Customer Segmentation in the Cognitive Age
Segmentation is well established in Marketing and CRM as a way to systematically subdivide a market or a customer base into discrete groups that have different needs or require distinct treatments.
The basic objectives of segmentation are not rocket science – better targeting and/or better service – and yet the means of achieving these seem to be becoming ever-more complex. Apparently, you’ve got to be a Data Scientist now before you’re allowed near an analytics tool!
MARKET VS. CUSTOMER SEGMENTATION
Firstly, I’d like to briefly draw a distinction between Market and Customer Segmentation.
Market segmentation operates at demographic or geographic levels, which when combined with researched psychographic and behavioural data creates distinct groups or personas that bring clarity a company’s product and marketing strategies. Market segmentation can be completely anonymous, merely referring to categories of product or types/personas of customers rather than any individuals. This has led to a mixed track record of overlaying market segments back onto customer databases. It also fuels the Data List Broking industry, which seeks to provide highly targeted prospect lists based on such characteristics.
In this blog, however, I’d like to address the subject of customer segmentation, which has many similar elements to market segmentation but approaches it in the other direction i.e. starting with your own database of prospects, customers and ex-customers.
Typically a customer segmentation project starts with the behavioural and value data available, which is ever-increasing and the best companies are surfacing and including their ‘dark data’. It groups customers along specified dimensions (e.g. ‘needs’ and ‘value’), and then looks for geodemographic and attitudinal discriminators to refine the parameters for segment allocation, with an expectation that all customers on the database will be reliably allocated to a segment.
Does this all sound very 20th Century? Well, yes, it is. Data has got ‘bigger’, channels have proliferated and analytics tools are more sophisticated, but the fundamentals of segmentation haven’t changed much since its inception and nurture in the second half of the last century.
IS SEGMENTATION’S TIME UP?
It could be argued that the combination of IoT and Cognitive learning power (exemplified by IBM Watson), and the enablement of true 1:1 personalization (at last!) is sounding the death-knell for segmentation.
This may well be the case for online or digital user journeys, where cognitive technology will instantly analyse all customer engagement and value data, apply predictive modelling, and create a unique personalized marketing treatment to present ultra-relevant content “in the moment” that might never be offered to anybody else ever.
Does that sound conceivable? Many blue-chip companies have already embarked on “digital first” product and customer experience development strategies incorporating Design Thinking to fast track towards this ultra-connected experience. In my estimation, it’s real-world achievable in the next 10 years (assisted by other innovations such as connected & autonomous cars) – starting with premium and high-value brands in digitally-sophisticated markets. Is segmentation’s time up? Not yet!
Until the point when every customer journey is digitally enabled or assisted, however, we all will continue to stand in the same lines/queues, drive on the same roads, and walk through the same doors!
This means that segmentation will still be required, at least to the medium term. This gives marketers and data scientists a clear window to improve, optimise and eventually replace their segmentation approaches i.e. to evolve alongside both the cognitive tools and the customer appetite/expectation for digitally enabled personal experiences.
There’s a wealth of helpful resources on the IBM Think Marketing website to help you succeed during this window of opportunity.
ATTITUDINAL, BEHAVIOURAL, OR BOTH?
One thing that marketers and analysts must “get over” is the argument about the competing benefits of behavioural vs. attitudinal customer segmentation.
The ‘behaviourists’ would assert that only behaviour is ‘real’ and therefore a valid basis upon which to predictively model. Advocates of the attitudinal approach argue that analysing behaviour gives the “what” and the “how” but not the “why”.
In my view this is one of the great opportunities of the Big Data / Machine Learning / AI / Cognitive revolution – that attitudinal data can be systematically captured and/or inferred. This enables attitudinal and behavioural data to be combined (please again see the Watson Personality Insights service referenced).
A simple example of my point is the Net Promoter Score (NPSTM), which gives three basic attitudinal segments regarding whether customers would recommend your product/service – Promoters, Detractors and Passives (AKA Neutrals).
Behaviourists criticise this as it’s only intention – they want evidence that Promoters are indeed advocates for the brand who stimulate incremental sales (and they also don’t have up-to-date NPS scores for every customer in the database).
I’m not advocating it, but it is now possible for a NPS-based combined attitudinal and behavioural segmentation model to be build and productionized should a company wish to ground its customer strategy in NPS. It would have up to 9 segments of Promoters, Detractors and Passives who do or don’t recommend, or publicly disparage (illustrated).
Wouldn’t it be great to add a value dimension to this?!
FOUR TOP TIPS
To successfully incorporate cognitive technologies in your customer segmentation, you need to:
- Basic rule of segmentation – don’t segment customers at all if you’re going to target and treat everyone the same! I have seen academically beautiful segmentation projects that go nowhere because they’re not supported by the people and teams that actually do the marketing, selling and customer servicing (who are still doing it the old mass marketing way)
- Be willing to invest in data scientists and ever-more-sophisticated analytics to identify ever-more-targeted groupings of customers (micro targeting)
- At the same time keep it simple as is humanly possible for staff to understand the categorization of customer types to enable customers to be treated differently by humans based on need or value or preference, etc. (a practical example is illustrated in this blog’s graphic). It is vital that your segmentation strategy does not ‘fall at the last hurdle’ in a physical channel because it’s too clever for its own good! I am currently conducting a segmentation project with a client that’s a relatively straightforward multi-dimension model. This generates tens of thousands of cells, which we have to boil down into a sensible number of discrete segments that mean something for marketing campaigns and customer interactions. In the future, it will be tens of millions of cells!
- Strive for ever-increasing personalization so that your marketing communications will never feel generic or ‘scatter gun’ by the customer. It is not comprehensively proven that customers will value (i.e. pay for) the ‘perfect’ 1:1 personalization that I described. Indeed, in the short term they won’t mind that they’re not the only person receiving the offer as long as it’s appropriate to their needs and circumstances, and presented via content that engages them