Embedded Files
Display attribution is easy to game.
When I first set off on this journey to measure lift for my brand’s paid media I sat down with our vendors to explain the new directive. I was met with confused lines of questioning from some, indifferent agreement from others, and one validating anecdote.
At the airline, Tuesday was deal day: the day of week our CRM team would always send out promotional emails. One of our top performing DSPs knew this, and not because they were devoted to the airline’s loyalty program. The DSP would load as much spend into that day as possible to game our attribution system into giving credit to those ads for site activity that should have been mostly attributed to the email program.
I've never seen display ads consistently generate conversion lift on any brand in the last six years, but some tactics can be moderately successful.
As I wrote about in Lift 101, the cleanest way to set up a lift experiment for display advertising is through an ad server. This was the first media channel in which I attempted to orchestrate this kind of experiment, and the result was alarming:
On the whole, our display investment yielded no incremental conversion. Control group conversion rates (people who were exposed to a placebo ad) were actually slightly higher than those who saw an ad for the brand. Said another way, the ads had no effect on the users to whom they were served. It’s the OPPOSITE of what you want to see.
Audience targeting had the greatest effect on lift.
This data included was programmatic banner ads only, and the partner or creative level data didn’t indicate why our ads did not generate behavioral change. We could see it when we looked at % lift by targeting though:
Our prospecting ads were broadly targeted with some light intent qualifiers. This was the first major lesson we learned that could be applied across most channels: stop retargeting. Retargeting ads go to people who have already been on your site and have likely already made a purchase decision. This is the most clear example of how easily display ads can be attribution gamed: if someone converts after being exposed to the ad, they’d already made that decision when they visited your site. Why would the ad take any credit? It’s like making a sign inviting people to come eat at your restaurant, and then placing the sign inside of the restaurant instead of out on the curb.
Display, especially programmatic display, is a crapshoot. Unless you are more sophisticated than the arbitrageurs and bot networks, save your money.
Across every vertical I’ve worked on, retargeting is the most efficient performance marketing tactic. Layering on incrementality measurement however, will usually knock it down to the least efficient. The Display channel is plagued with complex problems: fraud that Google or inventory providers are not incentivized to police, extremely complex sharks that arbitrage the cost of inventory to the point that it’s impossible to decipher the intrinsic value of an ad, ad blockers and growing banner blindness as people learn to just ignore the units.
When you stop and think about it, these units are really just digital billboards lining the virtual street as you scroll through content. I have yet to see a vertical successfully generate a positive correlation between any amount of banner spend and an action on the website, let alone sales. If I were CMO for a day at a brand that still spent money on standard banners, I would reallocate the funds immediately, even if it meant walking away from spend commitments that would incur some kind of penalty.
Page updated
Google Sites
Report abuse