“Data-Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.
Today’s column is written by Adam Heimlich, senior vice president and managing director of HX at Horizon Media.
Digital display advertising is effective. It works a lot like TV advertising, lifting awareness a little bit across an enormous audience. A few impressions per target over a few days improves recall, creating availability bias in a brand’s favor at the moment of truth. Avoid wasting impressions served to people not in market for what you sell or who bought it already, and you’re on the road to media buys more effective, per dollar, than TV.
That’s what I learned while focusing on controlled studies for three years while running a programmatic consulting and buying operation. The more newsworthy takeaway is this: Not only is almost no one executing effective digital media buys, the vast majority of big digital advertisers are on the opposite track, chugging like locomotives in the wrong direction. Many have arrived at zero lift.
Digital display is one of many areas of human endeavor where data is exposing conventions that yield suboptimal results. As in finance and baseball, data alone isn’t enough to convince most decision-makers to move away from orthodoxy in marketing.
Consider 2017’s plethora of retail bankruptcies, which include Toys R Us and The Limited among more than a dozen other national chains. Some, if not all, of those companies were still spending on digital campaigns that seemed to be “driving” thousands of weekly sales right up until the end.
Here are the three conventions that cause the most damage:
- Algorithms trained to correlative data
- Noncompliance with standards
- Incentives to waste
Marketers who want to influence audiences via digital media have to beat these. But it’s not easy to defy inertia. People are attached to routines and get upset when their value is questioned. Financially speaking, though, it’s definitely worth it. Digital ad effectiveness is a big opportunity.
Suicide By AI
Let’s say there is no internet and you have 10 sales leads. Seven of them are already committed to your brand, but there’s no way to know which three are on the fence. Your time is nonetheless better spent with all 10 leads than with 10 random targets. Even if you sway only one of three, it’s a good move, because you probably would have needed more than 10 random targets to create one customer. It’s smart to work your lead list.
In reality, we have abundant data and algorithms that learn. The seven that would have converted anyway became eight as prediction grew better, then nine, then a number pretty close to 10. Today, there’s a big problem with spending against your lead list. It’s not smart at all.
None of the artificial intelligence in ad tech is programmed to reveal this. It’s programmed to get that number as close to 10 as mechanically possible. Fueled by enough money and time to analyze billions of impressions, it gets there!
It is typical for digital marketers to believe multitouch attribution addresses this problem. But this faith can’t bear scrutiny with regard to non-incremental conversions: How would it help to divide up credit for conversions for which advertising should get no credit at all? It could (and does) make the problem of bad investments worse.
Consider, for example, a robot observing that humans carry umbrellas in the morning before it rains in the afternoon, mistakenly concluding that umbrellas cause rain. Is the robot clever for taking the corrective measure of also crediting rubber boots?
The real corrective measure is that miracle of the Enlightenment: the scientific method. Decoupled cross-device data makes it possible to run controlled lift tests in digital display. It’s well worth the cost of a PSA “placebo” ad to harvest rich results on incrementality by each individual targeting parameter.
Routinely, the last-touch and multitouch signals are upside-down. I’ve seen it across thousands of tactics. Take remarketing: Every automated system finds “efficiency” in very few impressions against very recent visitors, while all the incremental value comes from several impressions against not-so-recent visitors.
In video, conventions skew branding advertisers’ frequency too high. While the brand-lift studies typically used to measure pre-roll are at least controlled, I haven’t seen an algorithm trained to take cost and reach, as well as lift, into account. Many vendors spike lift by shrinking the audience. Without an eye on reach, reports mislead and video budgets are squandered in shallow pools.
The inverse relationship between reported and incremental ROI is among the most consistent findings and the reason I say with confidence that AI is killing performance for a great many brands.
Railroaded In The Wild West
Digital media grew up without standards, and most salespeople still like it that way. Last year, though, the ANA drew some clear lines on the question of cost transparency. The MRC has now accredited viewability vendors for all the important digital environments except mobile app. This year, WhiteOps found that MRC-accredited fraud tracking can get even sophisticated invalid traffic below 2% of impressions.
So, it’s not the Wild West anymore. It’s governed for those who demand consistent standards and verification – which should be every advertiser on Earth. Sellers who prefer not to comply should have no choice but to give in to advertiser pressure.
Look at how P&G’s voice-in-the-wilderness call for consistent standards inspired Google and Facebook to commit to MRC-accredited video tracking. A CMO’s threat to pull budget is very credible: You can reach every person on the internet without Google and Facebook via OpenRTB, which, despite its reputation, is the most conservative form of digital reach media. That’s why more voices will join P&G’s. Without standards, you can’t compare the value of various investments and minimize cost.
It is extremely common for big advertisers to go through a process of digital budget allocation without regard to standards compliance. In this farcical exercise, the most money ends up with the least compliant partners. By forgoing transparency, viewability and/or fraud protection, bad actors get away with appearing to deliver more value per dollar. How much this practice is fueling the growth of Google and Facebook isn’t exactly a mystery.
Advertisers’ finance and procurement departments should implement a compliance regime based on industry standards, with regular audits to keep marketers honest. This was very difficult before the ANA and MRC caught up to the marketplace. Now, it’s mature, and there should be no excuses.
The Greatest Headwind
Consistent standards and scientific testing won’t be enough. While they’re imposed by leaders from above, the trenches are where digital marketing decisions are made. And the day-to-day conventions that inspire waste are deeply ingrained in investment workflows.
Good luck revising standards and measures if employee bonuses are linked to misleading KPIs. Where this is not explicitly the case, it might be implied – decisions will follow the widespread belief that improving “ROI,” without incrementality testing or compliance, means career success.
The culture of marketing equates budget size with power and prestige, so we’re all disinclined to cut. Never mind the snowballing scale of waste in digital or the trend of complaining about it. Efficient digital is still a threat to most of the entities in a position to achieve it. Even startups make the mistake of putting search, social, display, etc. in competition for “driving” conversions. Agencies and tech partners paid on percentage of spend bring an additional incentive to waste.
Aversion to change is baked into every large organization. The trick is to alter perceptions of risk. Nobody gets fired for allocating most of their budget to Google and Facebook or for presenting a long-term plan that features the logos of every major ad tech player. Yet these are paths to failure. Consolidation of software and data strategies is as terrifying for established brands as relationship commitment is to 20-something millennials, for the same reason: fear of missing out.
Meanwhile, maturity has its advantages.
I hope it’s a relief to hear from a digital marketer that the topic isn’t as complex as it’s been made to seem. The list of things you need to worry about is brief compared to the abundance of pitches you can safely ignore. And the opportunity is genuine. Others have blazed the data-driven trail in the face of convention – Theo Epstein’s Red Sox or the heroes of “The Big Short,” for example. Like them, you just have to withstand everyone thinking you’re stupid or crazy until you win.