Making segmentation a success

By Rob Daley • 14 June 2016

Everybody loves a good segmentation. Whether it’s the analyst building the model or the marketers brainstorming the wants and needs of the segments, segmentation is often seen as the darling of the analytics world.

But, when all's said and done, how many case studies are there of a segmentation having a fundamental impact on a business? Or, more simply put, actually delivering against the ambition that was set out when the project started.

There are a few common reasons why segmentations can fail, this paper explores these and suggests ways to ensure your segmentation is a success.

1. What’s the plan?

Possibly the biggest reason why a segmentation fails is the lack of clear planning and requirements gathering up-front. All too often the requirements are defined as ‘to inform the targeting for direct mail campaigns’ or ‘to develop propositions for specific customer groups’. These are fine as a general statement of intent but lack the detail needed if they are to guide how the segmentation will be used.
For example…

  • ‘To inform the targeting for direct mail campaigns’ - how will they do this? How will they work alongside your current targeting – e.g. propensity models or triggers?

  • ‘To develop propositions for specific customer groups’ - proposition can mean different things to different people…e.g. a segmentation that is to be used to inform different customer service levels is very different to a segmentation that is to be used to inform creative and messaging.

All in all, the requirements need to be clearly thought through before any analysis or model building is done. A common mis-conception is that you need to see the segments before you can figure out how to use them. This is the wrong way around! Start with the problem you’re trying to solve and figure out what the segmentation needs to look like – in sufficient detail – to solve it.
Ultimately, this means greater focus is needed at the start of a segmentation project. A typical project with a high likelihood of failure might look like this chart below…

Percentage of time at each stage


Typically, the focus is on building the perfect model and the perception is that the majority of the work is done once it’s built. In reality, challenges can occur as the requirements haven’t been clearly thought through and, whilst the model is solid and robust, the chances of it being fit for purpose are greatly reduced.

In contrast, a successful, business-changing segmentation timeline might look like this:

Percentage of time at each stage


In this example, more time is spent up-front gathering detailed requirements, less focus on building the perfect model (does it really matter if you haven’t included a profile of which segments have a dog?!) and there’s more time for deployment – which we will explore later.

2. Is this my job, or yours?

The common mis-conception is that a segmentation project is the domain of the analytics team. Whilst they play a fundamental part, it’s the business owners that use the segmentation in earnest and drive value from the analyst’s work. A typical segmentation project might look like this:


The main challenges with this approach are…

Analysts are good at analysis, requirements gathering isn’t always a core strength. Business owners need to define exactly what they want from the segmentation. A conduit between the business and the analytics team is crucial here, to ensure the requirements are successfully translated into a robust, fit for purpose solution design.

The handover at stage 4 is a key risk point…if the business haven’t been engaged throughout the process then this is likely to be more difficult.

Typical challenges are:

- The business failing to see the value in the model - it’s not what they expected or;

- Resource constraints - not having the time to deploy the model.

Consider the following graph as an alternative model which increases the chances of successful deployment…


In this example, there is greater engagement from the business at the start of the project. Also, each party involved sticks to their core strengths.

3. So why are we doing this?

Fundamentally, a segmentation project is a driver for change. The business is asked to do something different as a result of knowing something they didn’t know before about a particular group of customers.

For example, segmentation can be used to drive culture change by changing who customer facing colleagues interact with and how they do it.

Take the example of providing segment details to a sales team to help them acquire more of your target audience…

It is critical that you understand the motivations and incentives of the sales team if you are to change their behaviours. If they’re incentivised by product sales, are they really going to be engaged with a customer segmentation that will change what they’re doing and how they do it? If however their bonus scheme was dependent on a combination of product sales and customer mix, you may find behaviours are changed quicker!

Change such as this can take a long time, a large proportion of segmentation projects under-estimate the time required to deploy a model so it’s important to be realistic about what is needed and how long it will take to deploy the model.

So, what have we learnt?

So in summary, key considerations for a successful segmentation project are:

Spend time thinking through the requirements to a granular level of detail. Create dummy segments and road-test how you’d use these as part of the planning phase.

Make sure the right people with the right skills are doing the right roles. Segmentation isn’t just an analytics project, it needs involvement from a wide range of individuals across the business.

Segmentation doesn’t stop when the model is built…the hard work starts there! Make sure you’re clear on how it will be deployed and that you have a realistic plan for deployment.

Back to top


  • Making segmentation a success

    Everybody loves a good segmentation. Whether it’s the analyst building the model or the marketers brainstorming the wants and needs of the segments, segmentation is often seen as the darling of the analytics world.

    Find out more