Those outside it frequently castigate the Direct Marketing (DM) industry for over-communicating with potential and existing customers (“junk mail syndrome”). Strenuous efforts to minimise mailing volumes and maximise response rates are therefore legion amongst DM practitioners. This article argues that it is time for a step-change in thinking about targeting that matches the UK’s current change from being a producer-led to a customer-led economy. DM practitioners need to move from using generic data for targeting that is provided by the market to finding and demanding the exact data that they need for the exact customers they need it for. The industry is the customer in this respect and must itself become fully customer-centric in meeting its own needs if it is to help its clients achieve the promised returns from CRM. We explain below one step-forward in this respect that we have been using for the past 5 years with great success. To help illustrate our key points we have used learnings gained from our work with a major European Gardening retailer over this time.
Marketing research, customer targeting, one-to-one marketing, usage & attitude study, database marketing, customer motivation, consumer behaviour
Targeting the Right Problem
In 1898 the British Navy found that for every 1000 rounds fired by its battleships only about 17 ever hit the target. Then Admiral Percy Scott studied the techniques employed by his gunners and discovered the most accurate were those who were better at allowing for the pitch and roll of the ship. Using this knowledge, he designed a new automated aiming mechanism that took their hit rate to over 50%.
Direct marketing is facing a similar dilemma. The information age has reduced the cost of delivering messages and increased exponentially the means of delivering them. The resulting deluge of communication has seen consumers donning their hard hats and rushing for cover. With eleven million tonnes of paper being dropped on UK consumers every year it is little wonder many are opting-out.
Current practice is a game of “Battleships and Cruisers”. Using a “Test & Learn” methodology dating back more-or-less to the era of the game itself, a never-ending series of “hit-and-miss” strikes uncover what works and what doesn’t. Admiral Scott would think we had learned nothing in 100 years and would probably have a point.
Modern armaments take Scott’s principle even further. By working backwards from the target, detailed intelligence ensures 98% accuracy, not 98% inaccuracy. DM practice needs a similar leap of thinking and making better use of market research is key. Depth of customer understanding is something each company can differentiate itself on; everything else is technology-driven and hence copy-able by competitors. This makes it the best (or only) way of achieving sustainable competitive advantage.
Taking the Consumer’s View
Working backwards from the target takes much greater effort. It needs close links with “special forces” close to the target who can uncover individual customer needs, pinpoint when, where, and how to contact them, and determine the best message to get the job done.
Using insights from market research would seem the obvious solution yet current attitudes within the DM industry would appear to be against this. In a recent edition of Marketing Direct , Holly Acland says she was told by the MD of one major DM company: “Market research is useful but it’s only the mental picture stuff. If I’m managing the customer, and the revenue that customer represents, I can do that without knowing the ‘why?’”. One only wonders how.
Acland concluded that market research has a valuable role in influencing the tone of the copy, influencing product development and measuring customer satisfaction but little else, quoting another senior DM executive as saying “There’s a limit to the amount of data you can collect and the number of people you can collect it from. So there’s a degree of assumption and therefore inaccuracy in applying the model.”
We disagree. Market research will tell you where to dig, what tools to use, and how rich a deposit lies beneath the surface. Done correctly, it identifies new measures of the customer, even from comparatively small samples. This article shows how this can be done. We begin with the underlying model.
Using Market Research to Understand Customer Motivations
Like Dick & Basu and Huddleston et al, we believe behaviours are driven by interactions between an individual’s attitudes and circumstances. Using this framework to model customer-specific data from loyalty card schemes we have found consumer purchase behaviour can be predicted using 5 common-sense motivations:
• The Relevancy of the Product or Service on Offer
• Identification with the Brand or Company providing the service or product
• The Accessibility of the Brand or Product
• Its perceived Value for Money
• Their Confidence they will be satisfied with what they will get
We present below examples from our work for a European gardening retailer where, with the respondents express permission, we have combined usage and attitude data, from a sample of over 3000 cardholders, with their loyalty card transactions.
Relevancy drives the level, mix and timing of engagement with a category and, unsurprisingly, factors such as age, life stage and other geo-demographics correlate with usage of a gardening store. Chart 1 shows category engagement (number of visits to the store over a 12-month period) versus customer age. The highest engagement in terms of both customer numbers and number of visits per year, occurs in middle age.
Chart 1 : Number of Visits per Year by Age of Customer
Interestingly, those under 25 and between 64 and 74 have disproportionately high engagement, implying these customers may have a higher interest in gardening than the average for their age group. Chart 2 supports this, where combining transactional records with data from the usage and attitude survey shows a strong and statistically significant relationship between the number of visits made to the gardening store and the customer’s claimed level of interest in gardening.
This is significant because it illustrates how U&A studies uncover attitudinal drivers that will predict category participation. It also enables us to find spend and visit behaviours that distinguish differing levels of interest.
Chart 2 : Number of Visits per Year by Interest in Gardening
This means we can understand the impact of customer attitudes without surveying the entire database and without needing to develop a projective model for attitudinal segment membership that is based overwhelmingly on geo-demographic overlaps. The U&A study delivers all the information we need to relate each individual’s attitudes and circumstances to their recorded purchase behaviour.
It also means we can find the measures we need to create from the data already on the database to identify customers with differing motivations, and what else is needed to reliably predict their engagement with the category and our brand on-going. And we can quantify how much better these models will be compared to existing models.
Chart 3 : Seasonal Visit Pattern Differences for Customers who “Dislike” Gardening
Lastly, it means we can use the model for “What if..?” modelling to evaluate the relevancy, appeal and hence “size of the prize” of potential new propositions.
b) Brand Identification
Many companies have conducted Usage & Attitude studies and have a market segmentation model that separates people according to their likelihood to consider using their brand. Couple this with information from the database and we can now derive how much they are likely to use our brand in specific circumstances.
Chart 4 presents an indirect example of this. It shows that there is a correlation (other things being equal) between number of store visits and number of loyalty cards held.
Not surprisingly, those holding many cards are less loyal but, more interestingly, those who claimed not to hold any loyalty cards also visited less. However, to take part in the survey they had to be cardholders. This lack of “top-of-mind” awareness of the card suggests a lower brand-engagement that appears to be reflected in behaviour.
Differences between more- and less-committed users enable us to tailor the message accordingly and better predict likelihood to respond. For customers who are more price conscious we may emphasise the price discount aspect of an offer and provide reassurance on quality. For more attitudinally loyal customers we may do the reverse. Combining research and database information gives greater confidence in and known reasons for response that can be leveraged in marketing communications.
Chart 4 : Numbers of Visits per Year by Number of Loyalty Cards Held
Accessibility measures the external circumstances that determine the consumer’s ability to take advantage of the proposition. For instance, I may get a mailing about a Mercedes but may not have the means to afford it, or may live far from a Mercedes dealer making servicing inconvenient. These factors are similar to those determining Relevancy, but either work at a more detailed level (e.g. the second Mercedes example); or relate to barriers affecting a particular individual (e.g. disability).
In retailing, a strong Accessibility factor is distance to store. Chart 5 shows this for our gardening retailer.
Chart 5 : Impact of Distance to Store on Number of Visits per Year
Existing awareness of the proposition, from TV advertising for example, is another Accessibility factor. Data from advertising research (Chart 6) shows how differing segments changed their spend according to their exposure to the retailer’s TV advertising (other things being equal).
Chart 6 : Customer Response to Advertising Pressure
This has important implications for integrated campaigns where mailings and other direct communications targeted towards customer acquisition are much more successful if harmonised with above-the-line campaigns. These generate awareness of the proposition; and the direct marketing component acts as the call-to-action.
d) Value for Money
Value for Money is the consumer’s judgement on whether the perceived benefit outweighs the perceived cost. Any lawnmower will cut the lawn but choosing which one to buy will also depend on the size of your lawn and how much you like mowing.
To understand which measures are most predictive of the consumer’s view of Value for Money, we typically consider 3 aspects: the individual’s attitudes to value; the relative price compared to competitor solutions; and the consumer’s view of the total benefit the solution provides relative to competing alternatives.
Market research is especially valuable in uncovering which behavioural characteristics best reflect these opinions. If I would not consider buying a Mercedes because of the lack of servicing facilities nearby (a perceived cost), I may become more likely to buy if offered a “free collect and return service”. This may be worth more to me than it costs the dealer to provide. Knowing the relative importance of differing value drivers in the eyes of each consumer gives us valuable clues as to their likelihood to respond to our new “free backseat satellite TV” offer.
Chart 7 : Percentage of Annual Spend that is Spent on Discounted Items
Charts 7 and 8 present examples of how perceptions of Value for Money and attitudes to life influence behaviour. Chart 7 shows a clear relationship between the percentage of the respondent’s total actual spend on discounted products and their claimed level of price consciousness. This allows us to compare many other behavioural variables against this dimension in order to understand which are most relevant.
It could be claimed that the behavioural data alone is all that is needed to do this and that the market research information is merely confirming what we already know. Tesco, for example, claimed that they used their Clubcard data to identify products that were signatures of people who shop exclusively on price. Accordingly it was identified that buyers of Tesco Value margarine were such consumers and targeting them with price offers was extremely successful. No doubt.
However, this is probably simply a better indicator of disposable income rather than perceived value for money. Someone might never buy Tesco Value Margarine but might readily switch store if they knew the cost of their average basket was going to be 10% cheaper at a supermarket nearby. Being price sensitive doesn’t necessarily mean someone shops exclusively on price.
Tesco does not know the income level of each of its customers but we do know that income is a key driver of Accessibility. We could therefore design far more effective and profitable ways to retain the loyalty of the Value Margarine buyers than price promotions. By working with suppliers we could offer retrospective non-price bonuses for staying jointly loyal to their brands and to our store. Couple this with promotions on “treat” items such as mid-priced desserts and we could generate long-lasting loyalty whilst growing their spending with us.
Chart 8 : Customer Profitability versus Attitude to Home
Identifying and validating such opportunities is where mixing market research and transactional data comers into its own. Chart 8, shows that attitudes to life influence perceptions of Value for Money separately from price sensitivity / price consciousness. Customer profitability data shows some consumers are willing to pay more to get the result they want. This helps us to predict their likely price sensitivity.
Following the Tesco view, this relationship should break down for the most price sensitive consumers but, as Chart 9 shows, the relationship still applies. Just because someone is more price sensitive than the average does not mean other considerations are ignored when judging Value for Money. Using our approach, Tesco’s customer management strategy could easily be outmanoeuvred.
Chart 9 : Customer Profitability and Attitude to Home for Price Conscious Customers
e) Expected Satisfaction
Expected satisfaction, or confidence, comes from previous experience, word of mouth recommendation, comments in the press and so on. For existing customers, current satisfaction is a good indicator, other things being equal, that they are willing to continue to deal with your company.
One concept appropriate to CRM is Ipsos’s Customer Delight Principle (Keiningham and Vavra). This asserts that customers’ satisfaction levels influence customers’ loyalty only when more extreme levels of satisfaction are reached (“Dissatisfaction” or “Delight”). And small changes in satisfaction between these extremes have much less impact.
In our view, this is because only these extremes cause customers to fundamentally reappraise. Dissatisfied customers dismiss future communications either because they doubt the company’s competence or feel the company does not care enough about its customers. Delighted customers consider the company is “good” because, even if things go wrong, they care enough to fix the problem.
Unfortunately delighting customers costs money, possibly lots of money. Modelling combined market research and transactional data helps evaluate the trade-offs.
Chart 10 illustrates these effects for two samples of customers using Customer Satisfaction as a measure of consumer confidence. In the first sample, 2258 respondents, we compare overall level of satisfaction with numbers of visits to the store in the twelve-month period prior to completing the satisfaction questionnaire. In the second sample, 2862 respondents, we compare with number of visits to the store in the twelve-month period after completing the questionnaire.
For the first sample, apart from the highly dissatisfied customers, there is little correlation between visit frequency in the year prior to completing the survey and satisfaction at the end of it. For the second sample, there is a statistically strong and significant correlation between their subsequent visit behaviour and their satisfaction at the start of it. Moreover, the largest differences in behaviour for this group occur for the Dissatisfied and Delighted customers for whom the index value compared to their neighbouring groups show the widest variation (7% – 10%). Amongst the “Merely Satisfied” groups the differences in behaviour are much smaller.
Interestingly, the visit frequency for the Highly Dissatisfied group who completed the questionnaire at the start of the year was not as low as for those who did not. It is therefore possible that by having had the opportunity to “Have their say” some of their negative attitudes were calmed and the potential negative impact on their behaviour moderated.
Similar results exist for their levels of spend, average spend per visit and likelihood to respond to CRM communications. So customer satisfaction should be a key element in our targeting models even though we may not be able to survey the satisfaction of everyone on the database. We must instead find measures of customer behaviour that indicate whether they are likely to be “Delighted”; “Dissatisfied”; or “Merely Satisfied”.
Chart 10 : Relationship Between Confidence in the Brand and Customer Behaviour
The proposed approach has several key advantages. It is easy to see and understand the key drivers and define the measures one needs. It evaluates the CRM proposition from the consumer’s perspective and intrinsically identifies the drivers of acceptance of a proposition not just its potential applicability.
Crucially, it is more predictive of actual consumer response than simple attitudinal segment flags (see Chart 12). For all the talk about how Tesco has leveraged its Clubcard data, the truth is transactional data are retrospective. Incorporating market research puts the customer’s opinions at the heart of your targeting. New and existing propositions; known and potential requirements; prospective and existing customers can all be explored in a consistent and marketing-oriented way, and valuable insights uncovered that would otherwise be missed.
Whether your customer data is derived from on-line or off-line transactions, or even if it only contains quite basic information about your customers, the same approach can be used. Bringing together the behavioural and circumstance data from your database with the attitudinal and circumstance data from market research will produce customer insights that will enable you to increase the returns from your CRM programme.
Armed with this knowledge, communications can be tailored to individual customer groups based on their potential receptiveness to your proposition and couched in terms they find most motivating and the proposition refined to fit individual customer needs more closely.
Other initiatives, such as gathering potential leads from contact databases, web-site referrals and so on, can be analysed and profiled using the same techniques. You can better judge their potential and, more importantly, justify the investment needed in your CRM capabilities to exploit them. One client, with a direct sales force and call centre operation, used the model to better predict what type of proposition would be most motivating and whether a prospect was likely to convert. This helped prioritise staff time on the best opportunities and increased conversion rates by 38%.
RedRoute CVM Model
RRI’s CVM model provides a structured framework for modelling the interactions between the behaviours, attitudes and circumstances. A database containing information on each dimension for a large sample of customers is compiled to both segment customers and develop predictive models of behaviour. By revealing which behaviours most closely correspond (given each customer’s circumstances) to particular attitudes, we can better predict their likely response to new propositions and their future needs.
Having obtained a respondent’s explicit consent, their market research data is either directly cross-matched with their behaviour and characteristics from the client’s database or data fusion is used to “marry” respondents from survey research with surrogate donors from the database. This is carried out confidentially by the research agency to maintain respondent anonymity. Naturally there are potential limitations based on the nature of the survey questionnaire and the ability to forge the links to the database. This therefore has to be undertaken expertly. Usually a mixture of both approaches is used to obtain a dataset covering around 10% of the customer base.
Diagram 1 : Motivational Modelling
For our gardening retailer this enabled motivational models explaining various aspects of garden shopping to be developed. The actual results are confidential but Chart 11 shows the type of model developed to explain visit frequency. As can be seen, a number of the key drivers are attitudinal, enabling the client to produce more tailored propositions. A marketing campaign can then be as motivating for a keen gardener as it is for someone who just wants to clear the weeds from their front garden.
In addition, understanding which measures of customer behaviour are most predictive of differing customer attitudes then enables us to project attitudinal data with confidence across the database without having to survey to every customer.
Typically we find some highly predictive attitudinal or circumstance information, usually between 3 and 6 key items, cannot be proxied using data already on the database. In this case the client can put mechanisms in place to obtain this information continuously and cheaply and do so just for those customers for whom the information will make a difference to our effectiveness. This saves money by avoiding the need to run regular large surveys. We also phrase the questions so as to provide direct links to the database. For example, not just “Are you interested in gardening?” but “How many times would you say you visit a garden store each year?”.
Chart 11 : Drivers of Visits to a Gardening Store
Naturally a company will only wish to enhance their targeting in this way if the rewards are worthwhile. The immediate benefits achieved by our retail gardening client, namely a step-change in response of more than 50%, generated additional profits and a project ROI in excess of 500 that more than justified their investment.
Chart 12 shows how the five factors described above contributed to this improvement (excluding “Confidence“ which was a later refinement). To make the communications more relevant and target customers who were more likely to respond, the model placed greater emphasis on Relevance and on targeting people with greater Brand Identification. They were less price sensitive and had the potential to uplift spend. More value-driven messages were directed at price-sensitive customers who were most likely to both switch and repeat. The result was more sales-effective and more profitable marketing campaigns, increasing ROI’s from 2:1 to 10:1.
Chart 12 : Response from the Motivational Model versus Client’s Original RFM Model
Conclusion: The Marriage of Market Research and Database Marketing
Using market research in developing one-to-one targeting models is both feasible and, if done correctly, highly productive. It requires considering the five key drivers of response and using the data to model the relationships between attitudes, behaviours and circumstances. In so doing it provides a new “ABC” of targeting effectiveness.
It tells you how to mine your data; what to look for; what extra data to collect and what themes and messages would be most beneficial in driving response rates.
This brings increased predictability and more easily justifiable database investment – typically doubling the returns from the CRM marketing budget and providing evidence to justify continued investment in database marketing that even Finance Directors cannot argue with. We have presented some straightforward but powerful examples of how circumstances and attitudes affect customer behaviours and ways in which this can be used to better meet consumer needs and increase company sales and profit performance.
However, when it comes to the marriage of market research and database marketing, one phrase says it all: “My wife doesn’t understand me!” It’s often difficult for Database Marketers to see how to use research to improve their ROI’s because it often covers only a few hundred people, the data are anonymised and they cannot usefully be projected across thousands of people in a database that is required for targeting.
The route outlined is harder than profiling a set of targets and picking a list. Unfortunately, because modern CRM systems mean we can now bombard consumers with communications many are now being used to do just that. Consequently there has been an explosion in the number of messages being sent to consumers with no corresponding increase in response per communication. In fact, one could make the case that response per message is now lower than when the industry was simpler!
So for the benefit of all concerned, and for long term consumer trust, what is needed is to replace what some commentators have termed “Carpet Bombing” with a situation where we know how each communication we send fits into the life of the consumer, what they are going to think when they get it, and where the expected response is 98% rather than 2%.
Nowadays at RedRoute, provided we have a respondent’s explicit consent, we often analyse cross-matched survey and customer data for clients. This enables us to understand both the “What?” and “Why?” of consumer behaviour and the learnings directly impact on our client’s database marketing strategies. Such analyses will only increase in the future as more and more digital communications are used for customer contact and tracking. The benefit of taking this more considered approach will be a step-change in targeting accuracy, and an improvement in effectiveness that we hope even Admiral Scott would be proud of.