CRM

A Nation of Thieves – The Dark Side

In June I spoke at the IDM/Precision Marketing Symposium 2003. I missed the first few speeches, but when I read the slides I was shocked. There was Hamish Pringle, Chairman of the IPA, claiming “the dawn of an ethical society”! He should have known better – still trying to sell his big “Cause–related marketing” (the other CRM). I discussed this at lunch with some senior colleagues, before I spoke. They agreed that there was no evidence either way. I countered by saying that there certainly were many reasons to believe that the nation as a whole was probably becoming less ethical, and being totally hypocritical in asking politicians and companies to be more ethical. Apart from the rising levels of financial crime (perhaps due to better detection, tougher laws but also greater opportunity), I suggested that the rapid decline in respect for authority, which is well researched, would cause a decline in ethics, on the hypothesis that many people (not all) are ethical because they believe it’s prudent – to please the authorities. My strategy for companies – be even more cynical – play the ethical game with the media while being even tougher about dealing only with customers who are likely to be of good value. Don’t drop your guard and let the current obsession for privacy turn into a fraudster’s charter.

It brought back memories. A few years ago I made quite a name for myself by being prepared to talk in public about the dark side of customers. I wrote a paper about strategies for distinguishing between good and bad customers. It was born out of my work with insurance companies, banks and the recently privatised utilities, none of whom can afford to be lax about bad customers. That’s because there’s a big difference between bad customers in, say, a retail business, and bad customers in these other businesses. Why? A bad customer for a retailer can do quite a bit of damage. They can walk out without paying. So the retailer has paid their supplier, but hasn’t got the money from the customer. How much can they walk out with? £100 a visit? Not in groceries unless they’re very strong and have got very large pockets. In clothing – easily, as a certain US celebrity showed last year, and as rich shoppers in the West End show too often. But still, you’ve only lost, say, £70 worth of merchandise. More dangerous if they make a habit of it – hence the insistence of always prosecuting when they are discovered. But bad customers learn quickly – it pays! So a more dangerous retail customer is one who works in collusion with shop staff – and 50% of shop theft is done by staff. It’s called “buy two, scan one”. However, the most dangerous bad retail customer of all is one who uses financial fraud – typically credit card forgery or theft. Here, the customer can make thousands a week, and it’s the retailer who usually picks up the tab. That’s why we’re moving to a smart card and PIN-based system in the UK – long overdue. Of course, retailers are exposed in other ways – for example to frequent complainers, who like to over-complain in the hope of over-recompense. That’s why most retailers have invested in complaints management systems that allow tracking of individuals who complain a lot – very sensible.

So, once you use start playing dirty, financial services is a great sandpit! Of course, it’s not just for financial gain. The “Know your Customer” rules being enforced in banking are at least as much to do with how money, whether ill-gotten or raised by seemingly respectable (e.g. religious) charities, is transferred and laundered, to be deployed in terrorism.

However, in financial services, “routine fraud” by “small people” can be extremely damaging, and it’s the subject of most of the rest of this article. Why now? Well, there is lots of formal and informal evidence that most individuals are happy to commit fraud. Just look at the demise of a certain well-known accountancy partnership, and the fines that were levied on big banks when they knowingly recommended bad stock because they were doing other business with the company whose stock it was. Of course, these cases only showed the enormous damage that can be done (particularly to pension funds) when governance fails. But there was a more sinister lesson. It showed that when there’s a big gain to be had, large numbers of people will join in an informal conspiracy to defraud others. There may have been one or two high profile sinners, but the conspiracy was widespread in these organisations. The same has been revealed time and again in antitrust cases.

So, let’s dig into this a bit more deeply. The first and most interesting point is that the concern with data protection and privacy is correctly forcing companies to think more clearly about when they need data about individuals. One reason for wanting to keep data is it tells you who is a good customer and who is a bad customer. Of course, “good” and “bad” are defined primarily according to the organisation’s objectives, rather than morally, though there is often a correlation between the organisational and moral definitions. You could use instead the terms “desirable” and “undesirable” customers, but this loses the emphasis I like to place on the relationship with the moral definition.

Technology and data sources now make it possible for companies of all sizes (not just large) and government organisations to differentiate much more accurately between “good” and “bad” individuals and groups. These developments also make it much easier for these organisations to predict likely “goodness” or “badness”, using a range of statistical indicators. Where use of these indicators is not permitted for some reason, this initiates a search for surrogates for these indicators. Following this approach successfully requires an organisation to define what it means by “good” or “bad”, and to keep this definition under review according to the performance of individuals.

“Good” and “bad” are relative terms. Their definition changes with time, with laws and with the strategies and target marketing of companies. In addition, we need to distinguish between character and behaviour. A customer who might naturally be “bad” for a company may be constrained to be “good” by product design or service management. Goodness and badness are not necessarily a function of the individual. Wider social and economic forces will at least in part determine goodness and badness, for example the state of the economy, marital breakdown, life-stage, neighbourhood etc. The state of the economy is particularly relevant because it will have a regionally or even sub-regionally differentiated impact on goodness and badness. The resilience of local economies to a national downturn will affect credit defaults (and one could argue other forms of badness as well) in a spatially dis-aggregated manner. These factors are seldom measured or input to models of worth, future value or risk.

In the private sector, good individuals are broadly defined as a mixture of some or all of the characteristics below. However, as we note later, companies can also specialise in dealing with customers who are “not good”. For example, some store cards are targeted at customers who are less creditworthy, possibly because of imprudence, or simply low income. So the list below should be taken more as an example of the spectrum.

This latter point is very important. Several examples have been given of how customers can be good in one domain and bad in another. It is the stability of this pattern, and perhaps more importantly the stability of individuals as members of groups which allows organisations to trade with them profitably. Although in theory all risks can be dealt with by insurance, when it comes to the balance of risk and value, stability may be the key. For example, a High Street retailer setting up a new store may reckon on a particular level of abuse (credit default, shop theft, staff fraud), based on its experience with similar stores, as well as a particular level of trade. As long as each new store displays a similar pattern, the loss level is deemed acceptable, and standard control procedures can be deployed. If a new store displays very different characteristics of risk and value, new approaches to customer management may need to be adopted.

Again, we stress the point that these are examples of characteristics that indicate “goodness” for many companies, but they may also be counter-indications for other companies. In some cases, the combination of attributes is important, rather than the possession of any individual attribute.

Bad customers have characteristics largely the opposite of those listed above. For most organisations, the key bad characteristic is current and/or likely future unprofitability, though there are circumstances in which unprofitable customers are highly valued e.g. as recommenders. Companies may include in their definition of bad companies debtors, switchers, liars, or those with court judgements against them. Charities, public sector bodies or private sector bodies acting on behalf of these organisations often focus on such customers.

As companies seek to manage customers more as individuals – but remotely, so some bad customers learn to exploit this tendency. This is most visible in the areas of credit and debit cards and the Internet, but also in insurance and banking. Bad customers learn very quickly because the incentive is often very large (the potential gain) and in some cases transfer their learning very quickly (often to other individuals within an organised criminal network, but sometimes also within their ethnic or professional group or geographical locality). For example, government benefit frauds are often organised within ethnic groups or families. Tampering with utility meters often spreads geographically. Some customers target particular types of organisation. Government is often regarded as a “fair game”, by individuals across the economic spectrum. Utilities are often regarded as fair game too. In the UK, recently privatised companies (e.g. utilities, rail companies), whose directors were labelled as “fat cats” by the government when they profited from big increases in salary or selling shares at great profit, became a “legitimate target”.

To give the reader an idea of the extent and variety of “bad customer” situations, and their typical correlates, here are a few examples that differ from the usual example.

• Certain apparel shoppers (mostly women – because they are the majority of apparel shoppers) buy merchandise knowing that they are going to wear it once, before taking it back to the shop to exploit the shops liberal returns policy. Shops develop strategies to deal with this, typically taking the merchandise to the back office and smelling it for signs of body odour, deodorant, perfume etc – sure indicators of the garment having been worn for an extended period. If this is suspected, management is called in and the customer is challenged.
• A very high percentage of claims on holiday insurance policies are fraudulent, particularly those involving claims of lost cash. Insurers are dealing with this by developing databases of frequent claimers.
• Certain flyers forge airline boarding passes to obtain frequent flyer points. One airline found that there was a correlation between forging and choosing ethnic meals.
• Certain utility users always pay at the very last minute, when they are about to be disconnected or have a prepayment meter fitted. One utility found this to be correlated with ethnic grouping.
• Certain customers make a habit of claiming that different types of cleaning and washing fluids damage the item being cleaned – or of course their skin! They are traced by keeping properly computerised records of complainers’ identities.
• Certain small businesses persistently claim that their suppliers are short-delivering them – maintaining that they were too busy to check the delivery in the presence of the deliverer. Companies deal with this by sending double-checked trial deliveries.
  

We could give many more examples of this.

Bad customers are not to be avoided at all costs. Some organisations have as one of their objectives to serve customers defined as bad (by these organisations and/or by other organisations). Examples include police forces, social welfare agencies and charities. As we have mentioned, private sector organisations can design products for bad customers (tamper-proof electricity meters and phone booths, insurance products with specific risks excluded, service products with prepayment tariffs). Similarly, slack product or service design can turn apparently good customers into bad customers, and indeed whole markets from basically good to quite bad. For example, recruitment of customers for cable telephony or motor insurance irrespective respectively of their propensity to pay their bills or to switch on price led to many more customers turning bad.

The knowledge of past patterns of risk is vital for organisations trying to reduce risk, so they are keen to obtain and use this data as early as possible – ideally even before they open a dialogue with a prospective customer. Here, de-marketing can take the form of selecting apparently risky customers out of marketing campaigns. For example, the Claims Underwriting Exchange, a database which helps identify fraudulent motor claimers, was originally designed to be used at the point of claim. However, insurers wish to use it at point of customer enrolment, or even before then, rather than incur the costs of targeting and recruiting customers with a history of fraud. Note that as patterns of badness change, the organisation will need to continue to refresh its knowledge. The organisation needs also to have enough good customers (at least one at any time!) to enable it to define good too!

Identifying risk levels and potential value is one thing. Developing and implementing policies to control risk and capture value is another. So how do companies and governments limit individual exposure to risks of all kinds, from moral hazard to less culpable risks? What systems, data and management practices are involved?

Here are my recommendations for dealing with this difficult area:

About the Author
Merlin Stone is IBM Professor of Business Transformation at Surrey University, IBM's Business Research Leader and Director of QCi Ltd and the Database Group Ltd.