November 16, 2015

Divide and rule?

 

The benefits of moving from customer relationship management to customer segment management

 

What differentiates customer segmentation from other targeting techniques?

 As marketers we all need to differentiate the ways in which we recruit and look after individual customers, and a myriad of different techniques have been developed to achieve this from badging products as gold silver and bronze, to refusing credit to individuals perceived to carry a high level of risk.

The key distinction however lies between individual differentiators like scores, and group differentiators like segments.

 

We segment the natural world by genus, species and sub-species, but at an individual plant or animal level there remain huge differences within a sub-species. Similarly in marketing; although we can use segmentation for many purposes that does not exclude our also differentiating at an individual level.

(And, to our great relief, whereas the definition of a species contains its capability to inter-breed, this is fortunately not expected of humans inhabiting the same marketing segment!).

So the question becomes how and where to use segmentation as a marketing technique, and how and where to use individual level discriminators?

Broadly, segmentation can be used almost universally except where the financial benefits of applying individual level discriminators outweigh their cost; examples where they do are often where an individual presents a high level of risk (such as for insurance or credit) or where individual level targeting for specific products or propositions is worthwhile.

 

 

How to use segmentations in customer marketing?

 It would be simpler to answer the question ‘where not to use segmentation’ as the applications for segmentation run through almost every area of marketing activity. We will aim however to highlight some of the key applications:

 

  • Recruiting

 Assuming an existing customer base with an applied segmentation, the marketer has a live sample within which measurements can be made of customer value, loyalty, product preferences, channel preferences, and responsiveness to different types of content, discounts etc. If further information like the segment’s interest in competitors is required, then research can be undertaken at a segment level. Some larger companies even build segment level continuous market research panels.

 

This segment level knowledge then allows the marketer to focus on recruiting profitable segments with known lifetime values, and product, channel and proposition preferences.

(At Berry Thompson we use a very simple segmentation between Clients and Marketing Service Providers (MSPs). We know that MSPs are twice as likely to accept a proposal as clients, but are less influenced by outbound marketing as they pick up the phone when they need us. This informs our recruitment strategy as we keep MSPs gently aware of our service range, but target client groups with specific offers).

Below is an example of a segmentation developed for the soft drinks industry to guide brand managers in their recruitment activities:

Segmentation profile

Fig. 1 A chart showing the volumes of customers in a segment, their consumption of soft drinks across brands, their value, and their lifestyle and life stage

 

A word of caution; for a segmentation to be able to be employed when selecting media, there must be a common currency between how media can be selected and how segments are defined. We deal with this later in our section on data integration.

  • CRM

 We suggest that segmentation is so important for servicing and selling to customers that each segment deserves its own business model and development plan.

At the business model level we know key financials like what it costs to recruit customers, their life-time value, their loyalty, and their responsiveness to buying more from us.

This in turn informs what we should spend on them and how they should be serviced.

Some work we did for a large UK home shopping company revealed that for each segment there was a specific and different level of spend on catalogues that was required to optimise ROI; some segments would tend to continue purchasing during a season as more and more was spent on them, whilst others had a ceiling and would not spend more in response to additional catalogues being sent out.

Contact density segmentation

Fig. 2 A chart showing how customers with different expected spend quintiles provide different cost to sales ratios from different levels of communications costs

 

 

For each segment we can consider:

– the suitability of products we have and whether new product development is required

– the level at which to pitch pricing in the context of the competitors they like to use, and also their propensity to respond to price offers

– the level of service expected and how this can be delivered

– the kind of content they read (when for instance browsing on the internet)

– the channels they like to use (e.g. mobile v laptop)

 

 

As these examples demonstrate, it is not sufficient just to have developed the segmentation structure, but it is often necessary also to conduct further tests, analyses, and research within specific segments.

For example product strategy will require research, pricing can involve live tests, content liked can be parsed using tools like Idio, channels used can be analysed. The odd man out is servicing, where your judgement comes in – you won’t want to test bad service until you find the tipping point at which customers attrite!

  • Data integration and reporting

 The segmentation framework provides us with the means for pulling data together from multiple sources.

For instance individual members of a segment can be matched up to external customer data providers such as Experian or Callcredit; these data suppliers can match in up to a thousand variables for each customer, from which an understanding can be gained at a segment level of e.g. affluence, lifestyle, credit risk, spend by product category, etc. etc.

In a retail environment, where there is not always a recorded customer base as such, a segmentation can still be used provided it has axes that can be matched into external sources. We have a preference for using life-stage and affluence as axes when building segmentations that need to be mapped externally because of their near universality.

 

If one considers a cell comprised of an affluence and life-stage grouping, it is the possible for a retailer to understand the passing volume of traffic via a mobile phone operator, and the value of it via a card processor. This valuable information can help with what is presented in store, and where to locate stores.

 

A typology of segmentation techniques

 There is no one-size-fits-all when it comes to designing the right segmentation structure for an organisation.

Step one is to brainstorm amongst people who know the business and its customers well, what they believe are the key factors that discriminate between different kinds of customer.

As part of this one needs to consider how the segment is to be attributed. For instance a market research based segmentation is not directly attributable (without modelling), a transactional based segmentation is only attributable to your customer base, where as a life stage and affluence segmentation can be transferred across databases and can be used as a currency.

Some examples of attributable segmentations are:

For a large book seller we worked for it was the type of book purchased.

For charities it can be the personality type involved.

For a media company it can be the content they are interested in.

For a life insurer it is often life-stage.

For home shoppers it is conventionally RFM (recency, frequency, monetary value, although we would challenge this when not mixed with other factors like the type of product purchased).

 

A robust segmentation often brings together multiple factors when clustering customers into groups that are most homogeneous.

There are other factors that we would see as helping to describe segments once built rather than as reasons for clustering customers together. These include their geographical distribution, channel preferences, longer term value, and loyalty.

To judge whether a segmentation is successful one has to look at whether it uncovers significant differences in customer behaviour, that are of real use to the business in differentiating how it presents itself.

 

 

How to develop and attribute your own segments

 There are a number of ‘traps for the unwary’ of which the most common one is developing a segmentation that cannot be attributed either onto customers or prospects.

In the early days of mobile telephony a market research company we knew had developed, using a number of external interviews, a typology of mobile phone users which they proudly called ‘tarts and taxi-drivers’. Sadly, or perhaps not, there was nothing in the company’s existing data that could identify a customer as belonging to either of these groups!

When designing a segmentation it’s essential to look at all the data and systems environments in which it will need to be implemented, to make sure that this can take place. For instance a bank with multiple product systems may find it difficult to join detailed data from each of these, although it may have a universal customer record; however the solution may lie in building the segmentation so that it can be imported from an external data source that can be matched to each customer record.

Another common trap is to build a segmentation on too small a sample to be properly attributed. This often happens when research data is used and there is a cost problem with interviewing a large number of people. If you are using propensity models to attribute segments to customers (often customers are allocated to the group for which they have the highest propensity) and say you have eight segments, then you will need at least 10,000 individuals on whose data the segmentation is developed. With less the propensity models are not robust or cannot be validated.

Another trap is to build the segmentation using data that is subject to rapid change, with the consequence of not being able to compare segment performance in one period with the next; this often the case with RFM segmentations where value is impacted by variations in product, pricing, and market forces.

Hence for a number of reasons we often find it safest to build our segmentations within axes of affluence and life stage.

 

Lifestage affleunce definition

Fig. 3    An example life-stage and affluence based segmentation

 

The segmentation is then built within these axes by looking at the averaged characteristics of the customers in each cell they form.

These can include any of the factors mentioned above such as product types and amounts purchased, content viewed, channels used, loyalty etc. The clustering is then done at a cell level rather than at an individual customer level, based on the chosen factors.

Each segment can then be sized on your customer base, against the UK or your market. This can then support planning and targeting across all channels. For instance, the marketing director could set a growth target for the young affluent segment which can be used by CRM, digital and above the line channel managers.

Other key advantages of this approach to segmentation are that:

– external data such as market research or street traffic can be imported and mapped onto the cells to build up a stronger description of them; and this process can be on-going

– the segmentation can be attributed onto any customer or prospect base, by being first applied to an external data source that holds life-stage and affluence from one of the large data providers

– because life-stage and affluence are such strong discriminators in the first place, little of the discrimination between segments is lost compared with using a conventional clustering approach

– it is simple to control the overall number of segments

– and possibly most important of all it is very easy to describe the segments to a group of marketers

 

How to deploy your own segmentation

This phase of the process can often be overlooked. In a marketing focused organisation you may already have several segmentations in place. But have the need to combine these into one unified truth.

In other instances there may not be a segmentation in place. Hence the concept needs to be explained across the business.

In both cases the relevance of the segmentation will be key in its adoption; hence the first brainstorming phase with all stakeholders is vital.

Once the segmentation has been built, it is then important to have two sets of deliverables. The first focusing on how the segmentation was built, and why it took its form. The next should be a detailed profile on each segment that can be shared across all parties; the insight in this second delivery should support all business functions.