Lessons from Facebook's "Glitch" Basic Campaign Hygiene

"The glitch caused many companies to fund Facebook ad campaigns that were essentially ineffective" is such a hilarious understatement, it caused me to take a closer look.

News of Facebook's latest "glitch" is still coming in. I was not running my own campaigns, so I am relying on information from others, primarily from this article: CNBC: Major Facebook ad glitch has advertisers asking about refunds . The details still emerging from this issue remind us about basic "campaign hygiene" -- basic things you should make sure to take care of in your campaigns to minimize your exposure to these "glitches" that seem to keep happening -- (2018) Facebook's Ads System Suffers Glitch Days Before Black Friday , (2020) Facebook inflated reported metrics that caused advertisers worldwide to increase spend .

Dayparting, frequency caps, CPM caps

Quoting from the CNBA article, he "noticed that it had spent 90% of its daily Facebook ad budget by 9 a.m. That meant it had only 10% left for the remaining 15 hours of the day. He then found that the problem was widespread, spanning his entire customer base."

"Data about the glitch provided by the advertising technology firm Proxima on 108 companies also revealed that these firms spent their “entire day’s budget in the first few hours of the day,” the company said. Companies that implemented cost caps, or limits on their advertising campaigns, were not impacted by the glitch."

Companies should set day parting. Day parting is meant to help ensure that your ads are shown evenly throughout the day and not used up in the first few hours of the day" as in this example, and the examples in the slide below.

This happens more than you think in programmatic campaigns too (not just on Facebook campaigns), but it is hidden from view because most advertisers don't ask for or don't get HOURLY details in their reporting. If they did, they'd see details like the 2 examples in the slide above. The green bars represent hourly volumes of ads. Notice the large spikes right at the midnight hour, and much lower quantities thereafter. Most of the ads are used up in the overnight hours when humans are sleeping, leaving few to no ads to show during waking hours.

"their CPMs at least doubled ... advertisers had essentially wasted most of their money for the day, spending roughly triple the amount they normally would to acquire a customer."

Companies should set CPM caps, so CPM prices don't spike out of control. If you let others' algorithms spend your money, that can happen either on accident or when they need to make more money. It's all black box so you have no way to verify. So the best practice is to "contain."?You should have your own analytics, like FouAnalytics, whenever possible so you can verify what is being reported to you by the platforms.

Companies should set frequency caps. Frequency caps limit the number of ads repeatedly shown to the same user. In the data table below, you can see an example of overfrequency in a programmatic campaign where ads were shown to the same fingerprint hundreds of times. Fingerprints are anonymous representations of unique device-browser combinations, which is a proxy for users. Although "f-caps" may not be perfect, they will limit the "out-of-control" situations like the many glitches documented over the years on ad platforms.

I'll keep this article short. If you want to look at additional examples and cases of suboptimal characteristics hidden from view, have a look at the supporting slides section towards the end of the deck below.

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