The Math of Loyalty Programs Part 4 Data / Analytics

Thanks for joining me here. This is perhaps the most important and yet least detailed article of the present series. A tough topic to cover without getting overly technical. Please let me know if you think it is helpful despite this limitation. Enjoy!

DATA

The core of a loyalty program is the data it generates which can be linked to an individual customer via a unique membership number in turn linked to a variety of unique keys. Being extremely granular has its uses, but loyalty data can also be totaled in myriad ways to build insight at various levels of understanding from the individual through the segment to the overall base.

Let's start our discussion by listing the broad classes of data a loyalty program can and should collect:

1. Profile data: Member-given (e.g name, address, preferences, interests, etc.) -- members puts up their hands and offer friendship, investing time to teach the organization about themselves. It the organization's responsibility to think this one through carefully, what will we ask? How will we use it? Is it overly intrusive given our brand values?

2. Transaction data: System-generated (e.g. typically spends and related categorization data etc.)

3. Interaction data: System-collected (e.g. visits, social media related, eDM responses etc.)

4. Point related data: System-calculated(e.g. points earned, redeemed, balance, expired, bonus etc.)

5. Financial data: System-generated (e.g. transaction and customer profitability, point liability etc.)

6. Environmental data: Variously-collected, as far as possible, systematically (e.g. season, TOD/DOW, advertising running at the time)

7. Derived data: System-calculated (i.e. data derived, usually totaled, from some permutation or combination of the data sources above e.g. customer value growth, customer life-time value, cohort analysis, clusters and segments, correlations and causation etc.)

I always find it interesting to learn how many of these data types are collected with the intent to be used, and then are used, effectively. (If you are a loyalty practitioner, be honest with yourself now.)

ANALYTICS

The objective of any loyalty program and hence analytics project is to:

[ Maximize the Sum of Customer Life Time Values across the Member Base]

In others words, to build insight and hence systemic organizational knowledge that enable a company to routinely make decisions resulting in cost-effective actions that drive customer relationships forward in the most profitable manner possible.

Analytics in loyalty tends to be a cyclical process: test and learn, test and learn and then roll-out cohort by cohort, segment by segment, individual by individual. How that is done, how test and controls are created and managed, and how much they can cost (a helluva a lot is the quick answer, is a topic for another article, here let's turn to what we want do.

Take cost-effective actions which basically means offering incentives as a reward for profitable customer actions. Incentives, I must stress are not necessarily discounts, which are regrettably the first refuge of the unimaginative marketer, but points, and most critically, intelligently planned emotional and intellectual reasons-why-to-buy. The idea is to drive customers into profitable and habit-forming behavior, away from "lowest-price-seek". The tools at our disposal are mainly communication and flexibility in program design. So we are interested Who we say or do What to, When we say or do it and How best we say or do it to help push the customer relationship forward. Why, though the most interesting of all, all too oftentends to be somewhat academic!

The tools we use are myriad -- varying from the statistical to the manner of marking the database (composites, pictures-in-time, etc.). But again, that is a topic for another article, here, let's just say there are primarily four stages of analytical sophistication in loyalty: Stage 1, seat of the pants, flying blind (folks in the organization often say we know how things work, we have been doing this for 30 years now), Stage 2, which is a little better, what we call the rear-view mirror school (the organization does have MIS systems, and perhaps OLAP type cross-tabs, here's what happened last month and hey, we can see this across various attributes), Stage 3, we are getting to be good now, analytics (where the company actually has a bunch of statisticians and is planning or has invested in AI tools), and last, Stage 4, embedded (where folks don't get out of bed to address a problem without getting the analytics run first, it is systemically embedded into the very culture of the organization). It is nearly impossible to get from Stage 1 to 4 without planning the journey carefully. But it is easy to tell after just a few minutes observation at which stage an organization is in. And whether or not it plans to move ahead.