It has been 10 years since Dr. V. Kumar exposed the myths of the loyalty movement (Reinartz & Kumar, 2002). Yet, to this day customer loyalty is still sought after as the holy grail of customer success. The age of social media has only served to drown out the voice of reason, since it provides an arguably richer source of behavioral (and attitudinal) data to sift through and analyze. Are we using all of this data effectively or are we just playing with new toys? Let’s see what he was talking about 10 years ago and you can decide if it still makes sense.
The loyalty movement claims a number of things about loyal customers, and we’re going to briefly present an alternative view, based on primary research from V. Kumar. What he found may surprise you. He found that the correlation between loyalty and profitability was weak to moderate in the four companies that were studied. In one study of a corporate service provider, they had setup a costing program in order to track the performance of its newly created loyalty program. After running this loyalty program for 5 years at an annual cost of $2 million, they found that half of the companies that they deemed as loyal (regular purchasers) generated little or no profit.
Since the loyalty movement users 3 basic arguments to support their position, he decided to test these three hypotheses to see how they fared with regard to profitability.
Loyal customers cost less to serve
The basic argument is that up-front acquisition costs can be amortized over a longer life of transactions. The problem with this argument out of the gate is that there is an assumption that these transactions are profitable. Another argument less used is that knowledgeable customers are better able to self-navigate the transaction process and should require less direct support from the company. Unfortunately, what the research found was that these customers could actually be more expensive to serve. Many of these long-standing customers clearly understand their value to the company and exploit it to either get premium service (such as expensive personalized portals) or price discounts.
Another interesting finding was that companies may assume that experienced customers would be happy to move to a less expensive channel to make their purchases, such as a web site. What the research found was that these customers then expected to pay a lower price in exchange for using this channel, which offset any savings the company could generate. Has anything really changed in 10 years? Has Amazon built its business on premium pricing or does it have a completely different business model to support the fact that we expect cheaper prices from them?
Loyal customers are willing to pay more than other customers
Does a loyal customer generate more revenue for the company? There is a belief that there is a correlation between loyalty and the cost of switching. Does a high cost of switching mean you’re loyal? That’s a little baffling. More baffling is the belief that they will pay a higher price because of the cost to switch. If this were true, the companies would charge these customers higher prices. In reality most customers will promise greater frequency of purchase as long as they get a lower price in return. The long-term customers of the corporate service provider in the study paid 5-7% less than new customers, on average. In the other studies, loyal customers either cancelled out higher prices with loyalty card discounts, or paid the exact same rate as new customers.
The first conclusion was that loyal customers are more price sensitive than new customers. Perhaps they are more knowledgeable and in tune with the true value of the product. It could also be that they take offense when a company attempts to profit from their loyalty. With information more widely available, and platforms more convenient (in tech industries), 10 years later the cost of switching is even less. Wouldn’t you agree?
Loyal customers act as word of mouth marketers for your company
When it was initially reviewed, it did not appear that there was a strong correlation between customer longevity and the propensity to market by word-of-mouth. However, there are two measures for loyalty; behavioral and attitudinal. While the latter is subjective and a measure obtained through surveys, it was found that customers who exhibited both attitudinal and behavioral loyalty were approximately 45-50% more likely to be active promoters and 25-30% more likely to be passive promoters of a company.
Loyalty managers need to understand that basing loyalty on behavior alone is a misleading indicator with regard to a customer’s likelihood to be a word-of-mouth marketer for the company. Of course, simply using simplistic attitudinal feedback, such as net promoter score, misses critical pieces as well. Having said that, there’s still more to the story.
Should You Actively Keep All Your Loyal Customers?
Now things start to get interesting as we take our first baby-steps towards a comprehensive and bottom-up view of customer lifetime value (which will be covered in a later post). The question isn’t whether we should discontinue our focus on loyalty, but whether the models being used are too simplistic to be helpful. One of my long-time favorite models has been RFM (especially the Jim Novo variant), but Kumar has thoroughly debunked RFM in general. Using such a model to determine which customers to keep and which to let go can be misleading.
Buying patterns for frequently purchased items are different from those that are bought infrequently, and RFM cannot distinguish between the two since it ignores pace. For instance, if two customers begin purchasing in the same month (using a 12 month buying cycle) and customer 1 purchases frequently beginning in the 2nd month (and following on in the 6th and 8th months) and customer 2 purchases less frequently by making a second purchase in the 8th month, RFM will determine customer 1 to be more loyal. Customer 1 wins in frequency and has a purchase just as recent as customer 2 (the 8th month).
Looking at this a little differently, customer 1 buys on average every 2.3 months whereas customer 2 buys every 7 months. Since by month 12 Customer 1 has fallen off the pace, and Customer 2 is still only 4 months into a 7 month cycle, it’s more likely that Customer 2 will make a purchase in the future. In the real world, many more factors will need to be involved (i.e. segmentation attributes, etc.) but here is a simple example that can be used to determine likelihood to purchase in the future. Dr. Kumar explores event-history modeling to explain the likelihood to purchase statistically. It’s actually a pretty simple, and time tested approach. The formula is
n = the number of purchases made in an entire time period (in the example that is 12 months)
t = the fraction of the period represented by the time between the first purchase and last period
In the example above, Customer 1 will have n = 4 and t = (8 ÷ 12), so 0.66674 = 0.198. On the other hand, Customer 2 will have n = 2 and t = 0.6667 which comes to 0.444. Unlike RFM, where Customer 1 was deemed likelier to purchase again, this approach shows that customer 1 has a 20% chance of buying again and Customer 2 has a 45% chance.
Now that we know how loyal (likely to buy) our customers are based on past behavior (remember, attitudinal only applies to WOM marketing, not likelihood to purchase), it’s time to understand whether our marketing dollars are worth spending on them. The problem with many models, including RFM, is that they look at revenue and not profitability. When we seek to simply grow revenue while treating each customer equally, we are almost certainly not doing so profitably. We are all in the business of being profitable, right?
A mail order company, which was in Dr. Kumar’s study, classified companies into 4 tiers of revenue.
- $50 or less
- $51 to $150
- $151 to $300
- > $300
The problem here is that 29% of their customers purchased small quantities of low margin products; and the cost of serving these customers actually exceeded the revenue they brought in. It was suggested at this point that individual customer profitability needed to be calculated for a typical purchase period. In the case of the example above it would be month, but these periods will differ from industry to industry and is driven by the natural purchase cycle. I’ll let the bean counters figure out how to do that for now.
So now that we have the likelihood to purchase and the per-period profitability, both calculated at the customer level, we can simply multiply the average periodic profit by the probability. If we were to try to understand and compare two customers over the next year, what we need to know is whether one customer should be actively courted over the other. As an example, let’s take two profitable customers. Dr. Kumar uses Adam Incorporated which has an average profit of $5,500 per quarter over the past two years. On the other hand, we have Even Limited, which had a $1,000 per quarter profit over the same period.
Here’s where the fun begins: let’s figure out what the profitability looks like over the next year.
|Quarter||Avg Hist Profit||Probability of Activity||Projected Profit|
|Quarter||Avg Hist Profit||Probability of Activity||Projected Profit|
As we can see, both customers look to be profitable for the company for the next year. The question that we have to ask is how much should the company invest in courting purchase activity for each of these customers? In this example, a visit by the full sales team is estimated to cost $5,000 and a visit by a single sales person costs $2,000. In this scenario, while you would be happy if Eve Limited remained active, it would make little sense to invest in even a single sales person visit because that visit costs more than the profit they are predicted to generate. Adam Incorporated, on the other hand, can easily justify a visit by the entire sales team and still remain very profitable.
It’s pretty clear that much of the business world has ignored findings that are already 10 years old. Dr. Kumar has progressed much further than this study and has a comprehensive model for customer lifetime value, and how it should be used for customer management; what he calls CRM. His research proved to be 30% more reliable than other methods, like RFM. How many of you are even using a predictive model at the customer level? This is another nail in the coffin of the one-size-fits-all way of doing business. It’s time to learn not just what percentage of your customers are loyal, but which ones to pursue and at what level of service.
Recall the service provider from his study. 40% of their profitable customers turned out to not be worth pursing as they were unlikely to purchase anything in the future. Nearly the same percentage of their loyal customers were unprofitable. 30% of all of their customers turned out to be neither profitable nor loyal. As he states, they were chasing customers they should have ignored, and ignoring customers they should have been cultivating.
How does your company determine which customers to invest in, and how much to invest?