Marketers, DON'T optimize your campaign for "performance"

Your first thought must be "he's crazy." Yes, I am "fou" (French for "crazy"). But, remember when I said don't optimize your campaigns for clicks? The reason I said that is because most programmatic ad platforms use clicks or click-through rates as input signals to gauge the "performance" of a campaign. Many marketers are also using "automatic optimization" where algorithms take these signals and optimize campaigns by adjusting budget allocation or bids to optimize for "performance." That means they are allocating more budget or increasing bids for sites and apps that show higher clicks and click rates. Why is this a problem? Right, bot activity.

Bots are programmed to click on ads at a higher rate than humans do. So fake sites and apps always appear to "perform better" than real sites with real humans, who click on ads just a little, not a lot. If you are allowing algorithms to optimize your campaigns, they are faithfully sending more money to the bad guys. Whatever you are optimizing on, the bad guys will cheat and do more of that -- for example, you are optimizing for viewability; bad guys tamper with the viewability measurement so their fake sites and apps always have higher viewability than real sites and apps. So you send them more money. See: March 6, 2018 -- Newsweek Group Ran Malicious Code to Alter Viewability Measurements to Commit Ad Fraud. So "performance" in quotes is not real performance. It's just clicks and click rates that algorithms use as signals. Don't optimize your campaigns for "performance" (in quotes) or for clicks. Optimize your campaigns for real business outcomes. If you're in a large corporation and don't usually get sales data (because that data is from another department), you can still optimize your campaigns. Here's how: How to Optimize Digital Campaigns Without Outcomes Data

But this article is not about campaign optimization. It is about how to analyze clicks using FouAnalytics.

Where did they click?

If you have FouAnalytics tags in your ads, you can see the clicks, what was clicked and where it was clicked. For example, in the chart below, you can easily check if the clicks even made sense. If your ad was 300x250 in size, common sense will tell you that legitimate clicks should be within an area that is 300x250 pixels. You can see these are. So no problems here.

Who clicked it? -- bot clicks vs human clicks

Next, you need to see if the user that clicked was a human or a bot. This is where the labeling comes into play. The following chart shows four different campaigns, measured by a FouAnalytics in-ad tag. Dark red means confirmed bot. Orange means declared bot (one that says its name -- i.e. tells you they are a bot). And dark blue means humans. That is what we care about - human clicks on ads.

You can clearly see there is HUGE variation in quality. Right now, all advertisers are getting reports (excel spreadsheets and dashboards) that report the NUMBER of clicks. You can see why that is problematic, right? In some cases above, like B and C, 55-63% of the clicks were from bots (add up the orange and red). If you didn't know these were bots, and you were just counting the clicks or click through rates, you would think these campaigns are performing really well. In fact you may even allocate more budget to these paid channels because you THINK it is working so well. This is why you need the details to check whether it was a bot or human that clicked.

In the above examples, if you only looked at the dark blue, THAT is what percentage of the clicks are humans. In case A, only 3% of the clicks were humans. That means you are effectively paying 33X higher costs to get human clicks than the CPM would make it appear. In case B, 4% is dark blue. So your effective cost per human click is 25X higher than the CPM would suggest. In case C, it's (1 / 0.15) or about 7X. And in case D, because you are getting 28% dark blue, your effective cost is about 3X to get human clicks. Hopefully this convinces you that just getting a lot of clicks doesn't mean the real performance of your campaign is better. It does appear to be "performant" if you mistakenly use clicks as your metric.

"Don't optimize your campaigns for clicks, otherwise you are sending more money to the bad guys."

What did they click?

Once you've studied the above, the next step is to verify WHAT was clicked. Even if the click locations made sense (e.g. inside 300x250 pixels), and you could see which clicks are from humans vs which were from bots, you should still check what was clicked. In FouAnalytics, you do a search for "clicked" and click [FILTER]. Then you look at the data grid below to see WHAT was clicked.

You will note that NONE of the items in the first 10 rows of this grid show a click on an ad! You have to scroll way way down to see that. Of the 1,017 clicks we see out of 1,003,266 impressions, about half (586) were clicks on the ad. Note the syntax in the screen shot below. For Google's adsystem the click tracker contains adclick. So you can isolate those with the syntax clicked:adclick (that means clicked contains "adclick").

So What?

Scary sh*t right? Only about half of the clicks were even on the ads. And only 3 - 28% of the clicks were from humans (dark blue). That tells you that real clicks are scarce and valuable. What has been reported to you in the past -- just the QUANTITY of clicks -- is not good enough. You need to see where the ad was clicked, what was clicked on, and whether it was a human clicking, in order for you to derive meaningful insights about your digital marketing performance.

Another place you can measure the quality and quantity of clicks is on the landing pages. See the following article for details on how to do this: How to Optimize Digital Media with FouAnalytics You optimize for the channels that show the largest portion of dark blue (humans) as well as the largest number of such clicks. You can calculate a "cost per HUMAN click" based on your costs, by channel and then allocate more budget to the ones sending you the lowest cost HUMAN clicks. Common sense, right? But only if you had the right data and analytics to look at.

If you need FouAnalytics for your sites or ads, please let me know.

Bonus: Facebook display ads (with FAN turned OFF) work really well.