How to segment, target and position (STP) in one integrated approach

Traditionally the STP process can be unwieldy and seem disjointed as the variables used to segment and target may be different to those used to position a product. They may also come from different studies.

In this article I’ll show you how conjoint analysis can be used to integrate all 3 aspects in one survey and within one analysis tool; while yielding the added bonus of linking to market shares and revenues.

The segmentation process, the targeting process and the positioning process are all formed from measures of the impact of different product features, brand names and pricing. More specifically these values are called ‘partworth utilities’. These are calculated per survey respondent based on consumer reaction to different product profiles and tell us how much impact each feature (i.e. attribute level) has on purchase likelihood.

Cluster Analysis: In Metrisim running a cluster analysis is easy.  Select the segment tab in Metrisim and indicate how many clusters you want and then click 'Run Analysis'. The cluster analysis uses consumers’ individual partworths to identify clusters of consumers with similar preferences, while simultaneously trying to make the clusters as different as possible. For example, if you use the UHT demo files, you will notice one cluster with consumers who are price sensitive but not as worried about brand or environmentalist issues such as recyclable milk cartons; and another who don’t seem to care as much about price, but are more sensitive to environmental issues.

 

Evaluate the potential of each segment by projecting revenue for your product in a competitive scenario.

The RPK stands for revenue per 1,000 customers. What customers means, depends on your filter selection - if 'all' then customers in the broader market, or if a specific cluster, then of course only customers in that cluster / segment. The RPK is simply calculated by the simulator as price per unit x (share % / 100) x 1000, so an adjustment is needed for volume (since each individual won't just buy 1 unit in this market (i.e. 1 litre in this case)). As an approximation, you could multiply the RPK by the average number of units purchased per annum. A more accurate approach would be to first use Metrisim's weighting feature. Open the partworths file (pws.csv) and replace the default 1s in the Weights column with each consumer's category purchase volume. The share %s will be more accurate since they now account for individual variation in units purchased. Imagine for example, that those respondents predicted to choose our brand - Kinder Green - had higher than average category purchase volumes; the weighting would then increase our share %. While the RPK is also different now as it is derived from the share %; you still need to multiply the RPK by the average number of units purchased per annum to get an estimate of Revenue per 1,000 customers in the segment.

This will give you an idea of which segments are most attractive for targeting.

While you got an idea of what segments to target in the previous step, in the positioning step you can take this further by running simulations for alternative product product positionings within each segment. By adjusting pricing and other attribute levels, you can see how changes impact projected market share and RPK (revenue per 1000).

For reference you should run a mass-market simulation – where you have a single standard product and all consumers are included on the filter. In this way you can verify that a segmented approach will lead to better revenues than a mass market approach.

The specific levels that work best in each segment, can also help determine what claims to focus on when advertising to different segments.