Predictive Personalization’s Greatest Pitfall
I’ll make it an easy read for you this week. Predictive personalization’s greatest pitfall is retailers themselves. Why? Because they don’t trust the technology.
There is plenty of historical context. Any time optimization gets applied to a process in retail, there is a big risk that retailers will not take advantage of it. There was pushback when price optimization first arrived. More often than not, retailers implemented fancy, high-end price optimization tools, only to override or ignore the recommendations – because they didn’t trust them. Price optimization became a very fancy way of applying pricing rules, which, rather than driven by data, became driven by the merchandiser’s “gut feel” for what consumers would be willing to pay.
It took a long time, but eventually retailers came to understand that what price optimization really did was let retailers get much more granular about how they could implement pricing rules – by zone, by category or sub-category, by consumer, and many other iterations. And, pricing solutions responded to the lack of trust by opening up the black box of optimization more, to help end users develop a better understanding of why the solution made the recommendations that it did – the solution basically had to train its users to see past “gut feel”.The solution had much more granular ways of looking at price than really, any human is capable of – not because the math is too hard, but because the business moves at a speed far faster than any human can really go through all that math across all those dimensions (product, category, consumer, location, etc.) and make decisions.
That same lack of trust still exists for merchandising solutions, like assortment optimization. In that case, though, it’s more about creating a process that is fluid enough to actually help human decision makers rather than just get in their way. The nature of the process involved in optimizing an assortment is iterative enough that there aren’t a lot of issues around why a solution might recommend dropping gray and adding wine as a color option in a given season. And there isn’t necessarily a driving need for merchandisers to try to override recommendations like that.
The challenges exist more around balancing design elements and target prices against what’s in a placeholder within an assortment. Merchandising end users aren’t that used to directly incorporating customer data – like checking to see if they’ve built an assortment that meets the needs of a specific strategic customer segment. The challenge with assortment optimization is changing the process to create enough space to consider these other ways of looking at assortment than the business is used to.
Personalization is a unique combination of the challenges inherent in both price optimization and assortment optimization. Marketers and merchandisers who are considering optimized (“predictive”) personalization have to both overcome their bias towards “gut feel” (as in, “I know what consumers want and I’m going to override this recommendation), and they have to change their process to consider far more angles to personalization (like consumer behavior or social media trends) than they’re used to when thinking about things like product recommendations or landing pages on the website.
That process challenge is complicated by the fact that personalization, by its very nature, also reveals things about the consumer shopping process that go against the grain of merchants’ and marketers’ gut feel. Part of it is driven by a more granular view of customers – averages from grouping consumers into segments hide a lot of variability, and anywhere there is variability, there is an opportunity to drive more value. Part of it is not understanding or having trust in why personalization wants to make the recommendations to specific users in the first place.
Right now personalization is mostly a digital thing. Retailers have a richness of data and the ability to shape or influence the shopping process online that they just don’t have in stores. But they, unfortunately, don’t have trust. That’s why in our latest digital selling benchmark report, we found that predictive personalization was at the bottom of retailers’ opportunity list for using personalization – right above the last-place opportunity of giving in-store personalization tools to use in customer engagement.
There are a host of things to say about that, for both retailers and personalization providers. For providers, learn the lessons of price optimization vendors and open up that black box more. Focus on training end users as to what personalization has learned about consumers so that the marketers and merchants looking at personalization recommendations can have a better understanding of the reasons behind recommendations, rather than leaving them in the dark and just telling them to trust the math.
For retailers, there is some responsibility here too. Optimization isn’t going to go away. Retailers have a responsibility to understand what it can do – both its capabilities and strengths, and its constraints. Today it’s personalization, tomorrow it’s doubling up on personalization and promotions, on top of channel optimization (not just who or what or what price, but which channel is best to offer it in).
This breakdown of personalization is just one recommendation we made in our report on digital selling. I encourage you to check it out.