Over 20 years ago, database marketers started formulating new segmentation objectives to reflect a meaningful and actionable blend of behavioural data and imputed, fused or collated attitudinal information.1, 2, 3 The goal was to organise knowledge and marketing strategy, as well as communication and service behaviours, by applying a combination of descriptive, predictive and operational targeting and execution capabilities. Marketers aimed to understand the brand's customers in terms of segments that were:
collectively related to the brand according to universal properties of the brand (in the absence of which it would make sense to create a new brand);
heterogeneous with respect to other segments with the differences rooted in brand choice, not merely in broadly held values, geodemographic or lifestyle characteristics;4
relatively stable over time;
relatively homogeneous within themselves, so that they might also be targeted as a meaningful community (or tribe) with predictable results;
although liable to member individuation, typically because of instability over and in time, for example with respect to timing of an offer or selection (eg changing lifestage status or from prospect to customer — since the goal is to find prospects like your best customers), enabling targeting and execution of appropriate actions at the right time;
insightful, descriptive and predictive, so that marketers and the organisation can understand the different segments (or communities) and develop appropriate strategies, innovations and communications messages;
employing systems and organisational capability that enable strategic, tactical and operational interventions at market, segment and individual customer level;
as well as having tracking and evaluation research capability.
The model relationships can be expressed graphically in several forms, including the two versions in Figure 1.
This three-dimensional model differs from both the competing models: one-solution-for-all, which communicates to everyone in the same way, and mass customisation, which assumes a mass of individuals receiving individualised communication from the brand, sometimes described as ‘segment-of-one’. Its purpose was to ensure understanding, insight and focus across the organisation and drive business and marketing strategy through differentiated value propositions and individualised tactical/operational execution. This became a de facto gold standard among experienced database marketers, and was implemented using a variety of techniques including clustering, factor analysis, regression and data mining, collating and fusing data from across the organisation, third parties, and research. This was successfully implemented on many occasions. For example, working with Ogilvy from 1992 to 2001 involved dozens of projects and a documented global methodology for Ogilvy practitioners in over 50 countries. It also remains a stated goal among many leading marketing thinkers.5
In 2007, I suddenly realised that the gold standard was progressively being abandoned as a result of a combination of project failures and spectacular advances in statistical, data mining and rule-based customer management techniques, particularly in the online and call centre spaces. Woodcock, Stone and Foss, the CMAT development team, were in fact already reporting the trend by 2002 in their CMAT global survey: for example, they showed that only 19 per cent of companies had carried out basic decile analysis, 54 per cent did not recognise customer behaviours in planning, and there was a broadly negative trend in analytical and planning capability since their previous study 3 years earlier (the only exception being customer retention).6 The occasion for my discovery was a meeting of the senior consultants and associates of a consulting division of one of the big consumer profiling houses in the UK. There was a 2-day meeting to discuss case studies, leading-edge practices, tools and methods. Everyone was experienced, smart and involved in significant projects in major client organisations. There was considerable discussion on customer journeys, uplift models, call centre performance improvement, next best action (or best next-action), collaborative filtering, campaign management tool kits, and analytical and process software and its application.
However, not one project involved gold standard segmentation. Yet, if rule-based action works so well in the absence of a strategic understanding of customer types, think how much better it would work in the presence of one.
A conversation with the chief consulting officer yielded this admission: ‘Isn’t that all debunked now? We can’t get it to work and we haven’t found any clients who want it. Do you have any evidence of success?’
Recent personal experiences support this analysis, while market research consultancy firm Incite also report in their Spring 2009 newsletter7 that segmentations form a key part of company or brand strategy in fewer than 40 per cent of companies, a fall of 16 per cent since their similar survey in 2004. They also report a 31 per cent increase in the perception that segmentation studies are complex and require experience to do well. It is true that ultimately segmentation requires expertise to do well. However, the rapid increase in perception of difficulty suggests a rising concern among marketers, consistent with the fall in strategic commitment.