Behavioural segmentation: lessons learned

22 May ‘16

A lot has been said lately about personalisation, especially segmentation based on customer preferences.

This is indeed a good way to stop spamming everyone at once. Personalisation helps us to to annoy our subscribers less (our emails become more useful), reducing unsubscribe levels and database burnout rates. It also improves marketing (open/click rates) and financial KPIs, e.g., campaign revenue.

A year ago, we introduced segmentation to our product. Now we want to talk about how our clients are using this functionality (including building real-time preference-based segments) and what kind of results they are getting from it.

They use all sorts of data (behavioural, check-based, personal data) in every possible combination. This is cool, because it provides a deeper understanding of what delivers results and makes it possible to constantly improve by creating more segments and testing them.

Four sample segments:

  • «Window shopper» – opened an email in the last 30 days, browsed photo albums, but didn’t buy anything
  • «Interested in educational games» – viewed a product from this category, a page with these products, or purchased them
  • «Doesn’t use mobile devices» – has never logged in from a smartphone
  • «Young and wealthy» – from 18 to 35 years old with a past-year shopping history of RUB100,000 or average check of RUB20,000




Mailings are tied to segments in the platform interface: regular advertising or trigger-based.

Mindbox’s built-in template engine helps to automatically add products, pictures and text offers, allowing for many segments with just one email template.


It works. The graph below compares three emails: a mass email, and two different segments. By testing several dozen different segments, the most effective segment was found, boosting the open rate by 4.5 times compared with the initial email. And this is a large segment with around 100,000 customers.


We believe that constant experimentation helps to build communication models with fewer emails but higher revenue.

This requires a lot of work, a thoughtful approach, and constantly creating and testing hypotheses. Automated algorithms based on big data are good, but they only unlock part of this potential.

The most important thing for us is the growing number of clients who share our point of view, are segmenting their emails and constantly improving by testing.

Join us and improve email together!

Tell us a little about yourself

We’ll respond within 24 hours

Partnership request

Typically we’re answering within 24 hours