Customer Segmentation

All companies have customers of various types, regardless of whether the business is B2B, B2C, e-commerce, retail, etc. For example, at a basic level, a B2B SaaS company can classify its customers based on the size of their monthly subscription. Or if they have an annual contract instead of a monthly subscription. By knowing which accounts are different, they can prioritize any needs based on what matters (usually large accounts with annual subscriptions).

Customer segmentation is simply this thinking process carried out on a larger scale. Instead of having 1 or 2 things when ranking customers (monthly versus annual and small versus large), you can include a lot of different characteristics in the equation. An online business, for example, could segment customers based on how they behave on their website. A physical retailer could segment customers based on the products being purchased. And so on and so on and so on.

So how can you do this and how do you use it? The truth is that there are many ways to perform customer segmentation, so I'll take an example and use it as a way to learn how to do it.

Let's imagine that we are an e-commerce store. We're trying to increase revenue per customer and retain more customers. How can we do this?

Step 1: Select What Matters

The selection of functions is decisive for successful customer segmentation. In that sense, the question we must almost always ask ourselves is: How does this connect with what we want to do?

If you take our current example, we want to increase revenue per customer and retention. The natural characteristics to include would be: How much revenue do we get for each customer? Another good feature would be: How long has it been since this customer made a transaction?

There are also features that it's not natural to include. For example, what state is the customer in? Or how old is the customer? You may still be able to include them in the segmentation, but in the end they may or may not be important. One way or another, any customer segmentation will have a limit on the number of features. A good rule of thumb? Probably no more than 10 functions are good. Anything more than that is difficult to understand.

Step 2: Segment and then do it again. Then again, then again.

This is the most Technique , in a sense, but I'll ignore the technical details. All you really need to understand is this: someone (usually a tool or a person) needs to analyze data to divide customers. In this process, there are usually a few steps:

1. Select the algorithm (this is the technical aspect)
2. Select the Number of segments (usually no more than 10, beyond that it becomes difficult to handle)
3. Run the algorithm and observe the results.

The truth is that steps 2 and 3 are usually repeated several times. This is largely because there's a level of randomness in these things, so we repeat it to see different results and find something that matches. However, in the end, after trying several times, the result should tell a little story. That story is the one we're going to use.

Step 3: Read the story

After running the algorithm several times and having a story, we can take that story and use it to make decisions. Let's use the example we mentioned to see how it works. Below, you can see the results of a customer segmentation:

So what is the story that the previous issues tell us? I think of a few things:

1. 1 and Done : Are these guys really customers? What did they buy? Why couldn't we turn them into something?
2. Hungry for Discounts : How cost-effective are these types? It seems like they want a discount all the time. Maybe they only like expensive products but can't afford them and that's why they always want discounts? Or maybe they just have less income?
3. Loyal Fans : they seem to really like our business. We should keep trying to do more business with them. For example, a new, more premium version of our products might be what they would love.
4. Lost Fanatics : this is pretty bad. These are former loyal followers who didn't stay. Reactivating these guys would go a long way toward achieving our revenue goals. They are also the most important to keep.
5. Slow and Steady : These seem to be the most standard clients. They're not the biggest fans, but they haven't been agitated or gone either. I think maybe we can dig deeper here in some other way to see what we can do with them.

Step 4: Rewrite the story

Now that we've read the story the data tells us, it's pretty clear that we have a few things we'd like to do:

1. One and that's it : something has to give way with these guys. We need to go deeper with these people to increase our revenue per customer and retain them. They are a missed opportunity in the business.
2. Loyal fans : they are clearly the company's biggest fans and the most receptive to its products and services. Since they are big fans of the business and don't like discounts very much, we should try some higher-priced items to see if we can increase our income with them.
3. Lost fans: we need to reactivate them and that's all. If we don't reactivate them, we are losing customers who are 9 times more important than other customers. They don't seem to be big fans of discounts, but what if that makes them return to the system? Valdria More than the penalty .
4. Discount Hounds: this group needs more research to see what to do with them. Maybe there's an opportunity here somehow, but I would prioritize others first.

Conclusion

In conclusion, customer segmentation is an excellent way to improve a company's operations. It requires some thought about exactly what type of data to include in the modeling, but once the segments have been defined, it leaves considerable space to act on the results.