by Byron Sharp
Given that Reichheld and Sasser's (1990) article was about profitability we might expect that they would have included some assumptions about costs, and they did – they assumed that the halving of defection is achieved at zero cost!
And finally they assumed that halving customer defection in the real world is perfectly possible. Indeed, why stop at half? Their article was titled “Zero defections ...” Can companies radically alter their rate of customer defection? Is it possible to reduce defections to zero, or even to halve defections? Empirical evidence shows that this is wishful fantasy.
Customer defection rates follow double jeopardy
No doubt you have heard the old maxim that it costs five times as much to win a new customer as it does to stop one leaving. There is no empirical support for this idea. In reality, permanently reducing defection rates is difficult and expensive, because defection rates (another loyalty measure) also follow the double jeopardy law.
This means that a brand's defection rate is essentially a function of its market share, and the category it's in. This defection level does not vary substantially between competing brands.
How many customers a brand loses in a year depends on how many it has to lose in the first place. Obviously a brand can't lose a million customers if it doesn't have a million customers to start with. Larger brands therefore can, and do, lose more customers each year, and they also gain more customers. But as a proportion of their customer base they lose (and gain) less than smaller brands, i.e. their defection percentage is lower.
Imagine a two-brand market. One brand is smaller with 20% market share (a customer base of 200 customers); the other larger brand has 80% market share (a customer base of 800 customers). If the two brands maintain their respective share of customers, then each brand's defections must equal its acquisitions. Imagine the large brand loses (and gains) 100 customers each year. Then, in this two-brand market, the small brand must also lose and gain 100 customers (see Figure 3.1). The small brand's defection level is 50% (100 divided by 200) while the larger brand's defection level is only 12.5% (100 divided by 800).
Figure 3.1: The smaller the brand the higher the defection (and acquisition) metric
Source: Sharp, Riebe, Dawes & Danenberg 2002 (Illustration by Dr Nick Danenberg).
In real markets there are more than two brands so things are more complicated. But the fundamental pattern, the double jeopardy law, still holds with brands with a larger market share having slightly lower defection levels (i.e. higher loyalty).
The empirical reality of the double jeopardy law is a fatal blow to Reichheld and Sasser's (1990) idea of easily and cheaply halving customer defection. Double jeopardy shows that it isn't possible to radically alter defection rates without massively shifting market share.
Look at Table 3.1, which shows defection rates for car brands in the US. The data comes from a survey of 10,000 new car buyers in the US in 1989-91. The survey recorded what car brand they bought and what brand they owned previously (if any). Defection levels for cars are much higher than in most service industries, though still surprisingly low considering the dozens of other new car brands that each buyer could have bought rather than staying loyal. Each major car brand in the US suffers a defection rate of about 60-70%, or two-thirds of their customers. No brand has managed to obtain a defection rate vastly different from this average, and certainly not without having larger market share.
Table 3.1: Customer defection rates for car brands in the US, 1989-91
US automobiles Market penetration (%) Buyer defection (%)
Pontiac 9 58
Dodge 8 58
Chevolet 8 71
Buick 7 59
Ford 6 71
Toyota 6 70
Oldsmobile 5 66
Mercury 5 72
Honda 4 71
Average defection
67%
Source: Bennett, 2005.
Table 3.2 is a similar table, showing defection rates for car brands in the UK. The data comes from a survey of 25,000 new car buyers in the UK and France in 1986-89. The survey recorded what car brand they bought and what brand they owned previously (if any). The UK and French markets are less fragmented than that in the US and the market shares of the reported brands are higher, and so, in line with the double jeopardy law, defection levels are lower.
Each brand suffers a defection rate of about 50% (i.e. a retention rate of also 50%).The double jeopardy pattern is again noticeable – the smaller brands have somewhat lower loyalty (higher defection rates). Ford, being the largest brand, has the lowest defection rate. But its defection level still isn't that much lower than its rivals.
Table 3.2: Defection rates for car brands in the UK and France, 1986-89
UK automobiles Penetration (%) Buyer defection (%)
Ford 27 31
Rover 16 46
GM 14 42
Nissan 6 45
VW/Audi 5 46
Peugeot 5 57
Renault 4 52
Fiat 3 50
Citroen 2 48
Toyota 2 50
Honda 1 53
Average defection
49%
Source: Data kindly provided by Renault France, and fully described in Colombo, Ehrenberg & Sabavala 2000.
Remember that Reichheld and Sasser assumed a cost-free halving of the defection level. It's difficult to see how it could be dirt cheap to do something that no other brand in the market has been able to do. Note that not a single one of these brands has a defection rate around 25%, not even Ford, which has double the market share of its nearest competitor. Asking a brand like Honda to halve its defection rate is equivalent to asking it to increase its customer base more than thirtyfold! Doing things that no other brand has been able to do, in spite of considerable investments in customer relationship management (CRM) and other customer satisfaction initiatives, is seldom cheap or easy.
Amazing maths
The implications of the double jeopardy law for growth potential are profound, and can be shown with simple maths. Consider Tables 3.1 and 3.2, each year a car brand gains about half its sales from new customers and about half from returning customers. If a brand like Toyota were to reduce its defection rate to zero then it would gain 50% more sales, which is one percentage point of market share. This one point of share is the maximum it can gain from improving retention. But each year about half of new car buyers switch brands, so each year 50 points of market share are up for grabs. This is the most Toyota can gain from improving acquisition, that is, 50 times the sales potential that retention offers!
Most service industries have defection levels far lower than these examples. Figures of 3-5% are quite normal, so even if a company could miraculously reduce defection to zero it would give them a sales gain of only a few percentage points. In market after market the potential gains from acquisition dwarf the potential gains from reducing defection.
Even growing brands lose customers
Halving defections is neither easy nor cheap, and permanently reducing customer defection to zero is fantasy. Also, the growth potential from customer acquisition is much higher.
The double jeopardy law describes normal markets where brands typically aren't growing or declining substantially. What about dynamics? What role does defection play in delivering growth? Does defection reduce dramatically while a brand is growing? Does acquisition rise dramatically? Or, does growth come from a combination of both? Is growth dependent on marketing strategy? It turns out that the answer to these questions is very simple: growth is due to extraordinary acquisition. Contraction is due to dismal acquisition.
Erica Riebe (2003) examined the dynamics of customer base growth and contraction in her doctoral research. She examined both growing and declining pharmaceutical brands over 10 years.18 For each brand she calculated the amount of acquisition and defection it should have had given the market norm and the size of the brand (i.e. if a brand's customer base was stable then its acquisition and defection levels should
be equal, and will depend on the size of the brand). She expected that growing brands might show excess acquisition and lower-than-expected defection. That is, together these two factors would contribute to the growth in the customer base. Surprisingly, she found that growth was almost entirely due to particularly high acquisition. The declining brands showed a similar pattern but in reverse: their defection rates were healthy and matched stable brands of similar market share, but their acquisition rates were poor.
Riebe (2003) replicated this research in France (looking at shampoo and chocolate bars). Using twelve months of panel data she compared each panel member's favourite brand in the first six months with his or her favourite in the second six months. This allowed Riebe to calculate, for each brand, how many 'first brand loyals' it gained and lost. The shampoo category turned out to be too stable, with the customer bases of the brands not changing. But in the chocolate category, where there was some movement, the same pattern was seen as in pharmaceuticals – customer base growth was mostly due to excelling in acquisition.
This research was later extended with further analysis and Nielsen provided four-and-a-half years of banking data. Again, it was good customer acquisition that led to growth, and poor acquisition that caused decline (Riebe et al. 2014).
Is defection largely outside of marketer control?
A simple explanation for the pattern described above is that customer defection is largely outside of marketer control – at least in terms of being able to alter it with customer service and other such initiatives. There is some very good evidence to support this. Consider the following data (Table 3.3) on annual defection rates from financial institutions in Australia. This data shows a very typical double jeopardy pattern: loyalty declines with market share. In comparison to the huge variation in market shares (a thirtyfold difference between Adelaide Bank and CBA), defection rates vary little.
Table 3.3: Defection rates (Australian financial institutions)
Bank Market share (%) Defection (%)
CBA 32 3
Westpac 13 4
NAB 11 5
ANZ 10 4
STG 6 4
BankSA 1 5
Adelaide Bank 1 7
Source: Roy Morgan Research.
The smallest brand, Adelaide Bank, is a regional bank with branches essentially only in Adelaide. CBA, the largest bank by a considerable margin, is thoroughly national with branches in every main city and regional centre. If someone moves from Adelaide to Sydney (about 20% of Australians move house each year) and if they banked with Adelaide Bank they now find themselves a very long way from the nearest Adelaide Bank branch. Hence, they are likely to switch to another bank, one with branches convenient to their new Sydney home. But if they had previously banked with CBA in Adelaide, odds are that they will have Sydney branches that are just as convenient as ever.
These differences in physical distribution seem to almost entirely explain the double jeopardy pattern in bank defection rates. So Adelaide Bank's comparatively high defection level (double CBA's) is probably nothing to do with differences in customer satisfaction, brand equity, nor any indicator of CBA having a superior retention program. It's simply that Adelaide Bank is smaller than CBA and has fewer branches, so it must have more defection. Therefore, Adelaide Bank should not worry about its higher defection rate; there is practically nothing that can be done about it unless Adelaide Bank dramatically increases its market share19.
Research examining the reasons for customer defection (Bogomolova & Romaniuk 2005, 2009; Lees, Garland & Wright 2007; Bogomolova 2010, East et al 2012) has shown that much defection occurs for reasons entirely beyond the firm's control (e.g. moving house, no longer needing the service, being directed by head office, etc.). Then there is the fact that any brand faces a great deal of competition and competitors are constantly trying to lure customers away. No matter how much effort and expense you put into looking after your customers, every now and then a competitor has to get lucky.
Acquisition is not optional
It seems hardly revolutionary that customer acquisition should be essential for growth, yet this fact is often get forgotten with today's emphasis on targeting, database marketing, social community management, CRM and loyalty programs.20 The real-world evidence is very clear: it is essential to acquire customers even to just maintain your brand. But what sort of customers should you seek to acquire? All buyers aren't equal, so who to target – among both existing and potential buyers?
The next chapter reveals a law concerning heavy and light repeat buying. These regularities in consumer behaviour underpin the stark patterns concerning the brand metrics we've seen in previous chapters.
Chapter 4:
Which Customers Matter Most?
Byron Sharp
The difficulty lies, not in the new ideas,
but in escaping the old ones.
John Maynard Keynes (quoted in Drexler, 1987)
Smart marketers know they need to reach all buyers (i.e. buyers of the category, and everyone from light to heavy buyers of a brand) in order to reinforce buying propensities, and to win new sales.
The death of mass marketing?
Professor Phil Kotler and colleagues (1998), amongst others, have declared mass marketing to be old fashioned. 'Modern' marketing is said to be about positioning; targeting segments; focusing on loyal, heavier buyers; focusing on retention (not acquisition) and return on investment (ROI). Oddly, textbooks have been preaching this 'new' message for decades. Today's fashionable media strategy uses new, diverse, low-reach media to deliver new 'consumer engagement' (see Nelson-Field et al 2012). Broad-reaching television channels and newspapers – while still very much used by marketers (to great effect) – are no longer de rigueur. Today the fashion is for loyalty programs, websites for loyalists (e.g. Budweiser's (ill-fated) 'Bud.tv'), targeting 'influencers', customer relationship management (CRM) and 'new media'.
Yet buyers are busier than ever, and many brands are vying for their attention and custom. Forming deep relationships with a substantial number of buyers seems more unlikely than ever. Consequently, it is logical to expect marketers to strive to become better at mass marketing, rather than abandoning it. This is supported by researchers who have been studying patterns in buyer behaviour and brand performance; they have concluded that mass marketing is essential for both brand maintenance and growth.
For decades marketing scientists have been studying buying rates and the statistical distribution of these rates. Buying rates show how many people buy a brand once a year, how many buy it twice, three times and so on. These patterns in these buying rates are rarely taught at business schools but they are used everyday in businesses. These law-like patterns underpin successful sales-forecasting techniques that are used by leaders in this field like BASES, as well as being integrated into some media scheduling models.
Today we know a lot about how often people buy, and how much variation there is between people's buying. We know that buyers differ from one another in terms of how often they buy and the different brands they buy. We also know how many of each type (weight) of buyer there are. These data provide benchmarks, allow for very useful predictions and deliver great insights. This is a tremendous scientific achievement, and something all marketers should be familiar with.
Light buyers matter
All brands have many lighter buyers. While these people are only occasional buyers of a brand, there are so many of them that they collectively contribute substantially to sales volume.
It is surprising just how light the typical buyer is. Let's look at a super big brand in a frequently bought category: Coca-Cola (see Figure 4.1). In many markets, Coke is the market leader by far in the cola category. Data from the Kantar Impulse Panel in the UK (this panel tracks personal purchases, for example, a person buying a can of Coke to drink themself) shows that the average Coca-Cola buyer purchases almost 12 times a year, around once a month. But this average is very misleading, because purchase frequencies show
a skewed distribution. The average buyer is not typical. There are some people who buy Coca-Cola morning, noon and night – that's more than 1000 purchases every year. So for Coca-Cola to have an average buying rate of just 12, each of these super heavy buyers must be balanced by many hundreds of Coke buyers who buy only a few times a year.
The Kantar Impulse Panel data shows that a typical Coca-Cola buyer purchases (for him- or herself) just one or two cans or bottles a year. That's half of all Coke buyers. Only a tiny fraction of Coke buyers purchase around the average amount of 12 purchases. Therefore, the average buyer is certainly not typical.
Around 30% of cola buyers don't even buy themselves a single Coca-Cola each year. Note that these are people who do buy cola (not just soft drink) – so most of these people do indeed buy Coca-Cola, just very occasionally and not every year!
Therefore, from Coca-Cola's perspective, a heavy buyer is anyone who buys three or more cans or bottles of Coke a year. Many of us who thought we hardly every bought Coke now turn out to be quite normal Coke buyers. Very light buyers dominate, even for Coca-Cola – which is a very large brand indeed.
Figure 4.1: Distribution of buying rates of UK cola buyers purchasing Regular Coca-Cola
Source: Kantar Worldpanel, UK 2005.
What do the buying frequencies look like for smaller brands? They are surprisingly similar. Let's look at Pepsi, which is a considerably smaller brand in this market (see Figure 4.2).
The average Pepsi buyer in this market buys Pepsi only nine times a year, which is somewhat lighter than the average Coca-Cola buyer (c.f. 12). There are many more cola buyers who don't buy Pepsi in a year – more than 50%, compared to the 30% who didn't buy Coke that year. This is because Pepsi is a smaller brand. But again, a heavy buyer for Pepsi is anyone who buys more than three times a year.