly.

You may not believe, but sometimes even if you set f-caps, the DSP or exchange are not enforcing it for you. See the example below. I had deliberately set a "lifetime frequency cap" of 1 in the campaign set up, but yet the FouAnalytics data shows hundreds of ads shown to the same user.

This recent data from Adalytics also shows examples of dozens to thousands of ads from the same advertiser shown to individual users per day.

Ad stacking and MFA ("made for ad fraud")

How many ads did they stick on the page? 10, 20, 50? Sound unreasonably high? Well, bad guys will be bad guys. How about 80 ads stacked on top of each other in the same ad slot, or 3,500 ads on the same page? Yeah, you heard me. This is common practice and they are getting away with it because legacy fraud verification vendors are not catching any of it. They're not even detecting basic bots anyway.

How do we see this shady activity in FouAnalytics? Easy. First check page-frame-count to see if there are egregious numbers of iframes on the page. Then check page-ad-count to see how many of those iframes were filled with ads, and therefore how many ads per page. And finally, for funsies, check out page-above-fold-ad-count to see how many of those were above-the-fold in the first place.

All of these are fraud techniques that MFA sites use to maximize their revenue while giving advertisers what they want -- 1) super massive quantities of ads to buy, 2) cheap CPM prices that are as addictive as crack cocaine, and 3) high click rates that give the appearance of performance. But the reality of it is the exact opposite. Ad buyers can buy as many billions of ads from MFA sites until they are blue in the face, they still won't get real business outcomes, even if they've reported good-looking vanity metrics to their bosses. Marketers, their bosses, and the CFO/CEOs are waking up to this. And they will soon be asking you "how long have we known about this (ad fraud) and why have we kept paying?"

See: Beyond MFA New Details in FouAnalytics (Viewer Discretion Advised)

Once you start seeing the reality of what the legacy fraud vendors have missed all these years, you will be itching to do something. That something is the systematic cleaning up of your digital campaigns with proper analytics for your digital ads. Just like you've had Google Analytics for your sites, why don't you have FouAnalytics for your digital media? The advertisers that have deployed FouAnalytics already are monitoring and actively managing their campaigns with better data than those who are still using legacy fraud verification vendors.

Don't take my word for it. "See Fou yourself" by doing a no cost pilot and copying pasting a tag into your ads.

Thanks to this community for your support and vote of confidence over the years.