The New Customer Analytics
During a radio interview last week, the host asked me why customer analytics are attracting so much attention lately. The question took me aback for a moment because the answer seemed so obvious — marketing using analytics produces far superior results than marketing done by the seat of your pants. We’re seeing a jump in business as more and more companies recognize this.
But the obvious answer ignores what is producing these superior results. On the radio show, I attributed the surge to two factors, the availability of useful data and the development of new tools and methods to work with the data. The combination of these two elements is creating a new era in marketing.
Everyone knows about the data: it’s piling up faster than most companies can use it. However the bigger (and less publicized) advances have been on the tools and methods side. One change has been a proliferation of metrics used to characterize customer behavior. The traditional metrics have been, for decades:
- Recency—how long ago was the last purchase
- Frequency—how often is the customer purchasing
- Monetary value—how much has the customer spent
Recency still rules the roost. Most companies (and just about all direct marketers) still characterize their customers as belonging to their “12 month file” (purchased within the last year) or “three year file” (purchased within the past three years) or whatever time period makes sense for their business model. The three metrics are often combined into an “RFM Score,” still the most commonly used customer analytics technique despite its many, well-documented shortcomings.
The Predictive Metrics
Now, advanced metrics are gaining traction, characterized by being predictive of future customer behavior rather than measuring past actions like the traditional metrics do. Foremost among these predictive metrics are:
- Customer Lifetime Value — how much will the customer spend with us over their future purchasing years
- Risk Score — the probability that a customer will not make another purchase before becoming inactive
- Likely Buyer Score — the probability that a customer will make a purchase within the near term (typically thirty days)
- Purchase propensities — probabilities or other measures of likelihoods of individual customers buying specific products, both of the type previously purchased (up-sell) and not previously purchased (cross-sell).
These newer metrics are driving a marketing revolution. Because they are predictive rather than backward looking, they are more actionable. Customer Lifetime Value measures the long-term worth of a customer, helping marketers concentrate on their more valuable customers. Risk Score is the most accurate, quantitative measure of customer loyalty and thus key to running retention campaigns. Likely Buyer Score lets marketers target those customers who are now ready to buy, answering the previously unknown question of when to communicate.
Purchase Propensities are probably the most important of the predictive metrics because they enable individualized marketing, with unique offers going to each customer based on their needs and wants. Campaigns based on these more relevant messages and offers have significantly higher response rates and deliver more revenue, and these are the numbers that are ultimately driving the more widespread use of customer analytics.
New Calculation Techniques
Underlying the new metrics are new calculation techniques, and here is where the heart of the marketing revolution is found. Long ago, marketing was “one-to-one” when the corner grocer knew exactly what each customer liked to buy. Mass marketing changed that approach, flooding our mailboxes and inboxes with what we now call spam. Everyone got the same postcard, the same email, the same flyer, the same catalog. Today we call this ‘spray and pray’ marketing.
The next wave brought segmented marketing, with statisticians dividing the customer population into groups based on demographics or behavioral data. Instead of one message to everyone, there are five or ten messages. Which one you get depends on the segment to which you belong. Statisticians use regression techniques to build models that hopefully predict behavior, while the marketers cross their fingers and hope that the segments hold homogeneous customers. Among many sophisticated companies, this is the most popular approach today.
Marketers who want to get ahead of the curve, however, are deploying a more powerful technology, individualized marketing. Instead of treating the customer population as one homogeneous group, or as a collection of multiple segments, each customer is analyzed individually to make predictions about when they are ready to purchase and what products they are likely to buy. Those marketers in the vanguard realize that you can’t produce analytics at the individual level (for example a different Risk Score for each customer) unless you model and calculate at the individual level.
The methodology needed to analyze this way is different too. RFM doesn’t work, and neither do neural nets or regression models. Finite State Machine modeling, which builds a mathematical model for each customer that describes future behavior in terms of the customer’s past history, is perfectly suited for the individualized approach. Analyzing at the individual level and then making individualized offers based on that analysis create exactly the conditions that enable individualized marketing and this new era of marketing.
Perhaps technology is finally ready to make today’s marketer as smart as yesterday’s corner grocer.