Building a Retail Value Segmentation Model in Less than 10 Minutes

Building a Retail Value Segmentation Model in Less than 10 Minutes

The first step in customer-centric retailing is to get a sense of which of your customers are the most important to your business and which are not. There is nothing unique about this approach or point of view and to rephrase the worn out analogy of the old time store keeper, he knew exactly who his most important customers were and who they weren’t.

This knowledge has not translated very well to the modern day multi-store retailer who finds themselves treating every customer pretty much equally for the most part. This of course is sheer lunacy in terms of resource allocation and with the razor-thin line that is the difference between profitability or not in retail today intelligent resource allocation is far from a luxury, but rather a must have.

Every retailer fundamentally understands that not all their customers are equal and some contribute significantly more to their revenue and their well being than other customers. The challenge always is to understand easily which customers are which, without embarking on a massive and costly analytics project.

In this short video we show the principles of a useful retail value segmentation methodology, as well as how 11Ants RAP can be used to build a retail value segmentation model in less than 10 minutes, using the 11Ants Customer Value Segment Module.

The first step in customer-centric retailing is to get a sense of which of your customers are the most important to your business and which are not. There is nothing unique about this approach or point of view and to rephrase the worn out analogy of the old time store keeper, he knew exactly who his most important customers were and who they weren’t.

This knowledge has not translated very well to the modern day multi-store retailer who finds themselves treating every customer pretty much equally for the most part. This of course is sheer lunacy in terms of resource allocation and with the razor-thin line that is the difference between profitability or not in retail today intelligent resource allocation is far from a luxury.

The challenges most retailers have with value segmentation models is where to start. The path of least resistance is to pick up the phone and get a consultancy to do the work, invariably resulting in an invoice for tens of thousands to hundreds of thousands of dollars for what generally becomes a static piece of work which become stale relatively quickly. Alternatively someone slaves away at in house for a while. Or the most common thing is: nobody does it, and we carry on with business as usual.

What we will show is how we can build value segmentation models in a few minutes utilizing 11Ants RAP. This is done utilizing the CUSTOMER VALUE SEGMENT MODULE.

Screen Shot 2016-02-16 at 3.30.34 PM

The first step is to define two equivalent measurement periods – most common would be 2 x 52 week periods (the last 52 weeks and the immediately preceding 52 weeks). You can use any periods you like, however keep in mind the longer the periods the more likely you are to have meaningful visit frequency numbers (i.e. customers are more likely to visit multiple times in a 52 week period than they are in a 1 week period). In this case we have selected the last 52 weeks, compared with the 52 weeks immediately prior. You can also select whether you would like to perform the segmentation based upon revenue or profitability (GP).

Screen Shot 2016-02-16 at 3.32.30 PM

Keep in mind that you can also filter by pretty much anything, which means that you could build an identical model for a specific store, specific region, specific category or even a product. In fact anything you can filter on you can build your model, and the process is pretty much identical. These filters appear in all 11Ants RAP modules. For this example we will not use any filters – so we just go ahead and click on ANALYSE DATA.

Screen Shot 2016-02-16 at 3.33.23 PM

The first graphic takes stock of how your customers looked in the Baseline Period. This could also be known as Period 1 or P1, but the most important thing is that it is the first 12 month period – which acts as our reference period.

What the model does is takes our customers behaviour into account on two dimensions Number of Purchase Occasions (How often do customers visit?) and Revenue per Purchase Occasion (How much do customers spend when they visit?).

Our model will assign all customers to one of four segments which you will notice RAP labels broadly based upon what our prudent strategy should be for each of them: ‘Take Good Care of Me’ (high visit frequency, high spend per visit customers); Give Me More Reasons to Visit (low visit frequency, high spend per visit); Get More of Me When I Visit (high visit frequency, low spend per visit); Help Me Take You Seriously (low visit frequency, low spend per visit).

So every customer is now assigned to one of these groups based upon their visit and spend activity. You can think of it as a the Frequency and Monetary aspect of a RFM model (with the Recency being handled by the date periods you selected).

RAP will then tell us some further information about the customers. We can see below what is pretty typically, yet generally startling new information for most retailers – that in this case 5,165 customers or 25.71% of the customer base account for a staggering 62.28% of our revenue. Hence the label – Take Good Care of Me! There are a few other metrics reported on each of the segments including Average Revenue per Customer, Average Purchase Occasions, and Average Revenue per Purchase Occasion.

Now a word on each of these segments and strategies to use this information to drive retail growth. As we mentioned the labels have been named to coincide with the prudent strategy for that segment:

Take Good Care of Me – These customers have high visit frequency, and high spend per visit. Look after these customers at all costs. These customers are the life blood of your business and you can afford to spend 23x as much on protecting Take Good Care of Mes as the customers in the Help Me Take You Seriously segment.

Give Me More Reasons to Visit – These customers have low visit frequency, and high spend per visit. They tend to spend a lot when they come in, but they don’t tend to come in that often. I feel like I identify with this segment when it come to clothes shopping. I don’t do it that often, but when I do I tend to spend a lot more money than I ever intended when I first walked into the shop. The strategy for this group is – as the name implies – give them more reasons to visit. It is probably well worth offering a free tee shirt to get a Give Me More Reasons to Visit into a clothing retailer, than it is to say a Get More of Me When I Visit. This because the probabilities are on your side that when they come in they will spend more.

Get More of Me When I Visit – These customers have high visit frequency, and low spend per visit. They tend to come in often, but not spend much on each visit. Our strategy with them should be to build their basket on every visit, and try to identify categories they could be buying from but aren’t and try to wean them onto spending more with us on each trip.

Help Me Take You Seriously – These customers have low visit frequency, and low spend per visit. These are unengaged customers. The ones around the fringes we should be trying to migrate to more profitable segments.

Screen Shot 2016-02-16 at 3.34.46 PM

The second graphic shows for the Post-Intervention Period. This could also be known as Period 2 or P2, but the most important thing is that it is the second 12 month period – which shows how we have progressed, or regressed, compared to the reference period. What we are trying to see is, has the past year of interventions (offers, promotions, discounts, events, etc. ) had the effect of growing our more profitable segments. We can see that in the below case all of the segments have grown – meaning the business has grown overall. At a different retailer (or a region within this retailer) we may see that some segments have grown and others have gone backwards.

Screen Shot 2016-02-16 at 3.34.46 PM

We can also study specific migration between the value segments between the two periods for example if you look at the second to last row you can see where all of your P1 Take Good Care of Mes finished up. You will note in the last column that 50 of them disappeared altogether. This would be a group (with an annual spend of more than $11,000) which would be well worth spending money on a win back campaign, where as the 1,749 Take Us More Seriouslys would not justify the same per capita spend. Again we are trying to most intelligently allocate our resources.

Screen Shot 2016-02-16 at 3.38.15 PM

We will likely wish to record every customer’s value segment in our CRM. We do this simply by clicking on the button DOWNLOAD CUSTOMER LABELS TO A TEXT FILE:

Screen Shot 2016-02-16 at 3.38.55 PM

This will generate a list which looks like this, which will for each customer show segments both for P1 and P2 – this can then be imported into your CRM system so you can filter customers by these segments and look at their behaviour changes.

Screen Shot 2016-02-16 at 3.39.25 PM

In conclusion we’ve very rapidly transformed raw transactional data into a powerful value segmentation model which provides us with an entirely new level of understanding of our customer base. This all done in a matter of minutes and without a reliance on understanding any sophisticated mathematical modelling techniques.

If you are interested to learn more you are welcome to get in touch with any questions you can get in touch here.