One of the longest-running debates in marketing has centred on the value, or lack of it, that service businesses derive from their loyalty schemes.
Retailers in particular have proved shy of specifying the cost/benefit ratio of running what have proved to be extremely expensive promotional and incentive tools.
Another criticism levelled at loyalty schemes is that they appear more obsessed with signing up new people (and hence acquiring new data) than actually rewarding loyalty.
This cannot be said of the Boots Advantage Card scheme. With 15 million households having one in their wallet or purse, it's one of the UK's most popular loyalty cards. Cardholders benefit from a generous reward scheme, getting four points for every pound sterling they spend.
Yet earlier this year the retailer acknowledged it needed to rethink the rewards it offers, and decide which customers should receive them, as part of an overhaul of the Advantage Card scheme. The card is seen as vital to Boots' efforts to fend off competition from supermarkets' expansion into health and wellbeing products and services.
"Customers have a very high level of engagement with the card programme, especially women who see the points as providing a treat for themselves," says John Wallinger, planning director at Boots DM agency Craik Jones Watson Mitchell Voelkel. "But we recognised there was a great opportunity to get a customer to buy a wider breadth of goods from Boots than they might be buying elsewhere."
Potential to buy
Boots' quarterly coupon mailings to cardholders, sent with the co-operation of Boots suppliers, typically selected the retailer's highest value customers, missing the opportunity to grow spend from the majority of the customer base. This left millions of cardholders, many with a high potential for Boots, receiving very little communication from the health and beauty retailer.
A new way of selecting customers was called for. Early in September, Craik Jones and retail analytics company 5One began work creating a targeting system that could identify potential customer spend across categories, brands and individual products. The solution needed to identify a person's potential to buy brands and make the offer selections.
"The objective is to reward customers to keep them coming back into the store and also to get them to buy in areas where they aren't already buying," says Ian Scholey, senior analyst and associate director at 5One. "We needed to work out which products they have most affinity with and how much we have to incentivise them to continue shopping."
Two models were created, looking at each individual's past transactional history, and their future potential (see box). The first model, named Peer Group Comparisons, calculates affinity to a product, brand or even category, based on demographics, product purchasing and, critically, the identification of related products. "If a customer was spending heavily in skin care but not in makeup, this would tell us to incentivise them with vouchers for make-up," says Scholey.
A key feature of this model is to stop unnecessary incentivising, because of the cost to suppliers. "It's great to continue incentivising customers to keep them buying one product, but also to encourage them to purchase products they're not buying," Scholey adds. The Peer Group Comparisons model would identify so-called secondary products - those that could be bought in conjunction with a customer's regular products.
The Share of Wallet model identifies and scores customers who are not purchasing at their optimum level, based on two years of purchasing history. It looks at purchase history within various product categories, working out potential spend and comparing this with actual spend.
Used together, the models allow communications to be tailored for each customer. The system runs through the database to select customers according to pre-set criteria and the value of the incentive they are then sent is personalised based on the customer's model score.
The new targeting system sits on the desktops of Boots' analytics team based in Nottingham. A key benefit is the help Boots can give its many suppliers, all of whom have vastly different propositions. "We might have an objective for a supplier or for Boots itself to cross-sell within a brand or steal sales from another retailer," says Wallinger. "So we look at customers' current spend within product lines and work out, looking at other customers, the average they spend and the potential to grow. For instance, you might be buying x amount of product y, but we know the average customer is buying 10 times that."
The project partners delivered the models within two months. Boots says it's too early yet to quantify the results, but 5One has conducted analysis on previous campaigns that didn't use the model and has discovered uplifts of 30-35 per cent in response.
"We're putting a huge amount of investment behind the Advantage Card," Helen Jeremiah, head of customer insight at Boots Advantage Card, explains.
"The direct marketing element is about making the most of that investment for our customers and for Boots' business. The models are the core part of this, helping to make relevant offers to customers that benefit them and our suppliers."
The model build was done using a five per cent sample of Boots transactional data, refreshed every month, with a lengthy model and IT testing schedule applying it back to the Advantage Card database.
SAS software was used to develop the models, combining various statistical techniques, including regression. For the Peer Group Comparisons model, 5One used the Yule's Q statistical technique to deliver accuracy.
"Yule's Q is a statistical measure of sequential association which says how well products relate to each other so that, for instance, two competing brands of skincare will be shown to have a weak relationship," says Ian Scholey, senior analyst and associate director at retail analyst 5One.
The share of wallet model looked at two years of purchase history within the various product categories stocked at Boots, to see if customers were spending to their full potential. "Every customer's share of wallet and therefore potential spend was calculated - how much are they spending now compared with in the past," says Scholey.
The result is a set of parameters for every product line and mailing, in terms of objectives and volumes of mail. The system is driven by SQL which runs through the database to select customers who fulfil the parameters and selects them for mailing.