UNTARGETED ads drive the most INCREMENTAL sales

I've witnessed the entire arc of the evolution of digital marketing - from 1995 when we needed to convince clients that they needed a website to (nearly) 2025 when we are just starting to get back to real marketing, in digital channels. You'll see what I mean in a few paragraphs.

This post was inspired by Dan Gee and his article on "waste" in advertising -- https://www.dhirubhai.net/posts/dangee1_advertising-marketing-media-activity-7264173989969727488-nFgQ HT Kathy Newberger for the heads up.

"Digital" was, and still is, seen as the most measurable medium for advertising. And it is. Digital is also the channel where in theory, it is possible to get "the right ad to the right person at the right time." This was made possible by cookies, small bits of information in browsers, which allowed advertisers to track and uniquely identify individual users. This means the advertiser can deduce what the user likes and serve ads to them that, in theory, should perform better than ads that were not meaningful or interesting to the user. All of this makes sense, in theory.

But what has happened, in reality, over the last three decades of digital marketing is what I would call the "deep, dark rabbit hole of hyper-targeted advertising." What started off as a good idea got further and further away from reality, because reality was too slow to drive the hockey-stick revenue growth that adtech companies needed to show to their investors to get the irrationally high valuations that were demanded of them. The further we got into digital advertising, the more bots and fraud that were needed to keep the revenue growth going up. Humans just don't click ads often enough. Humans just don't buy often enough. There are not enough actual sales for adtech companies to claim credit for, so they turned to media mix modeling and other blackbox algorithms to "show" customers that they drove lots of sales, even when the thread of causation was tenuous, at best.

Targeting, Retargeting, and Remarketing

Let me articulate more clearly how we fell into this rabbit hole and how we got deeper and deeper into the rabbit hold instead of pulling ourselves out of it.

Targeting. First we start with "targeting" in digital channels. Unlike on Facebook and Gmail, most humans are NOT logged into sites that they visit for content. That means the sites/publishers don't know who the users are. So they started using tracking cookies in browsers to identify the users who visited. With these cookies, adtech companies could harvest the list of sites that a user visited and deduce what the user liked and possible who the user was. For example, if a user visited Sports Illustrated, ESPN, Maxim, and Playboy, it would be easy to deduce the user was male. Conversely, if the user visited Victoria's Secret, Tampax, etc, one could deduce the user was female. But this kind of inference and deduction get far less accurate when we move beyond these over-simplistic examples. What can you deduce about users that visit Amazon dot com or Walmart dot com? Right, it's much harder.

Retargeting. So adtech companies tried to get more and more granular. What product did the user look at? If they looked at it, they must like it. If I looked at a Sony digital camera on Amazon, within a few moments, that exact Sony digital camera shows up in banner ads I see on other sites. This is called retargeting, because in theory, if I looked at the product, I must be interested in it. But what if I were looking at baby cribs or toys? It's not that I just had a baby, it might be for gifts for a friend who did. Retargeting also applies to users who visited an advertiser's website. They must be interested because they visited, right? So adtech companies charge the advertiser extra costs to retarget the user - show ads to the same anonymous cookie that previously visited the advertiser's site. But what if bots knew that and deliberately visited the advertisers' sites before they visited cash-out sites to cause retargeting ads to load there? This means the bots helped the cash-out sites get higher CPMs than they normally would, by tricking the retargeting mechanism into thinking the bot was part of an "intender audience." This was comically illustrated in 2018 when a data seller purported to have 300 million "auto intenders" for automotive advertisers to target.

Remarketing. And finally, wouldn't it be super smart to target ads to users who already purchased from you? Aren't they the most likely to buy again? Yes they ARE. And didn't some marketing class tell you that it costs far less to get an incremental sale from an existing customer than it costs to get a sale from a new customer? What they didn't tell you was that you can't get a consumer to buy a fifth quart of milk a week, when their family can only physically drink fou quarts of milk a week. No amount of ads can get the consumer to buy something they don't want or need. Most consumers can't buy another big screen TV after they just bought one. Most consumers can't buy another computer monitor when only one fits on their desk. Most consumers wont buy a third pair of running shoes, when they are not even using both pairs they have now. And so on, you get my point.

Good ideas in theory, but far less effective in reality

Remarketing (showing ads to users who bought from you already) is a good idea in theory. But in reality remarketing looks so successful because the consumer was already going to buy again anyway, and the reporting claimed too much credit for sales that happened when remarketing campaigns were running, but were not caused by the remarketing.

Retargeting (showing ads to users who visited your site already) is a good idea in theory. But in reality retargeting looks so successful because those users already know the brand and are likely to buy anyway, and the reporting claimed too much credit for sales that happened when retargeting campaigns were running, but were not caused by the retargeting.

And targeting is a good idea in theory. And some targeting is useful to prevent obvious waste (like advertising a beard-trimmer to women). But hyper-targeting is not useful because 1) the derived, deduced, and inferred data is inaccurate, and 2) the more targeted you get the smaller the number of users that match all the targeting parameters. Let me reprise an old slide to illustrate this. If you use 1 parameter -- gender -- for targeting and you chose "male," you've cut the targetable audience in half. If you add a second parameter -- age range -- and you select one of the five, you've cut the targetable audience further, so it is only 10% of the total audience. Say you add a third targeting parameter -- geolocation -- and you target 1/5th of the entire country, you're already down to 2% of the original audience. You can see what happens when you layer on more and more targeting parameters -- what we call "microtargeting" or "hypertargeting." The addressable audience gets too small to be meaningful.

Self-fulfilling prophecy, the myths we chose to believe

How do adtech vendors address this "hyper-small audience" problem? Right, they start to invent things like "lookalike audiences" and "audience extension." They purport to help you find other users -- i.e. anonymous cookies -- that supposedly "look like" the tiny audience that matched all the targeting parameters. The theory was that if you showed ads to these folks, you would get more sales. This is where we really took a left turn from reality. Can you see how easily this can be gamed? Adtech vendors can virtually make up as large of a targetable audience as they want, and you'd have no way to verify whether those audiences were actually consumers that fit all those targeting parameters or were entirely fabricated to absorb all your "ad spend." Like I said above, some adtech vendors tried to stay on the "scrupulous" side of the line, but many were under too much pressure to show revenue growth, they crossed over to the "unscrupulous" or outright "fraudulent" side of the line.

These were the myths we chose to believe because they made us look smart and made the campaigns look like they were working really well.

Keep in mind that little to none of these hyper-targeted ads drove incremental sales, even though the reporting appeared to show that they did. This was the downward spiral that many advertisers got caught up in over the last 10 years. They invested more and more into targeting, retargeting, and remarketing. The reports appeared to show they were getting lots of sales, when in reality those sales were simply occurring at the same time; they were not caused by those marketing tactics. These were the myths we chose to believe because they made us look smart and made the campaigns look like they were working really well.

Waking up to a new reality, and better digital marketing

So the reality was that far far less incremental sales were being driven by these digital marketing tactics than was shown in the reporting. I am highly encouraged to see more advertisers focus on "incrementality" now, finally. In other words, did the digital marketing tactic drive MORE sales (incremental ones) that would not have happened in the absence of that marketing tactic? Some marketers arrived at this by running their own "turn off" experiments where they paused digital advertising and saw that sales continued. This made them realize that those sales were not caused by the digital tactics that they paused. eBay saw sales continue, even though they turned off paid search. P&G saw sales continue even though they cut $200 million in digital ad spend. Uber saw app installs continue even though they cut $120 million out of their $150 million of app install spend.

1. eBay (2012) - https://hbr.org/2013/03/did-ebay-just-prove-that-paid
2. P&G (2018) - https://www.adweek.com/brand-marketing/when-procter-gamble-cut-200-million-in-digital-ad-spend-its-marketing-became-10-more-effective/
3. Uber (2021) - https://www.warc.com/newsandopinion/news/ubers-former-performance-chief-charts-extent-of-ad-fraud/en-gb/44527

So what the heck was happening? Advertisers spent more and more in digital tactics thinking it drove more sales. They started to realize the sales were over-attributed to those tactics, either innocently (media mix models, view through conversions, etc.) or maliciously (attribution fraud), and that those tactics actually drove far less sales than were reported. Not only this, they started to understand that repeatedly beating the same customers over the head with more ads would not get them to buy more. Targeting, retargeting, and remarketing to the same customers had long past the point of diminishing returns. Some customers became annoyed (seeing ads for the product they already just purchased) or pissed (seeing way too many ads for the same store they already buy from regularly). By spending all their budgets on ads highly targeted to a small group of users, the advertisers neglected to spend money on ads that made new users aware of their products -- i.e. broad reach or brand awareness ads.

Put simply, by spending way too much budget targeting the 1% of users who already knew or already bought from you, advertisers were spending way to little budget making the other 99% of the market aware of your product. Even if only 1% of that 99% ended up buying, ALL of that is incremental (sales that would not have occurred in the absence of the advertising). And you've basically doubled the 1% you were desperately trying to get more sales from, but who could not buy a fifth quart of milk when their family could only drink four quarts, or who would not buy another digital camera because they just bought one. I am encouraged to see more advertisers shift budget back to broad reach, awareness tactics -- dare I say "untargeted" ads -- to tap into the other 99% of the market that was not aware of their brand and have not purchased from them before.

Even if only 1% of that 99% ended up buying, ALL of that is incremental (sales that would not have occurred in the absence of the advertising).

This is why I said "UNTARGETED ads drive the most INCREMENTAL sales" in the title of this article. And this is why I said "You'll see what I mean in a few paragraphs." Keep in mind UNTARGETED == "broad reach" or "brand awareness"

Tell me I'm smoking something, or tell me I'm right. In any case, tell someone else who needs to hear this, so they can wake up to better digital marketing too.

Happy Friday (after Thanksgiving) Y'all!