Business

Let’s get Personal – Data Science and a Letter from Mom

By January 20, 2019 January 27th, 2019 No Comments
Reading Time: 3 minutes

Personalization is very near to my heart. After all, I’ve spent a large part of my career helping companies achieve it when interacting with their customers. Tens of millions of them!

But when I think about personalization, I always think back to those form letters someone else’s mom used to send out, circling the right options for the person in mind. Let me have a go at one…

Dear Friends,
You’ve been a [good neighbor, perfect relative, old friend I miss]. It was really great [reading that letter you sent, seeing your baby pictures, watching you on that YouTube fail video compilation]. I hope you [have a nice summer, don’t fall off your bike again, get married soon and give you mom some babies!]. Take care and all the best!

Love, Shirley

Wow, remember those? The concept is terrific; write one letter but leave room to personalize elements and make everyone feel special. But while this might be funny, it’s far from really personalized, and will hardly make anyone feel that special.

Personalization is a concept I talk to a lot of stakeholders about. From CEO’s who want to see a lift in overall product engagement to marketers who want to see higher open rates or click through rates or retention.

Personalization is often the key to these initiatives. But it’s also lacking, even in large companies with data science teams.

I’ve worked for years on 1:1 personalization, from sending tens of thousands of personalized trunks of clothing with Trunk Club to millions of app recommendations with SweetLabs. This stuff is very important to us and our business.

Don’t be fooled! How many times have you received an email with generic products. Maybe, what’s “new” or what’s “trending”? This is not personalization, and while it will perform better than picking random products, it will inevitably underwhelm, as the same emails or offers come out week after week.

Let’s look at the simple case of recommending products to users at an e-commerce site for Acme Corp. Let’s say Acme has hundreds of thousands of customers and tens of thousands of products. It wouldn’t surprise me at all if Acme wasn’t using predictive analytics; after all, the majority of e-commerce storefronts don’t do any intelligent personalization.

After all, at that scale, it’s not trivial to build an intelligent recommender that provides unique messaging to each user. Yet that’s exactly what is expected by shoppers today.

Amazon has raised the bar, so that two day shipping and highly-personalized product offerings are the norm. I can’t help you much with the shipping part, but I can tell you a bit more about personalization!

The goal of good personalization, be it for product recommendations, marketing messages, super-user prediction or churn, is to develop predictive models that provide and crystal ball, accurate for each user, under each situation.

The key to building such personalized models is stores of historic data, showing purchase or behavioral patterns. This type of data can lead to the, “people like you also like…” type of recommendations. These are very powerful, and sit at the base of personalization from Amazon and Netflix, among others.

In closing, please think about what would you rather receive, an email of “trending” products, general enough that anyone might like them, or an email of “trending” products that you’ll probably really enjoy; products that excite or delight you? This is what good data driven personalization provides. It’s the difference between, meh, and, “oh, that’s exactly what I needed.”

Zank Bennett is CEO of Bennett Data Science, a group that works with companies from early-stage startups to the Fortune 500. BDS specializes in working with large volumes of data to solve complex business problems, finding novel ways for companies to grow their products and revenue using data, and maximizing the effectiveness of existing data science personnel. https://bennettdatascience.com