Lessons from the real world: Five best practices for driving sustainable growth from personalization

After spending many years as the VP of Marketing at one of the leading personalization technology providers, I have a pretty good understanding of the benefits and limitations of personalization. And while there are very clear advantages for retailers that adopt a well-designed personalization strategy, I worry about the personalization bubble that has been created. There is irrational exuberance in the marketplace today, an almost mystical belief in the impact that personalization technology will have on a business.

Messages from analysts and thought leaders are fueling this fire. Here are excerpts from an email I just received promoting an upcoming analyst webinar on personalization—sound familiar?

“Personalization is being redefined as individualization—structuring interaction, functionality, and content around the real time individual needs of customers.”

“Discover why individualization, not segmentation, is becoming the new standard for personalization.”

The implication is that with the right technology investments, brands won’t need to do any heavy lifting in marketing. They won’t need to dig deep to understand the differences across customer segments, they won’t need to develop personas, and they won’t need creative breakthroughs to propel their business. Just leave it to the algorithms to crunch the customer data and deliver the perfect message to each consumer at just the right moment. Sounds like magic, right?

As my mother always said, “if it sounds too good to be true, it usually is, unless it’s me.”

Continue reading

Custora at DataGotham

In September I had the pleasure of giving a five-minute “lightning talk” at Data Gotham, an annual conference that brings together New Yorkers who work with data in their professional lives. Speakers came from private businesses, the public sector, and academia. Videos of the speakers are available on YouTube, and if you have time to listen to two or three of them, I’ll also recommend Igor Elbert from Gilt or Jeff Hammerbacher, currently at Mt. Sinai.

But if you have time for just one, I’ll (not so) humbly promote my own talk (also embedded at the bottom of this post). I give a quick, high-level view of how we approach data analysis at Custora, intended to be understandable by people who do not necessarily have a statistical background or familiarity with the retailing or marketing industries.

For those who are new to the idea of data-driven e-commerce, the main takeaway is our longstanding mantra that every customer is different, and that businesses that identify these differences and tailor their strategies accordingly have an advantage over those that do not. We reflect this through predictive modeling at the customer level. This allows businesses to identify the value of their existing base, accurately budget for new customer acquisition, and customize the way they communicate with different people.

For those more comfortable with statistics, I’d highlight the way we generatively model CLV with a dependency structure of intermediate variables. For example, we treat CLV as a function, in part, of order rate and of “lifetime,” the duration of the customer’s relationship with the business in question. These are in turn assumed to vary across customers according to a model-assumed probability distribution (whose parameters we fit to the data). By modeling these intermediate dynamics, we create a framework that can be used not only to estimate CLV but other statistics with meaningful interpretations in the context of order rate and lifetime: the probability that a customer makes another purchase, expected number of orders over the next month, and so on.

I hope you enjoy the presentation, and if you are interested in data and in the New York area next fall, I encourage you to look out for the next iteration of Data Gotham. And if you are interested in learning more about customer lifetime value, check out our Custora U courses on the subject.