How to Optimize Digital Campaigns Without Outcomes Data

I've said repeatedly over the years that digital campaigns MUST be measured and optimized for business outcomes (e.g. sales) not vanity metrics like clicks, click-through rates, traffic to the site, etc. Otherwise, bots happily generate more clicks so they can absorb more of your budgets. But there are some scenarios in which you just can't get outcomes data to use. I've seen this in large advertisers where sales occur offline, and they simply don't get the data back in time, or where the sales data reside in other departments and they don't share it back to the digital marketing department. So, the question is, can you still optimize your digital campaigns absent this data?

Yes.

I'm going to invoke the "theory of relativity" -- no, not the one by Einstein -- the one by Fou. Let me explain.

Relative numbers instead of absolute numbers

In the chart below, you see eight different paid sources: three paid display, two paid search, two paid social, and one "native." You can see the RELATIVE amounts of dark blue (humans) versus dark red (bots/fraud) in each type. With just this information, you can already optimize your digital campaigns -- by allocating more budget to the sources that have more humans and away from sources that have more bots and fraud. Even if you don't have outcomes data to work with and close the loop completely, you can still make your digital campaigns better because you will be showing more ads to humans and less to bots. And that is the critical first step to improve outcomes; after all bots don't spend money and showing ads to them is useless.

Let me focus on the difference between "relative numbers" versus "absolute numbers" for a minute more. Absolute numbers are almost always inaccurate, if not outright incorrect. This is because in most cases assumptions and extrapolations have to be made to arrive at those numbers. A perfect illustration of this is in the industry-wide estimates of ad fraud -- either the dollars or the percentage of fraud. No party has 100% coverage of the entire universe of digital ads; in other words, they don't have measurement in 100% of all ads of all types across all advertisers, etc. They only see a slice of the universe; I only see a slice of the universe. In these slices, there is a certain amount of fraud and the levels of fraud depend entirely on how well the campaign is managed. As you can imagine, there are hundreds of variables that contribute to this. For any party to say a number for fraud across the entire industry is irresponsible because it's an extrapolation based on a limited set of data. Furthermore, even if they said fraud is a large number, everyone just thinks it doesn't apply to them, ignore it, and go back to their daily business. So it's useless.

You MUST measure for humans too, not just for bots

However, those parties like the ANA (Association of National Advertisers) and TAG that knowingly cite low IVT numbers like 0.6% - 0.8% year after year, are causing actual harm; they have misled the largest advertisers for years by telling them fraud is low and their initiatives helped solve it. This has caused advertisers to lose billions of dollars of their ad budgets to fraud; and worse, some of those dollars flowed to terrorist groups, nation states, organized crime, hate speech sites, disinformation outlets, piracy and porn sites, etc.

The key problem is they rely on fraud detection vendors whose tech can't catch most things. To be blunt, the 0.6% is all they could catch. You can surmise that the bots used by fraudsters are skilled enough to evade their detection. One of the simplest ways they do this is to block the detection tags. Large portions of campaigns are simply not measured by these fraud detection vendors; so when they report 1% IVT (invalid traffic), the other 99% should be read as "THEY DON'T KNOW" instead of "everything's fine."

In FouAnalytics, not only do we show you what portion of the data is not measurable (white area in our charts), we also measure positively for humans (dark blue in our charts). You will notice in the examples above, that aside from the dark red (bots and fraud), not everything is dark blue. This is where the concept of "relative" is important. You can judge the RELATIVE quality of the clicks arriving from different paid channels by looking at both the dark red and the dark blue. And you can optimize your campaigns by progressively adding budget to those channels that show more dark blue and less dark red. This means you lower budgets to those sources that have more red and less blue.

Judging the RELATIVE quality of clicks

Once you have both dark blue and dark red to work with, you have the data to judge the relative quality and effectiveness of different tactics you use. In the three donut charts above, you will notice that "retargeting" has the most dark red. This is related to the form of fraud where bots deliberately visit an advertiser's site first, and then go to "cash-out" sites to cause ads to load; they make higher CPMs from retargeting.

For practitioners who have stuck with me this far, there are two more articles on my Substack with some more examples and details -- e.g. full funnel measurement, organic versus paid sources, etc.

https://fouanalytics.substack.com/p/detect-humans-optimize-for-humans

https://fouanalytics.substack.com/p/simple-cross-channel-optimizations

Let me know what you think, and if you want to use FouAnalytics to measure and optimize your own campaigns. FouAnalytics remains free for small and medium businesses.