CU Employee Community Chair

Member lifetime value modeling

In his influential book, Customer Centricity, Peter Fader urges marketing analysts to identify and understand the traits of a company’s most valuable customers – and argues the merits of using that knowledge to guide the company's most rational use of resources and its attraction of high-value future customers.

His customer valuation model doesn’t limit itself to customers’ profitability at a particular point in time – but rather seeks to estimate (probabilistically) a customer lifetime value, acknowledging that some customers are likelier to remain loyal purchasers over longer periods than others and their purchase behaviors predictably yield different revenue streams over time.

Even though credit unions are cooperatives, these concepts seem relevant. Most of us who lead credit unions seek rationality in our product pricing as a matter of fairness. We look to encourage a level of member engagement with our products that presumably drives increased value to the individual and the collective. And, with occasional exceptions, we look to strengthen the franchise as we increase the size of our memberships, attempting to assure that growth delivers more shared benefits than burdens. A customer-centric approach seems likely to deliver insights to advance all of these.

Nonetheless, my own progress has been complicated by five practical challenges that arise as I start to compute member lifetime value in a credit union context:

  1. With financial institutions, revenues and costs accrue very differently than they do for the retail store scenario that Fader engages. I'm concerned that RFM analysis doesn’t quite fit.
  2. With many credit union products, the greatest source of profitability arises from members who make irrational choices that would be wrong for credit unions to optimize around.
  3. With other credit union products, the greatest source of unprofitability arises from our own irrational product pricing.
  4. Credit unions’ broadest source of member value/profitability is very often centralized in just one or two product categories (or so I'm finding).
  5. The data with which we might derive insights about our members’ future value is very uneven in its availability and that may bias our optimizations

Beyond those, there may also be opportunities to marginally improve on Fader's model, including recognition that many members' product usage morphs more or less reliably through different life stages.

I’ll try to expand on each of these issues in future posts to this thread, and perhaps together (in dialog) we can noodle our way toward a model that addresses them, even if it requires some greater sophistication than the simple foundational one that Fader proposes in his book. Many of you heard Nate Derby engage customer centricity as a topic at the 2018 CU Analytics Summit in Redmond. With luck, we'll coax him into contributing to this discussion as it marches along. I hope that many others will freely voice their thoughts, as well.

Dale Davaz
STCU R&D Strategist
CULytics Community Chair

You need to be a member of Credit Union Big Data Analytics Community and Summit - CULytics to add comments!

Join Credit Union Big Data Analytics Community and Summit - CULytics

Email me when people reply –

Replies

  • CU Employee Community Chair

    Hello folks... time for another addition to this thread, harkening back to issue #4 of my original post.

    Set aside the woefully irrationally priced products I talked about last time, and how they uniquely complicate measurement of member lifetime value. Turning our focus to the credit union products that are left, what if almost all are priced to essentially “break even” and only one or two are priced for profitability? This, too, undermines much of the strategic promise that I think Fader would claim for the lifetime value concept, certainly when compared to situations where that’s not the case.

    One can quibble at length about the data and the methodology, but when (for pure expediency) I pull household retail product profitability data from our MCIF system and look at how it’s distributed across our membership, a VERY interesting picture gets painted:

    Egads. With one or two notable product exceptions — we’re losing money with over 80% (looks like a couple standard deviations, anyway) of our member households in every other product category. We’re not losing much at the peak — it rounds off to just $100 per household per product per year. But there it is. Gulp.

    The undisputed money maker? Mortgages. Set aside the funky spike at $2600 that smells a lot like an artificial product profit cap that the MCIF has enforced. In that one category, there’s very nice positive profitability distributed along a nice range of members.

    Digging a tad deeper, a case can be made that home equity shows some promise, also, despite the gnarly spike at -$600, which no doubt represents idle, zero-balance home equity lines of credit that have cost us plenty to set up, but haven’t netted us much (or any) income quite yet.

    The problem? Whatever strategic insight we’d hope to derive from a member lifetime value exercise devolves to this — we’re essentially a mortgage origination and servicing company that just happens to provide other financial services. Fader might rightly advise us that the purpose in having all those non-mortgage households is just to give us easier access to sell them mortgages. The whole credit union experience falls into three phases — before you have a mortgage, while you have a mortgage, and once you’ve paid it off. Stop sweating everything else — it's all about the home loans. The perspective is only a wee bit more interesting when home equity is the alternate source of profit.

    I shouldn't get hyperbolic. It’s true — we can see some of the upper tails of several products situated in the +$100, +$200, +$300 profitability areas. There are likely traits in those households that should be studied and encouraged. But when one product dominates the story as mortgages do in this one, it tilts our priorities significantly.

    And, true, it’s a very first-blush assessment. There are lots of refinements that could adjust the conclusions, perhaps significantly. Maybe the graph above looks very different for your credit union. But if we're NOT unusual and some of this effect is a feature and not a bug, thanks to how cooperative principles have played out in a thousand ALCO pricing decisions, this consideration certainly raises an important cautionary signal in our early going on the member-lifetime-value path.

  • CU Employee Community Chair

    Hi everyone,

    I noted in my previous post (a reply to Justin) that one important caveat we want to consider in building a model of member lifetime value in a credit union context comes from understanding quite common — and unfortunately irrational — sources of household profitability. Members with products ill-suited to their needs and members with poor money management habits tend to top our lists of most highly profitable folks. Yet very few credit unions consider it wise (or ethical) to optimize their operations around those segments.

    The reciprocal concern that I expect is nearly as diabolical is how so few of our most UNprofitable households are that way because of their relationship depth or channel usage behaviors. Our most unprofitable households are those who took the credit union up on promotional (often, questionably “loss-leader”) pricing that tanked their otherwise profitable relationships.

    Technicalities come into play — especially how models apportion profitability between deposits and loans, typically using the rates of matching-term treasuries to assess how much of the difference between interest income and interest expense go to borrowers versus savers. Both deserve some of the credit of those earnings. But when we set a certificate rate special a little too high, we’re dooming all but the most profitable folks who take us up on the offer to the lowest circle of the profitability “inferno.”

    Is it fair? Not really. But even more to the point, in member lifetime value thinking, is it wise to say we want to steer the credit union altogether away from households who would take us up on our own (perhaps foolishly-priced) certificate specials? It’s even easier to say to that, certainly not.

    In what other product areas are credit unions vulnerable to cheating rates in directions that unfairly skew our members’ lifetime value calculations? Who among us hasn’t tended to get a little extra aggressive with indirect auto pricing? Who hasn’t felt the market tugging down on our home equity line of credit rates? Add those (and others) to the deposit specials we offer, and you can see how quickly run-of-the-mill, far-from-cherry-picking members can find themselves in red ink from a marketer’s perspective. But it’s far more a result of OUR doing than these members’.

    Can corrective allowances be made? I expect so. But I worry that credit unions who neglect to consider biases introduced by these all-too-frequent pricing outliers do so at their analytics peril.

    — Dale

  • CU Employee

    Dale,

    Great topic.  I am interested to see where the conversation goes.  Below are a couple of my initial thoughts.

    • Here is a recent artile that argues it is time to move away from RFM models.  If nothing else, it has some interesting points.  For your purposes, your RFM model might still provide better insights as is, or with just a little tweaking, compared to excluding the model all toghether.
    • Have you defined "profitable"? Maybe members that meet a monthly card transaction minimum, have loans that produce interest revenue of a certain amount, generate regular fee income with limited risk / cost, have stable low cost deposits allowing the CU to make additional loans, a combination of these, or something else?
    • I like the statement that "...the greatest source of profitability arises from members who make irrational choices...".  I assume these are irrational choices from a data point of view.  Can the rationality be explained using additional datapoints; life stage, recent loan closure, recent loan application, cyclical buying behavior, etc.?

    Justin

     

    Segmentation Models Are Outdated: How to Update Your Marketing Segmentation Practices
    For decades, marketers have segmented audiences based on demographics, but those methods are growing irrelevant. There are better predictors of buyin…
    • CU Employee Community Chair

      Thanks, Justin, for your additional thoughts.

      In the context of member lifetime value, I’m thinking of profitability in a broad — but financial — way. It’s about net revenue to the credit union (with all the considerations you suggest), to the extent we can reasonably attribute income and costs to individuals — or, better, households. Sure, some members bring intangible value through their bigger or smaller social networks, brand ambassadorship, etc. But in this analysis, I tend to discount those things as secondary, and focus on the financial fundamentals.

      So yeah — profitability arises from interest income to the credit union from loans, fee income from interchange and overdrafts and whatever, and give deposits some transfer price insomuch as they’re necessary for funding loans of roughly corresponding terms. Still, with the latter, the lower the cost of funds, the better. If I can associate costs to channel engagement, that’s an improvement to the model.

      When I note that many of our most profitable members are irrational, I’m really talking about the limited number of overdrafters who make up the lion’s share of that huge fee income category — at least a fairly substantial line item on our financials. For these folks, overdrafting isn’t an occasional “oops” — it’s a lifestyle. Set aside all the ethical questions surrounding the service of these types of members — one can’t dispute that they’re among our most profitable folks. But it’s probably not right that we want to set as a goal finding thousands of more members just like them, irrespective of Peter Fader’s dictum to build the business around our most profitable segments.

      That’s not the only case. I daresay STCU isn’t the only CU where our most profitable credit card members are folks who, for whatever completely mysterious reason, have situated themselves in the wrong product. They end up paying irrational levels of fees for benefits they ultimately don’t use — and don’t switch to products that would help them optimize their finances. They persist in product choices don’t match their needs — and while they’re profitable to us, it’s not right that credit unions would follow the example of for-profit outfits (cough, Capital One, cough) who are said to encourage folks into misfit products just to maximize revenue.

      The third, perhaps most common case, is the number of folks with idle funds in checking accounts that could be earning sometimes substantial dividends in dividend-bearing savings accounts with some of that cash on deposit — but there the balances sit, in zero-dividend free checking.

      All of these irrational behaviors have non-trivial impacts on our real-world experience of member profitability and (I worry) are apt to seriously skew our attempts to estimate lifetime values if we don’t make important adjustments in our models. They’re one of several issues that I suspect are just not as salient in other industries, like consumer goods.

      As I noted, it’s just one special credit union wrinkle (#2 in the original post) among several. I’ll see if I can’t add a little extra exposition around #3 — irrational pricing — next time I chime in.

      — Dale

      Segmentation Models Are Outdated: How to Update Your Marketing Segmentation Practices
      For decades, marketers have segmented audiences based on demographics, but those methods are growing irrelevant. There are better predictors of buyin…
This reply was deleted.