Jim answers questions from fellow Drillers
(More questions with answers here, Work Overview here, Index of concepts here)
Topic Overview
Hi again folks, Jim Novo here.
So Jim, this customer behavior profiling / prediction is great for consumer businesses, but what happens if you’re running a long sales cycle B2B biz where buying decisions take months if not years, and may involve a dozen decision makers? Well fellow Drillers, the answer is not as complicated as you might think – it’s about where to look for the predictive behavior outside of the sale transaction. Interested? Let’s get to the Drillin’ …
Q: I read your section about how “R” and “F” are better indicators than “M” which I agree. But for the problem I face, do you have any ideas on how I can redefine “F” for my purpose? If not, I can always use RM, but will face the drawbacks you mentioned in the book which I think are legitimate concerns for predicting potential value.
(Jim’s note: this Driller is referring to the modified RFM model used in the Drilling Down book. For an overview of what he is talking about see this description of what is in the book and this outline of RFM.)
A: Just to ground this discussion, I assume you are talking about Company XXX …
(a major enterprise software company with many products. He said Yes)
You should look for R and F in other places, if “short term” prediction is what you are after (I’ll discuss long term in a minute). Long cycle businesses like enterprise software can be more difficult to model because the variables you are looking to do an RF scoring on are not as obvious. The sales activity may not be particularly predictive of customer behavior because the nature of the business precludes frequency of purchase.
For example, think customer service. Where in your organization would you see RF show up relative to customer satisfaction? Perhaps at the call center, help desk, or “outstanding issue” logs of the implementation team? There could certainly be other areas, depending on how customer care is set up. The question is: how does the Recency and Frequency of customer care predict the likelihood of customer defection?
Despite the fact you sell a “product,” one could imagine you are really in the service business. This type of product sets up (hopefully) a very long Customer LifeCycle and ongoing service relationship with upgrades, add-ons, customization, and so forth. Perhaps most of the profit is really in the ongoing relationship, not the initial sale. If true, this is where the focus on RF profiling should be.
You want to go where the transactional behavior is, because this transactional behavior is predictive. So you have to find out where it is and run your profiling there. For example, once the installation is over (is it ever over?), what is the Recency and Frequency of calls for assistance? Does the RF of “trouble calls” predict the likelihood of additional sales in the future, or is it a negative predictor – the higher the score, the less likely a customer is to upgrade? Many times in a service business, high RF scores indicate negative satisfaction, as you probably can imagine.
Somewhere in the organization there is transactional data predictive of likelihood to buy additional services / likelihood to defect. Your mission (should you choose to accept it) is to figure out where it is, or if it does not exist, create a way to capture it.
Now long term. Over very long Customer LifeCycles, one simply has to extend the time horizon. Remember, RF is a relative, not absolute, scoring system, which is why it is useful across such a broad range of businesses. It compares and ranks activity between customers, not against an external benchmark. So even though “frequency” may be every 5 or 10 years, it is still predictive relative to other customers.
For example (and I certainly don’t know your business, so I am making this up as I go) say there is a “base” package, an ERP Accounting / Planning / Forecasting module. It’s the product you are well known for and has high customer satisfaction; the product most companies buy first when they engage in a relationship with you.
Let’s say satisfied, best customers tend to add on to this base module as the years go by. They add Human Resources, Warehouse Control, CRM, e-business marketplaces,. etc. This may happen every 3 -5 years. But some customers do it more quickly the others, and this is where you see high RF scores, as compared with others who do it more slowly. So you still get an RF ranking, and you still get predictive power in the model, even though the transactions are spread out over decades. Your challenge may simply be this – you don’t have data that goes back over decades.
What you want to know is this: once you have identified high scoring customers, what is it about them that is similar? Is it who made the initial sale, the type of business they are in, geography? If you compare high scoring and low scoring customers, what are the differences? What kind of business adds on to the base module every 2 years as opposed to the kind of business that adds on every 5?
Plus, can you use this knowledge to predict defection, or in your case, a low likelihood of further upgrades? If the top 20% best (most profitable) customer businesses make their first add-on by year 3 after the initial install of the base module, what does it mean when a business passes by year 3 and does not add on? Is this a red flag? Should you send in a “specialist” to find out why the add-on has not happened? Are they experiencing problems with the base module which were never documented, or worse, never fixed? Setting up this kind of “early warning system” can be very helpful in a customer retention effort – the behavior of the customer is telling you, flashing a signal, that something is wrong relative to other customers.
I hope the above answers your question. Long cycle B2B is not as simple to profile as B2C, but the behavior is still there. You just have to look a little harder for it. Here’s some additional resources on my site.
The first goes deeper into behavior profiling for service-oriented businesses.
The second reviews Latency, first cousin to Recency and another “early warning system” metric which for some organizations is easier than Recency to “sell” internally and implement. The “didn’t add-on by year 3” example above is a form of Latency tracking.
Good luck with it! Let me know if you have further questions.
Get the book at Booklocker.com
Find Out Specifically What is in the Book
Learn Customer Marketing Concepts and Metrics (site article list)
Download the first 9 chapters of the Drilling Down book: PDF