Marketers get little benefit from many adtech offerings, warns Dr Augustine Fou, and they need to pay more attention to what they are buying.
CMOs should consider whether they and the marketers in their organization have fallen victim to adtech fraud. In the past decade, some, if not most, shiny adtech offerings have yielded little to no benefit to marketers, despite the increased costs. As advertising budgets face more scrutiny and cuts due to the COVID-19 pandemic and recession, marketers need to spend more wisely and prove ROI. The chart below, (from May 2020) already shows the trend of decreasing “brand spend” (dark blue) and increasing “performance spend” (light blue) across industries.
With the need to prove real ROI at the top of mind, let’s review a few examples of marketers taking action to do just that. In 2017 and 2018, marketers like P&G and Chase turned off $200 million in digital ad spend and reduced programmatic “reach” by 99%, respectively, and saw no change in business outcomes. If those adtech offerings were delivering incremental business outcomes, at all, these large marketers would have seen sharp decreases in business activity and outcomes when they turned off the ad spend and reduced the reach. But, alas, there was no change that was detectable.
Further, Uber is currently suing 100 mobile ad exchanges for fraud. Uber paused its digital media spend that was meant to drive more app installs; but the app installs continued at the same rate! Upon further investigation, Uber discovered the mobile exchanges were both falsifying records to make it appear that ads were placed on reputable domains (when they were not) and fabricating records to show ads being served and clicked, when no ads were even served. This is what Uber is now suing to recover.
Why are these adtech offerings so enticing? Marketers have seen higher “engagement,” in the form of higher click-through rates (CTRs). Marketers have seen lower costs, in terms of unit costs – CPMs (cost per thousand ad impressions). Their media agencies have been able to buy record numbers of impressions for them through programmatic ad exchanges. And their ads are achieving unprecedented “reach” in terms of the numbers of sites they are being shown on. But do more clicks, higher click rates, more sites showing your ads, and more targeting parameters add up to quantifiable, incremental business outcomes, compared to doing nothing?
Behavioural targeting is not worth the premium
Academic studies designed to assess this – e.g. whether “behavioural targeting” drives any incremental benefit for the extra costs to marketers – can’t seem to agree. Widely cited reports like the Acquisti study show that “when a user’s cookie is available (needed for behavioural targeting) a publisher’s revenue increases by only about 4%” compared to an industry-sponsored study that showed that “no cookie present yielded an average of 52% less revenue for the publisher than traffic for which there was a cookie present.”
These are actually both true, but they drew their conclusions by looking at different things. On the one hand Acquisti was focused on the increase in publisher revenue (4%) when targeting cookies were present, while the Google study focused on the decrease in CPM prices (52%) when user cookies were removed. CPMs were indeed lower for ads without targeting cookies (e.g. on Safari browsers which prohibit such third-party tracking cookies); put another way, advertisers paid higher CPMs for using targeting. But that is not the same as publishers getting higher CPMs, or conversely losing 52% of their ad revenue if they removed privacy-invasive third-party tracking cookies. This is because all of the increased CPMs accrue to the adtech middlemen not to the publisher. When marketers pay more for shiny things like behavioural targeting, that extra money goes into the pockets of the adtech companies, not to showing ads. In fact, a third industry study shows that for ads purchased through programmatic channels about half of every dollar goes to middlemen, instead of to publishers (to show the ads).
So how should marketers think about behavioural targeting and whether the extra expense is driving a positive or negative return for them? The chart below shows data from the Beales study, The Value of Behavioural Targeting, that looked at the extra costs advertisers paid for behavioural targeting (e.g. 2.1 times more cost) and the average change in conversion rates, an indicator of outcomes (e.g. 2.4 times higher). If an increase of two times in costs drove an increase of two times in outcomes; that’s literally breakeven. From the data, we wouldn’t be able to see any incremental benefit to marketers, if we didn’t use an extra decimal place. A more recent meta-study by Garret Johnson et al, The Online Display Ad Effectiveness Funnel & Carryover: Lessons from 432 Field Experiments confirms “that [digital ad] campaigns increase conversions with median lift of 8%,” again only slightly better than breakeven.
Proponents of behavioural targeting will of course tout its effectiveness, because they earn two times higher prices when it is used and paid for by marketers. But if marketers are only getting an average of 8% “lift” (some campaigns saw worse outcomes), is it worth the trouble? Without getting into all the math, let’s apply some common sense to judging just how “shiny” the following three adtech myths really are: 1) myth of the long tail; 2) myth of behavioural targeting; and 3) myth of hypertargeting.
Debunking common adtech myths
1. Most humans regularly use 5 - 7 domains per day, few linger on long-tail sites
Can you name ten sites that you visit every day? After how many did you start to slow down, 5 - 7? Same thing with mobile apps. Most humans regularly use 5 - 7 domains on a daily basis, and a similarly small handful of mobile apps daily. Few humans spend a large amount of time on long-tail sites – sites that have niche content. So the idea that there are tons of humans on tons of long tail sites, to whom a ton of ads can be shown, is a myth. The Chase case study proves this. They reduced their ads appearing on 400,000 sites to 5,000 sites (99% decrease in “reach”) and saw no change in business activity. The “reach” they got may not have been human reach – because those long tail sites had few humans, but many bots. It sure was “shiny” to believe there was huge reach when their ads were being sprayed on hundreds of thousands of websites.
2. Targeting has eroded privacy, fed the surveillance economy, and enabled bots and fraud to flourish
Targeting based on people’s behaviours sounds like a good idea – what sites did they visit, what search terms did they type, and what products did they look at on Amazon? But have you experienced a product following you around the internet, right after you looked at that specific item on Amazon? Everyone has. That’s because they tracked your behaviour and assumed you needed diapers and baby clothes, when you were just looking for a gift for someone else. Further, what can you accurately deduce about someone who visited cnn.com or walmart.com? And do any of those behaviours lead to more relevant ads for the user and better marketing outcomes for the marketer? Not much – see the above section.
Don’t forget about the bots that deliberately visit medical journal sites to make themselves appear to be doctors; or the bots that look at backpacks, add to cart, and abandon to make themselves look like the desirable back-to-school student segment. Targeting lots and lots of users based on lots and lots of behavioural parameters is problematic.
3. How many targeting parameters is too many targeting parameters?
A simple thought exercise will tell you the answer. Starting with 100% of the addressable audience, if you choose a gender – M and not F – you just cut the audience in half to 50%. If you further selected one out of five age ranges, the targetable segment is 10% of the original total. And layering on just one more targeting parameter like one in five geographical regions means you’re down to 2% of the total addressable audience. What if you applied 5, 10, 20, or more targeting parameters? What would be the size of the resulting segment that matched all those criteria? Right, very very small; so small in fact, that there wouldn’t be enough new customers if you successfully advertise to just those that matched all the parameters. The bottom line: more targeting parameters does not mean better business outcomes.
What should you do?
With marketing budgets under greater scrutiny, digital marketers must wake up and smell the coffee. Privacy regulations, the death of the third-party cookie, and the virus pandemic are all forcing change. This is the perfect opportunity for you to get a step ahead, by cutting the money wasted on buying adtech snakeoil. Run your own experiments to see what to cut.