Back in the early aughts, I remember getting worked up about losing around 5% of the impressions I paid for.
In other words, I paid for 100 ads and could only see 95 in the server logs. The usual explanation was plausible–before the ad was rendered, the viewer left the site. (I still didn’t understand why I should pay for the 5 missed impressions, but that’s another discussion.)
The problem was that I didn’t know if 5% was reasonable. What if site A had 3% missing ads and site B had 11%. What did that mean? If the SAC was $50 on site A and $47 on site B, how happy should I be? Could we get the un-served rate from 11% to 3% and yield a few more $47 SAC acquisitions? It made me crazy and some of my marketing analysis friends can vouch for my constant questions.
Ahh, it makes me long for the good old days. It turns out that my pre-programmatic problem was nothing. Bennett Rosenblatt at Uber did an interesting piece of analysis, detailed here at AdExchanger. The conclusion?
If you assume your ads are actually being served, you’re a chump.
He said it in a nicer way. But he detailed a set of circumstances where:
- The publishers misrepresent the type of ad to yield higher CPM
- The ad served mismatched the inventory, causing the ad to fail to load
- The SSPs had no incentive to check the inventory they were representing for sale
- The ad exchanges also didn’t bother to check the inventory on offer
The advertiser is, of course, left holding the bag and paying for the incompetence and malfeasance. Mr. Rosenblatt provides some good recommendations on what to look for.
My recommendation is, at your Monday 8 am meeting, discuss load rates (impressions served/bids won) at the publisher/audience/ad exchange/SSP level.
Please note the high level of granularity. It will take some time to do this line-by-line. Fire up the coffee pot first.
If your agency is buying the same publisher inventory at multiple exchanges, as resold by various SSPs, you should ask why. Secondly, if there’s any statistically significant variation in load rate, you should demand specific answers. (Some if it will be due to the ad exchange running bogus inventory, of course.)
Be prepared for a long slog with your agency and marketing teams. They won’t like it one bit. But looking into your supply chain more closely will pay off.
Takeaway: Don’t assume that winning a bid means the ad runs. Examine your supply chain carefully every morning. Be ruthless when asking questions–it’s your money. And win.
P.S. Here’s more on source-level triage and de-averaging your media buy, both good guides to help you develop a process to address the load rate problem.