Everyone is talking about segmentation, personalization and the software involved.
But what if you can’t use advanced software or hire a specialist? what if you just don’t have the money? is your business doomed?
The answer is no.
Even though using the proper tools and asking experts for advice is always a good idea, you can take initial actions on your own to ensure relevancy of your messages and improve KPIs
Where to start your segmentation:
Start by trying to find a simple (and easily trackable) way to split up your audience: arbitrarily choosing users will be harder to track later.
A good example might be to split paying users vs. non paying (in case of a freemium mobile app, for one). Another might be the interests in different section of your site.
A third way can be by demographics: Gender, Country, age range, or a combination.
Whatever you choose, make sure you have a different message to each group, otherwise the segmentation will lose it’s purpose.
Don’t worry if more than one of the above looks tempting. There’s (almost) no wrong answer at this stage.
Not sure how to
What if your choice isn’t that plain and simple? what if you opted for all active users in your free to play app and wanted to separate them by their rank? should you choose arbitrarily?
If there’s no easy, straight forward way to segment try using a graph. Sometimes the mere visualization makes differences pop up:
In the above graph, you can see there’s a decline in each rank, but there’s a very obvious drop just between rank 9 and rank 10. Try to understand why that is. If there’s a logical explanation (it appears that rank 9 is really hard to overcome in the given example), you can use that to your advantage. Otherwise, try and look for a different split and come back to this at a later stage. No sense in risking your revenues over what can be a mere coincidence.
Once you have decided on the first segmentation, and you’ve got a different message for each (for an e-commerce site, non-payers might get a store wide discount to tempt them to make a first purchase, while payers will get a coupon to be used on a purchase for a large sum), you should get to the testing part: A/B test each of the above segments to determine if the test group did better than the control: choose the KPI for each test (conversion for the non payers, average spend for the payers) and send the new message to a small group.
The purpose of testing on a small group is to avoid causing damage to a
large section group by sending the wrong message: offering too big of a coupon will result in loss of revenues, too small an offer will result in anger or disappointment in the tested group, which will lead to churn.
Choose a group size that is large enough to determine if the test was successful, but not too big as to cause too much harm in a worst case scenario.
After determining the size of the groups, send the message. Try to avoid holidays, weekends and other factors that can influence the test (unless the message is a holiday sale, of course).
The final stage in the test is to determine the results: if the test group won, congratulations, you have a profitable segment! if it lost, you can either try another message, or try a new split.
Once you have your first segmentation, iterate: look at that segment as new whole, and try to find easy, intuitive ways to split it again: Purchasers can be split again by the total amount spent so far (you can even divide
into more than one group). Users who didn’t purchase can be split into those with items in the basket, and those without, etc.
Keep your groups under control:
An important thing to remember through the process is that over-segmentation can lead to fracture, which is a state where you actually earn less from each group than by sending the same message to the whole population.
Finally! you’ve a good number of groups of a proper size, you’re done testing right?
Make sure you refresh your groups by re-testing. People change, and probably so will your business. Always make sure the offers are
still relevant and appealing.