Common Definitions and Optimization Opportunities

FouAnalytics is analytics for your digital media, just like Google Analytics is analytics for your website. Beyond ad fraud, there are other aspects of digital campaigns that advertisers simply didn't have the right analytics to see, so they previously could not optimize. But as more and more folks adopt FouAnalytics and use it themselves to monitor campaigns, I have seen some common themes and have made similar recommendations to many parties. So I am summarizing some of the common optimization opportunities below, along with the simple definitions of the observed phenomena.

Over-frequency

Define over-frequency: too many ads shown to the same user/device

Most advertisers assume their agency set frequency caps for them, or assume the DSPs or ad servers enforce frequency caps properly. It's better not to assume; it's better to check. In the data grid below, from FouAnalytics, each fingerprint is an anonymous representation of a browser/device. More specifically Chrome on a PC and Firefox on the same PC will have a different fingerprint. The count next to each fingerprint shows the number of ads shown to each browser/device -- i.e. a user.

Those numbers are not good or bad in and of themselves. But knowing these counts, allows the advertiser to check against their campaign set-up -- i.e. context matters. For example, brand advertisers selling soup and soda may want to show many ads to the same user because they know that simply reminding the person of their brand and product increases their likelihood of buying the product in the grocery store. In other cases, advertisers may want to limit the numbers of ads shown to the same user to low numbers, like 2 - 3 per week. Having this data allows the advertiser to check if the campaign was set up according to their plan.

Over-concentration

Define over-concentration: too many ad impressions shown on too few sites and apps

Most advertisers say they went into programmatic media buying because of the "reach and frequency" that was available compared to buying from a small number of reputable publishers. But the data shows the opposite. When buying programmatic media, we often see the following phenomenon -- over-concentration -- i.e. too large a percentage of ads shown in a small number of sites and apps. In the table below, you can see that a single app -- com.truecaller -- is eating up nearly 63% of the impressions of a campaign. Our recommendation is to split off that one app into its own campaign line and adjust the CPM bid and budget allocation to control for this phenomenon.

Pixel-stuffing

Define pixel-stuffing: ads served into a window or iframe that is 0x0 or 1x1 pixels in dimension, which obviously can't be seen by humans

In the FouAnalytics data grid below, we can see the window size which corresponds to the ad size, since this is measured in-ad. You will notice normal ad sizes like 300x250 and 320x50. But you will also see 0x0 (yellow highlight). This is pixel stuffing if there is no corresponding resize event -- i.e. responsive ads. These ads are marked as non-viewable in FouAnalytics and it is not clear whether the legacy verification vendors catch pixel-stuffing properly.

Day-parting

Define day-parting: running ads during certain parts of the day or hours of the day

In the FouAnalytics time series chart below, we note the green bars, which indicate hourly volume, are roughly the same height every hour of every day. That means ads were served in the overnight hours when most humans are asleep. A best practice would be to turn off the overnight hours like 1 am - 5 am, to save 1/6th of the daily budget (4 hours out of 24). That means more ads can run during waking hours when humans are online.

Pacing

Define pacing: the rate of serving ads over the course of the day.

In the example below, the green volume bars show that right after midnight, the volume spikes. Most of the ad impressions are served between midnight and 2 am. Then we see lower green volume bars and by 12 noon, there's no more volume left to serve for the rest of the day, when humans are awake. This is an example of an opportunity to optimize pacing, so the ads are spread more evenly throughout the day and not used up in the overnight hours.

Advertisers that are getting month-end Excel spreadsheets don't have this basic data, so they would not have known this optimization opportunity. By measuring your ads with FouAnalytics in-ad tags, this becomes readily apparent. Time to upgrade your tools.

Ad-slot refreshing

Define ad-slot refreshing: refreshing the ad slot at exactly 1.01 seconds so that ads meet the MRC definition of a "viewable ad" but allows the shady publisher to sell more ad impressions

see the Ad slot refreshing section in this article -- https://www.dhirubhai.net/pulse/what-you-didnt-know-dr-augustine-fou

Ad-stacking

Define ad-stacking: stacking more than 1 ad in an ad slot where only the top one can be seen by users

see the Ad stacking and MFA section in this article -- https://www.dhirubhai.net/pulse/what-you-didnt-know-dr-augustine-fou

Let me know if I missed anything above. I am happy to add.

Further reading: https://www.dhirubhai.net/today/author/augustinefou