Market segmentation

Segmentation is a major element of marketing. At its simplest it involves finding groups with different preferences and different levels of willingness to pay and creating products and offers that target these different groups - better matching what your offer against what the market will pay for.

Segmentation is often used to describe the process of finding the segments. This can be through market or survey research, or through database or other data analysis. However, it is also important to realise that segmentation also involves using the groups for business development. As such, it normally comes with a cost and complexity overhead so segments need to be robust, replicable and have sufficient potential sales volume to make them worth addressing.

At a research-level there are four major ways of segmenting a market, partly according to the level of precision required and the type of data and analysis available about your customers. However, in finding different market segments it is important to keep in mind that the business will have to use the segments and implement segment specific investments and plans for segmentation such as tailored products, pricing or service to meet the needs of each target group.

Important questions are therefore how are you going to place customers into each group and how are you going to target and track each group. Do you leave it up to the customer to select themselves into a segment, or do you have specific segment sales managers?

1. A Priori segmentation

A-priori (pre-existing) segments are the most basic way of creating market segments. In A-priori segmentation, the market is split according to pre-existing demographic criteria such as age, sex or social economic status, or other criteria available, particularly if using databases or lists as a starting point.

More sophisticated versions include lifestage (which combines information about age, presence of children and working status) and geodemographics such as Experian's Mosaic or CACI's Acorn classification systems where households are allocated to specific clusters on the basis of typical household make up and housing type. For website analysis, segments can include country, browser type, or previous history.

A priori segments are easy to define and easy to target with advertising and media. For some sectors, for instance technology, there are such strong relationships between age and use, that a priori segments are all that are needed. However in other markets - for instance drinks, it is more difficult to use pre-existing variables for segmentation.

A priori segmentations are also the simplest segmentation to apply and use. A database can be flagged or sorted on the pre-existing data and that data used to drive sales and marketing campaigns. Similarly, on a website basis, segments can be identified from session histories and IP identifiers.

However, although better than pure mass marketing, even the most sophisticated a priori systems are quite crude. In geodemographics there is the assumption that you buy or think the same way as your neighbour which is clearly not always the case.

2. Usage segmentation and use cases

A usage segmentation is based on prior behaviour. There are two main approaches: firstly customers can be split according to their weight of use, or weight of purchase - heavy users/buyers being more important targets than light users for most businesses (also known as decile analysis or pareto analysis)

The usage history can then be used to tailor products or services, for instance tiering extra services, or account management for the larger purchaser groups.

A simple transaction history or visit history on a customer databases, or visit histories for online behaviour can be used to split the database according to weight of sales, value of purchases, number of visits etc. Decile analysis splits users into 10 evenly numbered groups, which Pareto analysis splits the top 20% from the bottom 80%. This is normally used in business-to-business markets and is a core part of database analysis for consumer markets.

Secondly, usage can be considered in terms of time, place and reason. A café might sell sandwiches at lunchtime but main meals in the evening because the purchasers are looking for different things. It may even be the same purchaser just in a different "mode".

Similarly, a photographer may have different cameras, each for a different function or use-case. Understanding 'use-cases' may be an added layer to understanding individual customers.

Usage studies are also extremely common in market research, but normally to determine measures of market share and other market metrics. However this information can also be used as the basis of a segmentation approach - checking heavy versus light user groups for different attitudes or behaviours.

Often usage segmentation is used to try and establish underlying driving forces from other demographic variables. So if women are more likely to be heavy users would it be easier to convert more light users who are female, rather than target their male counterparts. This focusing of market activity on groups that are similar to heavy users gives rise to measures such as "uplift" - the improvement possible over a purely random approach.

3. Attitudinal research and cluster analysis

When market research is used for usage studies, it is also often accompanied by attitudinal research - what do customers think or believe about the category in question. This is commonly achieved through banks of agree-disagree scales or ratings out of 5,7 or 10. The aim of these studies is not just to understand commonalities in opinion, but also what makes one group of users different from another.

To understand how attitudes affect purchase statistical techniques such as "cluster analysis" are used where people with similar attitudes are combined together. For instance grouping those for whom the environment is important separately from those who think price is more important.

This information can then be used to target groups by what they think and how they feel, rather than just who they are. This is particularly valuable in determining branding strategies and keeping a brand in tune with consumers.

However, attitudinal clusters do not fit easily into database or conventional media targeting which are more often than not based on demographics. The translation from attitudes to demographics means that some of the usefulness of an attitudinal segmentation is lost. Companies can reach different attitudinal groups by offering a range of products and a range of communication, but clearly the lack of a clear definition means cross-over between the targeting of segments.

Attitudinal grouping also suffer from some problems with regard to their robustness and replicability. Cluster analysis cannot be carried out in the field so scoring systems (similar to credit scoring) or surrogate measures and variables are needed to allocate individuals to a group. These additional measures can be guessed at, but normally need to be defined and tested post-hoc. Repeating attitudinal analysis successfully can be very difficult and expensive.

Attitudinal groups may also change or move over time as some views become fashionable or unfashionable. It is possible to find a segmentation that quickly disappears or is superseded by events (imagine the music market). There is also debate about how attitudes change - is it the advertising and the product that create the attitudes, or do the attitudes lead to the choice of a particular product. In particular a single individual in different circumstances or mode, may fit into a different segment. Capturing this complexity in a single dimensional study is difficult.

Nonetheless, attitudinal clustering remains one of the most common forms of research-type segmentation. If you find different attitude groups - for instance an environmentalist group versus a buy by price group, there can be very clear distinct differences in messages, distribution and product for genuinely different attitude groups.

4. Needs based segmentation

The fourth method is to try and determine fundamental drivers for the decision to create what is known as a needs based segmentation.

Most needs-based segmentation uses Conjoint Analysis to split a category into different levels of functional performance (see Conjoint Design). By understanding what elements are key drivers for individuals, specific needs and requirements can be identified from the trade-offs that each person makes. Using cluster analysis, this information can be drawn together to find different segments with similar preferences and needs from the product category in question.

Needs based segments are typically the most actionable forms of segments as you know what drivers and performance the product or service has to satisfy. These are normally more stable than attitudinal groups as they should directly reflect and predict existing market share.

However, like attitudinal studies, because cluster analysis is used, targeting each of the underlying groups can be difficult.

Nonetheless you also have the benefit of being able to produce a market model or market simulation using the Conjoint output.

5. Other forms of segmentation

Those these four approaches describe the main methods for segmenting a market, alternative approaches can also be taken. For instance, using event or behavioural segmentation is easier and more common for businesses operating online. For instance, an e-commerce system will be designed with different routes for different types of customers - from fast offers for buyers wanting speed, to more detail for those browsing and considering. This can include targeting by mood or event - for instance focusing on convenience for buyers in a rush, or concentrating on presentation and packaging for customers looking for a treat.

Choosing a technique

The type of segmentation you use will depend on a lot of factors including the cost not only of conducting the research, but also of implementing the solution and the investment the business wants to make. Consequently ideally for each segment or group you want to know what the economic value and the economic potential for each group is and have some idea as to whether this is increasing or falling. Consequently most quantitative segmentation studies are detailed and complex.

Mixed approaches

In practice, a good segmentation will include several different levels of analysis and validation. Since segmentation have to be strong and clearly defined, a cost-effective approach is to develop groups based on qualitative research. Typically a business wants to minimise the number of segments it has as each costs money to target properly (database marketing and digital printing techniques allow for far finer targeting without too much additional cost). With small numbers of big segments, a good researcher will be able to identify these groups within a programme of qualitative research. This will not gather economic data, but it enables deeper insight into each group and, if monitored over time, provides core information about how segments change and develop.

Our segmentation research is aimed to be holistic in approach. We look at existing sources of data and factors such as database, purchase and usage information available. We also like to use qualitative research to pre-identify potential bases for segmentation. From this we design research to address the specific market circumstances, optimising any sampling for different groups. We then use a range of statistical tools to investigate and validate segments, building models and forecasts for targeting.

For advice on the best type of segmentation approach for your market contact info@dobney.com.