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.
Where does all this kind of thinking on customer retention and value over time eventually lead you? Well, often right to Finance. Because you see, the more you can do to convince Finance the activities you are in engaging in are increasing profits for the company – and Finance truly believes this because they participated in the creation of the systems, models, and validation – the more likely your budgets supporting these activities will increase.
Simple as that. You’re not afraid of Finance, are you? Good! On to the Drillin’ …
Q: Been reading through your site a bit. I run the CRM and online marketing at (large airline) – a business with roughly XX million customers and X.X million members in our loyalty program. Interested in your thoughts about RFM algorithms as well as aggregated scoring. My predecessors set up ranked scoring along spend – essentially taking paid purchases and ranking people from high to low in R, F, and M and then built programs around this.
My issue with this approach is that we find very different behavior in our top 20%, 10%, 5%, and even 1% (e.g standard deviation of population is large). Additionally, rank ordering often grouped individuals with the same underlying behavior in different categories because of the arbitrary nature of where the snap lines fell. So I altered our scoring as follows…(long description of new model)
A: Did you by chance see this article?
Latency may be a better way to go for an overall approach to airline behavior in the business class; Recency in the tourism class.
It sounds to me what you have done is a similar idea – recognized the generic RFM model is broken for your needs, extracted the essence of the RFM idea, and rebuilt it into a model that works for you. Nice job!
Q. But,if someone spends $400 on a flight that is 400 miles vs. 1000, the revenue has differing implications – both in terms of customer and non-customer driven fixed and variable costs. If someone spends $400 on a flight that sells out – we are potentially spilling revenue (not holding inventory for a bigger spender) – and thus the opportunity value is greater than the collected revenue. But if the flight doesn’t sell out…this may not be true?
A: For sure, it’s a complicated business. I think you need to ask yourself if it is the responsibility of your model to deal with this complexity. If the model is a “targeting” model, then targeting is what it can do. It can’t possibly deal with a bunch of semi- random margin scenarios. That you would probably push back into the Financial Analysis of program / marketing outcome – after the targeting model did that job.
Put another way, RFM and the many variants are fundamentally about response rate. The “M” is a predictor variable, and a weak one at that; it also tries to align the cost of marketing with customer value potential. The financial implications of the response generated by the RFM model is a different issue, one that would be best dealt with in an offline analysis, outside the response model itself.
Q: To even further complicate matters – we can sell product on other airline partners and vice versa. Should we be looking at profit vs. revenue – even if some of the costs were not driven by customers?
A: Sure, you can look at profit instead of sales if what you want is to target higher profitability customers for response. But again, RFM is not a financial ledger system, it is not meant to make financial decisions, only targeting decisions.
Q: Customers can redeem miles to fly rather than pay. How should this behavior be integrated? The value we derive from redemption is often the switching cost we create with customers to ensure high share of wallet. What about scoring how customers earn miles – by flying on us or a partner, by earning bonuses, by using bank affinity credit card. Then there is other shopping behavior issues – the buyer may not be the ultimate user. What if they choose a sister brand? What if they are someone who flies little but influences spend of others or other revenue (corporate travel manager or C-level exec or purchases cargo revenue).
A: Yikes! OK, let’s talk about “scoring”. The kind of scoring you really want to do here I don’t think has much to do with RFM at all. What you want to build is a customer profitability model, and I’m not sure that has anything to do with response at all. In fact, in many cases, the highest responding groups – usually best customers – often do so at a negative profit. This happens because you frequently eat into margin dollars that would have occurred anyway: I spend $1000 a month, you give me a 10% discount, I spend $900 that month and then $1000 the next month. No incremental behavior.
Q: How to manage issue of purchase channel – with its differing cost and loyalty implications – direct vs, agency, online vs, offline; promotional response rates – only buys with promo or doesn’t change behavior with promo or uses promos for travel they would have likely taken anyway; form of payment – with its different credit risks and merchant fees; purchases of main product vs. ancillary (hotel, car, cruise, vacations, gift cards – some of which we are a supplier on – and some that we are not); market contestability – if they are flying in a market where either there are lots of equally sized competitors – lots of competitors but of unequal strength and share – or only a few competitors; relative market strength – are we a dominant competitor, in the running for significant share, or an also ran. Is it a new or mature market? Do we have a better or lower positioned product? And what of data on the experience (late or cancelled flights, long lines, call center hold times, lost or late luggage) and satisfaction of customers?
A: OMG !! Not just a customer model, but a financial model of the entire business, with all the moving parts! Jeepers, are you the Marketing guy or the CFO?
Q: Would be interested in your thoughts on how to approach these issues – to eventually end up with a single or small set of KPIs to use for customers in marketing applications.
A: Here’s what I think. That’s one heck of a complicated business and you really need to sort out what you can have an impact on. There are certain parts of the business that you really have no control over, and so the question is, if I can’t do anything about it, does it belong in a marketing model?
Seems most of your questions really have to do with net profit, which clearly is a marketing issue, but in a different way. More similar to a fundamental ROI question: if I sell 1000 miles, how much money did I actually make? Then if you know the answer to this question, you can go back to the financial analysis and plug that number in.
If I’m understanding the issue correctly, here is what I would do. If your need is short-term, create a proxy variable like “Net Income per Mile” or something like that right from the balance sheet of the company. This gets you pretty close to where you want to be as a generic starting point to build out from.
If you have a little bit of time, I would simply start having this very same conversation with the CFO – what number, in the CFO’s mind, best represents the profit “flow-through” from an incremental mile generated by marketing?
There are a heck of a lot of lines on an income statement and any one of them might not be appropriate for this particular situation. What does the CFO want to generate? Cash flow? Earnings? EBITDA? That’s what you really need to know. From there, you can start to look at income scenarios – start with the largest / highest volume ones first – where you perceive the income metric to be higher or lower than the baseline “Cash Flow per Mile” or whatever provided by the CFO – sales on other carriers, source of sales, market conditions.
We did this when I was at HSN. There was a desire to get to the net productivity of marketing efforts, and we worked closely with the CFO to develop the model. What was the true profitability of an incremental sale generated by Marketing?
This model was loaded with every variable to order cost you could imagine – cost of the order phone call, cost to pick, pack and ship an order, cost of customer service calls after the order, cost to process the projected return, etc. These costs were deducted from any net gross margin (after returns) generated to arrive at “flow through” from a sale, where THEN you deducted the cost of marketing, discounts, etc. Yet we still often generated a 60-day ROMI of 2 to 1 or better – even though the flow-through was about 20 points below product gross margin. Makes me really want to chuckle about incredibly weak online marketing metrics like ROAS – what a joke.
It was tough to generate profits, but the flip side was every time we told the CFO we generated a profit, he knew precisely what we meant. That’s when Finance started really throwing money at us, because they had such a high degree of confidence in the results we were producing for the company.
If you’re going to go as deep as you are considering, you first need a financial model of the business. Then you can build a customer scoring application. And then you can build a response model that will do what you want it to do. Nice challenge for a Marketing guy, plenty of work to do. But I would really start with the CFO – or whoever in the Finance chain is appropriate. You just might find they have a lot of these models already and that will
save you a ton of time!
Besides, rooting around in the business model like this will probably raise some questions that don’t have answers – hey Finance, why do we do this? Now you’re really into the guts of it and are participating at the Strategic level – where Marketers ought to be, in my humble opinion.
Jim
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