I'm a big fan of programmatic, just like the next guy...

... how else could I afford >50 million ad impressions?

Given everything that I have documented and written about the problems related to digital ads purchased through programmatic channels, how could I possibly say I am a fan of programmatic? See: the bad and the ugly of programmatic - https://www.slideshare.net/augustinefou/alternative-to-anas-end-to-end-supply-chain-transparency-study-v-final

Well, have a look at the campaign below, which IS working, somewhat to my surprise. Programmatic works at 1 cent CPMs. You know how adtech proponents like to say “Don’t worry about it, fraud is priced in. You’re paying far lower CPMs now than ever before.” Let’s take that to the extreme — 1 cent CPMs (the DSP doesn’t allow me to specify any bid lower than this).

1 cent CPM campaign in EU

DSP stats - 1 cent CPM experimental campaign targeting EU only (approx 1 month)

51 million ads served in 1 month

EU ad impressions measured by FouAnalytics (less strict labeling)

EU ad impressions measured by FouAnalytics, more strict PCA labeling.

Once we see where the red (fraud and suboptimal placements) came from, we added them to block lists. This allowed us to progressively clean the campaign and increase the portion of ads shown on good sites that have human audiences.

64k clicks to the site (0.1% CTR)

Clicks that arrived at eu.fouanalytics.com from this 1 cent CPM campaign

The valid clicks that arrived on the site, also took an action (touch events below).

From these arrivals, I have been getting a steady stream of requests for invite to use FouAnalytics and also a steady cadence of sign ups for this Substack Practitioners’ Newsletter. Some of these were clearly spam, completed by bots. But some humans did see the ad; some of them clicked through; and some “converted.” Note that these are not big numbers. But it does show that programmatic works - to an extent - when ads are shown to real humans on real sites.

Fraud rates are similar/same at all CPM price points

Why is buying super lower CPM ads OK if you have monitoring in place? From experiments over the years, I have observed that ad fraud is NOT necessarily lower if you pay higher CPMs if you are still buying through programmatic channels. This is because the mix of sites and apps is still crappy. And no matter what CPM you pay for the ads, the sites that are cheaters (use bot traffic, stack 5 ads in the same slot, do popunders/popups, etc.) will still take your money. If you bid $10 CPMs, they will take it; if you bid $1 CPMs, they will take it; if you bid 1 cent CPMs, they will take it.

Can you see any real difference in the dark red (bad) and dark blue (good) across all the price points - -from “less than 10 cents” all the way up to “less than $20” CPMs?

The best way to get your ads on good publishers sites — ones that are not cheaters and intent on ripping you off — is to buy direct from them. Yes, you can buy direct using programmatic exchanges; just be sure to target a specific dealID set up by that publisher on one specific exchange. This minimizes “supply chain leakage.”

So What?

Run an experiment for yourself. Set up a new campaign line and buy programmatic ads for 1 cent CPMs (or whatever super low CPM you choose). If you have monitoring in place, you can see where the fraud is coming from (which domains or apps are cheaters and eating up your ad dollars). Turn those off by adding them to a block list.

So programmatic works for me — at the 1 cent CPMs level. I would not pay (much) higher CPMs if buying through programmatic channels. Note that my campaigns were strictly branding/awareness. This means I did not pay ANYTHING extra for targeting, audience segments, fraud detection, viewability, etc. It was just 1 cent CPMs, run of exchange, targeting EU countries only. Try it and let me know what you find.

Bonus

Of those that arrived on the site from the 1 cent CPM campaigns, 3% went on to take an action on the site. I can see what they clicked.

What did they click? they clicked to read the articles and slide deck linked from the front page of FouAnalytics.com and some of them submitted the form to request invite. See the data about what they clicked below.