Commentary
We Need 'Act-Alike' Ad Targeting, Not Look-Alikes
- by Dave Morgan , Featured Contributor, September 8, 2022
Two weeks ago, I wrote about the fact that look-alike modeling of audiences in digital advertising has gotten out of control. The models are applied so loosely, by so many today, that everybody ends up being in every look-alike model.
Companies create look-alike models to extend the potential target audience from a small kernel of known customers or audiences into much bigger pools to help marketers get more scale.
Theoretically, the models should be built in such a way that the new, look-alike audiences are as desirable and performant as the audiences they were built from. Or, at the least, significantly more desirable than general audiences overall.
However, as everyone in the digital ad business knows, look-alikes today are hooey. In a desire to create audiences that are large -- that have high “match rates” and can be acquired cheaply -- suppliers and buyers most often now build look-alikes that look almost nothing alike.
advertisement
advertisement
The problem, as media and marketing research legend Bill Harvey told me while discussing the problem over lunch recently, is that what we really want and need are “act-alike” audiences not look-alikes.
There is a reason that Bill was the wizard behind the invention of the ADI (Area of Dominant Influence) and the notion of media-defined regional marketing areas, which was later copied in the creation of the DMA (Designated Marketing Area). Bill is brilliant and can see directly to the root of a problem, and its solution.
Our industry needs to stop talking about look-alikes, which mean nothing but sound great, and focus on delivering act-alike audiences when we are trying to expand a target audience sample into larger audience delivery that is not only efficient, but is actually effective for the marketer.
Finding new prospective customers that “act like'' existing customers is what marketers really want when they are defining target audiences for advertising campaigns. That is what we should be finding for them, and that is what we should be held accountable for.
Do you agree? Isn’t it time to get rid of look-alikes and time to focus on act-alikes?
Dave, you have hit on one of the great flaws in digital targeting.
We are told that tracking the activities of consumers on the internet--what websites are visited, what buys are made,etc. offers granular targeting capabilities---which may be correct---but only for relatively small numbers of consumers whose activity can be evaluated in this manner. Not every intended sports car buyer visits automotive websites; not every person looking for a new brand of toothpaste, searches for more effective brands online, etc.
What most branding advertisers desire is coverage of all of their category users as well as likely new users---not just those few thousands---or few millions ----who can be pinpointed specificlly.So the sellers and buyers have been resorting to what amounts to profiling---even though this practice for TV when applied to "demos" like adults aged 18-49----is routinely derided as stupid and counterproductive.
I don't think that profiling is totally useless as there are distinct correlations between demography---if used properly--by fine cells----plus regionality---that are evident regarding singling out a brand's prospects. But only to a point---it may get you half way there. Why not try for 100%---even if you dn't attain it, you may improve your targeting efficiency. The question is how do you go about it? There, I see no clear cut solutions---nor the data to support them.
Ed, I'm at fault for not keeping you up to date on our latest work with RMT. Simmons established that their best lookalike models accounted for only 6% of brand adoption - using advanced demographics and geographics - and that when psychographics from their own questionnaire plus RMT DriverTags (psychographics deduced from content consumption based on unique content coding system) are added, lookalikes accounted for 21%. In a random control experiment, IDs identified as motivationally resonant with a specific ad outperformed Google Affinity IDs by 95% per Neustar. In another test for selling my new sci-fi novel the RMT/Semasio motivational IDs outperformed the benchmark by >5X. This is what I was telling Dave about at lunch. Third party validations have also been done by NCS and by 605. Most recently in a study by the ARF Cognition Council the RMT Motivational Types were found to explain 48% of IRI sales data - in other words just by knowing the creative used at the time (study was a five year study) 48% of the sales results could be predicted/postdicted, even without knowing the adspend nor the media types. In MMM ones has the reverse situation and leaving out creative, advertising normally receives about 7% credit for sales. RMT partner Semasio following all privacy laws and standards is able to place RMT IDs on 700 million people globally - nearly ten percent of the human race. So far we are actually doing such work in the US and Canada where we have 168 million and 16 million motivational IDs respectively..
Bill Harvey remains among the admired visionaries of our industry. Hopefully, he can make more progress here and convince the powers that be to deploy what's necessary. Still, we find our industry making up "lookalikes" to create cheap impressions and many buyers lining up for the same. Advertisers deserve better? Yes! Will they demand it?
Bill, thanks for that. When I referred to demos explaining 50% of the targeting I was not referring to any particular source as I have seen many of these studies, including one I did myself where I could play with the stats to estimate brand users by employing a combination of finely segmented demos plus mindset information
I tend to question a finding that such methods acount for only 7-8% of the answers and think that somewhere between 25-50% is more typical---again, if the correaltions are done correctly. Of course there are variations between product categories as well as large versus small brands, new versus old brands, etc.
As for how we obtain data to use in media buying, this remains an issue as all too often the media folks are out of the loop where creative and product positioning is concerned and, for upfront TV buys, these are unknowns as the ad messages for the upcoming season haven't even been developed as yet. Also, these are corporate not brand-specific buys so targeting is not really possible in the sense that we are discussing.
I have seen numerous ROI studies that attempt---often with incomplete data----to estimate the contribution of advertising on sales and these usually give ads about 5-15% of the credit, though they are unable to properly calibrate the long term effects of advertising's branding equity function and are focused mainly on short term data. Still, the average marketer allocates only 2-3% of its revenue to ads, which means that ads are a very cost efficient way to generate sales---you spend 2-3% for a 5-15% gain.
Jack, I agree with you that advertisers deserve better----but the agency clients ---despite what they say in public---don't seem to mind the way that TV---in particular, upfront TV----is being bought and they approve all of the major transactions.
More to the point, the only way that improvements will ever be made to tilt the balance of power away from the sellers ----with their big, inflated, "impressions' stats---back to a point where the brands' marketing interests are better protected will be if advertiser CMOs understand that money talks and they actively enter the frey, meaningful amounts of cash in hand, to fight for better targeting and better data. In the latter case they should start with attentiveness, then move on to other relevant changes. So far, there are no signs of any such initiative with the result that advetisers are getting what they pay for where improvements are concerned---namely, nothing.
Ed, I agree. I would say start with attention and resonance. Both are affordable, proven, work with the ecosystems so as to not impact workflow, and both increase sales and brand equity. Also, they complement rather than duplicate each other's effects. Since they both enter optimizers as impact weights it's natural and easy.