"How do I know whether my wine club members are at risk of leaving the club? And is it a high risk or low risk?"

"How do I identify those people?"

"It seems like my data is kind of a mess, and that there's not nearly enough of it to do any kind of analysis. What do I do about that?"

In a nutshell, those were the three top questions for Enolytics that came from those of you who tuned into last week's webinar we did in partnership with colleagues at WineDirect and Sovos ShipCompliant.

Let me bottom line our responses below. The complete blog post, along with a link to the full recording of the webinar, will be available shortly.

In the meantime, please have a look at answers to three of your most pressing post-pandemic DTC questions, and thank you as always for reading.

Stay safe, and naturally be in touch with any ideas or feedback —

Cathy

Q&A from the Webinar on Post-Pandemic Planning for DTC

How are you cataloging risk? What variables go into a high or very high risk for leaving vs. low? Is it solely previous purchase history?

To catalogue risk, we create cohorts and analyze their behavior. By “cohort” we mean a group of people with similarities, such as females in a particular wine club who have been members for a certain number of months. This is the “pre-marketing” segmentation that lays the groundwork (and adds real-life teeth) to the kind of personas that Adrienne Stillman from WineDirect referenced in the webinar, in terms of targeted marketing outreach and creative communications.

The next step is to ask, what is the likelihood of that cohort leaving the club at this period of time? Every cohort has a risk profile, that our algorithms define based on variables (churn rate being one of them) on a month-by-month basis.

What are the metrics we should be looking at to recognize at-risk customers?

See response above regarding cohorts. An additional variable worth considering is personal purchase history.

This gets “in the weeds” from a technical perspective. The algorithms we’ve developed are beyond the capabilities of Excel and are in the realm of artificial intelligence, but let me share an example.

When a customer's purchase history aligns very closely with the more general cohort's purchasing history, we find that those customers are at greater risk. When customers deviate from the cohort and purchase outside the wine club (through telemarketing and website channels, for example), we find that those customers are at less risk.

Great info, but building unique profiles and sorting by bad data, segmenting by generation or gender etc. seems to be dependent on extensive manual database entry (at least for us as our reports are unable to automatically pull those fields). Any thoughts on to streamlining the process?

You might be surprised at how much we can do with relatively limited fields of data. Birthdate, for example, is critical in order to identify generational differences. We use first name to determine gender with a high mathematical probability. A third example is contact information like email or phone, and also the zip code or street address where you ship the order.

These three pieces of data (birthdate, first name and zip code) are the basis of Enolytics Lite, which we created as an “emergency kit” during COVID-19. You can also see it as “streamlined” analysis of data that you are very likely to already have.

Birthdate, first name and zip code are the three most essential pieces of information you want to be sure are in your database, because that enables segmentation by generation, gender and geography. Email address and phone number are also very important in order to enable targeted campaigns via email (which drives traffic to your website) or telemarketing.