Reading Time: 3 minutes

by: Zank, CEO of Bennett Data Science

Personalization means offering your users or customers something different than you’re offering everyone else. This can be as exacting as a different product or action recommendation for every user, or it can be as broad is treating younger users differently than seniors.

In this short post, I’ll show the kind of simple segmentation we see companies miss all the time. If you don’t have some level of personalization in the way you’re interacting with your users, it won’t take long for another company to come along and treat the better.

Free Weekly Data Science Tips For Your Business

  • See the real value of machine learning
  • How you can monetize your data
  • Exclusive subscriber content

If you’re interacting with your users or customer in a homogeneous manner, here are a few things you can do right away, with no data science or advanced analytics required:

ONE: Look at anything you know about your users, such as:

  • Age
  • Gender
  • Geography
  • Earnings
  • Marital status
  • Profession (or student)Then, segment your audience by one or many of these and look for behavioral differences in areas such as purchase, messaging or adoption

TWO: See if you have any data correlated with the highest and lowest performers. At a small scale, this sort of thing is even available in spreadsheets by making plots of parameters and looking for relationships

For example, let’s say you have a group of customers and you are collecting some on boarding data that describes them. You may have a couple parameters that look rather random across high and low spenders (blue and green):
It doesn’t look like we can tease much of a relationship out of these variables. Going one step further, let’s look at the X axis with the Z axis. The Z axis here, is the axis we can’t see…it goes into and out of the screen. Plotting x vs z looks like this:
Now there are clear segments. (Note that I generated this data to make my point. Almost nothing in life ever separates that well!)

In this case, we have a clear break, showing two user/customer segments. Based on the parameters plotted here, we can investigate these two groups and immediately treat them differently, based on their positions in the plot above. For example, imagine a basic product recommender that uses popularity to make its recommendation. That’s fairly straightforward to build and trivial to use in practice when making recommendations to customers. But imagine now, building two different popularity based product recommenders; one for the green customers and another for the blue.

In a nearly negligible amount of additional work, there’s now a way to treat different customers differently. Think of the case of a huge age gap. This is important stuff!

In advanced data science, we aim for one-to-one personalization, where each user receives a unique experience, similar to what we’ve become used to with Netflix or Amazon. And in practice, this advanced personalization pays off. But doing something simple to start with can help young companies realize big gains in a hurry and pave the way for more advanced and accurate methods.

If you want to know how you can achieve greater personalization, please get in touch. We’d love to talk to you!