The Jigsaw Guide To Segmentation
The business world is constantly looking for the next great marketing segmentation. Building great marketing segmentation is a lot like building a difficult jigsaw puzzle. I like puzzles with 3000 or more pieces.
First you need to define your borders
I have yet to meet a puzzle enthusiast who doesn’t start by doing the straight pieces. First off, they are quite easy to identify (not always true in segmentation) and secondly, they make doing the rest of the puzzle so much easier. If you’ve ever tried a so-called “borderless” puzzle, you know what an advantage defining the border really is.
Defining the borders or constraints on your marketing segmentation can be a bit more difficult. Your customer base is not likely filled with straight edges (or anything truly analogous). But you need to start by determining who is in and who is out. This starts with defining “customer”, “marketable customer”, or “existing customer”, but will quickly lead to determinations about the ease of identification and attribution.
Customers with limited data are a lot like the edges of a borderless puzzle. There is little to identify where they belong in your segmentation. Unlike the jigsaw puzzle, this can often be solved just by leaving them out. If you are trying to build “one segmentation to rule them all”, this advice won’t sit well with you. But then, you have bigger problems… like figuring out why you think a single segmentation strategy provides any value at all.
Using easy to see distinctions
Step two in jigsaw puzzles normally involves using colors and easy to identify patterns to connect pieces that you can clearly place to one area of the puzzle. In segmentation, this is an exercise in clustering or grouping. In either situation, these easy division allow you to pinpoint position and association. Although puzzle builders can take those terms more literally.
While nearly everyone starts by building the edges of the puzzle. There is much more personal choice associated with early pattern recognition and matching. Both analysts and puzzle enthusiasts may opt for the easiest patterns to recognize, the largest, or perhaps the most interesting. Regardless, after completing the first selection, you simply move on to the next until you run out of distinguishable pieces.
The End is often very manual and mechanical
In a typical puzzle, there are often a few remaining areas of background with limited visual distinction. At first there may be subtle shading or coloring to help but in any larger puzzle, there will be a considerable area that lacks these traits.
Here segmentation can have another advantage. An analyst may be able to group these pieces as their own segment. Problem solved. But often these areas are spread out across the puzzle and may not group so easily into a single unified group of their own.
In this situation, the analyst and the puzzler once again have similar alternatives. The first is to use association. Put simply, find the piece that your your piece connects to. Unfortunately, this is much easier to say than to do.
To facilitate the final stage of a puzzle, I often group the remain pieces based on their shape. Standard puzzles have six different standard shapes. Dividing the pieces in this fashion speeds your ability to find where they connect. Segmentation analyst can often employ an analogous process using more subtle statistical divisions of their remaining populations.
In the end, the puzzle needs to match the picture on the box. In the end, the segmentation needs to fit the purpose you built it for. Now if that was one to rule them all… well, good luck with that.
One final similarity
If you can’t stand puzzles, I have met quite a few people who can’t, you probably won’t enjoy marketing or customer segmentation either. You may be tempted to leave the segmentation to a neural network or a black box solution courtesy of your local data science department. Typically this doesn’t end well.
As noted, the success or failure of any segmentation is based on its ability to provide value when targeting customers, defining treatment strategies, or measuring performance. If you are not familiar with the construction of the segmentation strategy, as in the case of our non-puzzler, you are not likely to see the early adjustments that might save a slightly flawed model. This will lower your chances of success.
Hopefully you enjoy a good puzzle. Use the tips above to enjoy a good segmentation strategy, too. In the end, if it doesn’t succeed in creating the value you needed, you can simply me happy that you get to puzzle it all out again.
