Doug Garnett’s Blog


Past Browsing Does Not Indicate Future Purchase

Past Browsing Does Not Indicate Future Purchase

Read a RetailWire post today about personalization. The idea of personalizing offerings based on data is a very common theme at retail – but it’s not a very good one. (After years of articles about how everyone’s efforts fall short, perhaps it’s time we realize it’s not the retailers fault but that it’s a faulty idea).

The following was my comment on the post:

Past purchase is no indication of future interest.  I’ve written about this truth in a blog post about how data and machine learning tell us less than we think.

Where the idea goes wrong is with this sentence: “The most successful companies understand that personalization is at its core a data problem.”

Data only reveals a very small portion about us. Something that lives up to the idea of “personal” is based on far more than what’s available in the data.

Only when we realize this can we pull back from pestering consumers with ads for things they’ve already bought.

Still, many marketers are subject to a naively widespread belief that “data” will reveal everything needed to send people “personalized” suggestions or use digital retargeting to ensure the ads they see are “meaningful” or “relevant”.

What marketers miss by not challenging the underpinnings of the ideas is…

  • These assumptions about “meaningful” or “personalized” ignore human realities. People have tons of interests and we shift between them quickly. Even if I searched for a trash can last week, there are a lot of reasons that ads about trash cans won’t be personal or meaningful this week.

  • In other words, something is only personal for an instant of time. As soon as the second hand ticks to the next hash mark it’s no longer personal.

  • This is made worse because online shopping is heavily buying – not shopping. In other words we go to the web to buy and quite often (usually?) do so very quickly.

  • As a result, by the time algorithms see that I bought a trash can online and decide to send me ads about them, any shopping interest I have is for something OTHER than trash cans (and it’s highly likely what I’m doing on the web isn’t shopping related at all – my mind is not in that space).

  • There seems to be an assumption that “once a trash can shopper, always a trash can shopper”. For trash cans, that’s the least possible reality.

  • Yet perhaps there are enthusiast categories that I search (say flyfishing supplies). Except, I’m human. While I might want to look at flyfishing stuff today in a coffee break, tonite I’ll need to be buying a staple for the household and tomorrow I’ll order a new clamp for some work I do in the garage.

So when your agency’s algorithms decide something makes an ad (or email or product recommendation) “relevant” it’s highly likely they’re wrong.

And let me recommend that more scholarly researchers than I take that on – perhaps the Ehrenberg-Bass team. The question being: What percentage of the time is a targeted ad “relevant” to the person receiving it? I expect the number is very, very small – as low as 10% of the time it’s relevant.

That said, online direct response marketers DO find that digital retargeting is often a cost effective way to generate online sales.

And that makes sense – but they aren’t justifying their choice as a way to be sure that consumers only see things that are “personal” or “meaningful”. They just know that for every 1000 retargeted ads they feed, a few go to people who are interested and a maybe one of those buys it.

They also don’t really care that their ads were irrelevant or meaningless to over 900 of the people who saw them – as long as the cost per order is right.

So let’s start getting real.

One of the buzzwords about targeting has been that it makes ads “relevant”. Reality of targeting is that algorithm chosen potential consumers are relevant to the advertiser – but the algorithm chosen ad is only relevant to perhaps 4 or 6 people out of 1000.

Online direct response marketers should use these tools – it makes sense to find people who are relevant to you.

Brands or retailers wanting to do something good for your customers:  Don’t retarget or feed me ads based on past browsing (or purchase). My past history is not likely to predict my future purchase interests. And that means you run a very high risk of pissing me off and damaging your brand. (I get incredibly tired of weeks of trash can ads just because I accidentally pasted the word “trash can” in a search window.)

If we take care, there are great options for advertising online. Seeing them takes putting a critical eye on what we’re told are “common sense” assumptions about targeting.

Copyright 2018 – Doug Garnett – All Rights Reserved

Categories:   Advertising