Loyalty’s Royalty

FOURTEEN MILLION BRITISH SHOPPERS watch the mail every three months for a check from their grocer. Then they spend like it’s Christmas.

The quarterly dividend checks are the backbone of Tesco’s nine-year-old loyalty program. Clubcard holders get a 1% rebate on purchases — an average GBP 150 million per year (about $100 million), more than GBP 1 billion since its 1995 launch. Along with their checks, members get a slick four-color magazine and offers that are so carefully tailored to each recipient that Tesco prints four million variations for each quarterly mailing.

Clubcard has made Tesco the No. 1 grocer in the U.K., boosting its market share to 27% in 2003 from 16% in 1995 and catapulting it over longtime competitor Sainsbury’s.

That was just the beginning. Tesco has used Clubcard to launch a bank, telephone service, and a wide-ranging Web site that rivals Amazon.uk. A handful of “sub-clubs” cater to niche audiences, from new mothers to wine lovers.

In the U.S., Kroger Co. is adopting the Clubcard model and could launch as early as this year. The Cincinnati-based grocer is testing pieces of a program in different markets, and could have a sizable program by 2006.

Relationship and loyalty marketing increasingly appeal to retailers who are tired of competing on price. Tesco’s model matters most to retailers who are looking for a way to make an end-run around Wal-Mart.

“There are going to be two types of retailers: Wal-Mart, which focuses on supply chain, squeezing costs out and competing on price and location, and Tesco, which competes on customer knowledge and relationships. There won’t be any in-between,” says Don Schultz, Northwestern University professor and marketing consultant.

The real value of Clubcard data is invisible to its members. Tesco analyzes shopping baskets to segment shoppers, then dives into the data to see who might buy more, or what non-grocery services may appeal to members. The data prompted Tesco Personal Finance, now serving 3.5 million members, and Tesco Talk phone service, with an estimated half-million (and counting) subscribers since its mid-2003 launch.

“Clubcard is not just a loyalty card; it is a business system,” says Schultz, a former board member of dunnhumby, the marketing analysis firm that helped Tesco launch Clubcard and still manages the database. “They are way beyond rewarding customers and retention; they are using data to drive business decisions. Tesco is the most sophisticated marketer in the world.”

Much of the credit goes to Clive Humby, one of Clubcard’s architects and co-author of Scoring Points: How Tesco is Winning Customer Loyalty. Humby is chairman and founder of London-based dunnhumby, which has functioned as Clubcard’s in-house database division since Tesco bought a 50% interest in the firm in 2001. Humby is a mathematician who pioneered geo-demographic segmentation nearly 20 years ago. He spoke from his London office with PROMO Senior Editor Betsy Spethmann about the mathematics of getting groceries into baskets, about rolling balls and about making TV obsolete.

PROMO: Describe Clubcard for U.S. readers.

HUMBY: It’s the most widely used loyalty card in the U.K. It’s rather different than typical grocery cards found in America, which tend to be discount cards. The basic personality of the brand is, “This is a club, and because you’re in the club, we will reward your regular shopping with Tesco by giving you a 1% rebate.”

One of the big questions is, Why mail? It’s an expensive way to deliver it. But if you mail people, you get a bigger uplift. We send a statement that delivers money. Shoppers come in to spend that money and they spend more because they’re feeling good about the store. That mailing then becomes a retail medium. We’re talking to 35% of U.K. homes in any mailing.

Over the years we’ve gotten smarter, realizing we should encourage people to trade up in the category they’re in, based on products they like, or try products that, based on everything else they buy, suggests they’d like.

That 10 million mailing has got nearly four million variants now, so you’re nearly one-to-one.

We’ve not got four million offers; we got a variety of offers, and we optimize the mix of offers that we give each customer. As a result, we get tremendous uplift — it’s not uncommon to see product coupons get 20% redemption.

PROMO: How do manufacturers participate in that?

HUMBY: They decide which shoppers they’d like to make. We optimize the offers for the customers. We work out the best mix of offers to give each customer, and then we tell the manufacturer how many customers we’ll be giving their offer to.

PROMO: How much of the program is funded through vendor dollars?

HUMBY: None of the physical reward is funded through vendor dollars, although the vendor dollars do cover the execution costs of the mailings.

[Each Clubcard mailing] is such a powerful medium for manufacturers that it’s not treated like trade marketing dollars anymore. It’s seen against the advertising dollars rather than trade marketing dollars alone.

PROMO: Clubcard mailings have supplanted traditional media: At holiday time Tesco didn’t even do any advertising.

HUMBY: Tesco used to use a lot of TV, was one of the biggest TV advertisers in the U.K. As we grew the amount of direct communication with customers, it was so relevant and powerful that it gave a far bigger uplift than TV advertising. It was directly measurable, which meant people were more comfortable spending money on it, and as we’ve learned more and more, it’s given them the confidence to use the direct communication as their principal medium.

In grocery in the U.K., it’s not about winning more customers, it’s about winning more of the dollars of the customers you’ve already got. Because they’re in the [Clubcard] scheme, we know exactly what they’re like, and effectively we use the shopping basket to really understand them and be very, very relevant to them. That’s made the direct communications very cost-effective and very powerful.

PROMO: You segment Clubcard members in an unusual way, scoring products on different characteristics and then tracking how shoppers buy those products in combination — your “rolling ball” method. Why did you do it that way?

HUMBY: We all see it in the queue waiting to pay, when you look in someone else’s shopping basket and you create a picture of the person based on what’s in their basket. My background is in demography, and it struck me that if we could apply the same principles to shopping baskets, then we could create a language of the consumer.

You want to give the basket a series of profiles: How much of the basket is things on promotion, how much is things the shopper regularly buys, how much is adventurous. The problem is you’ve got so many lines — 40,000 to 50,000 SKUS stocked in a typical store — that you can’t just sum up sales by line. You’ve got to get the lines to group together. That’s where the rolling ball idea came up. We can classify something fairly obviously — if I look at pot pies or own-label toilet tissue or exotic fruit, I can classify those easily. So there were things I knew I could classify intuitively, with about 30 or 40 labels.

PROMO: So your scored each SKU on 30 or 40 different characteristics?

HUMBY: Yes. Now, you can’t do that for 50,000 items, it would take far too long. We took the top tier and started off by scoring some very obvious products. And then the rolling ball technique picks the things that, depending on how often they’re bought together, scores the things we’ve not yet counted. So it basically forms groups of products that look like they’re bought in a similar manner, and gives them similar characteristics. And that enables us to classify 10,000 to 12,000 SKUS with all of these attributes.

PROMO: How long does it take you to classify that many products?

HUMBY: A computer algorithm classifies it. We only have to start with about 50 very obvious products to classify by hand, and then the rolling ball technique allocates the rest. It probably took us about a week to classify the products.

PROMO: I’m picturing someone sitting at a desk, looking at a papaya, trying to rate it on 30 characteristics.

HUMBY: That’s a good example, actually, because it’s easy to classify: Adventurous, fresh, premium price point, short shelf life. We can give that four or five attributes almost instantly. You can do the same thing for another 50 or 60 items that have reasonable volume and are quite distinctive, and then the rolling ball technique looks at product associations.

PROMO: So it looks at other things in the basket of the person who bought the papaya?

HUMBY: Yes. You look at the occurrence of the papaya with a pineapple, or large pack of frozen chips, and are they associated with both, or only one? It goes around that process over and over again [to define the key products used for tracking].

PROMO: Then how do you use that information? You’re using shoppers’ behavior to make deeper assumptions.

HUMBY: What people do is much more predictive than what people say they’ll do. What you do is changing all the time. So if we come up with something purely behavioral, we get far better leverage because it’s constantly up to date. If I can look in your shopping basket and it shows me how you’re behaving, then I can track trends. It becomes a window onto the customer that you’re in complete control of.

PROMO: How many groups of consumers did you define?

HUMBY: We’re looking at perhaps 30 different types of shopping baskets by five different levels of promotional activity, by eight different preferences for when you like to shop, by a number of different life stages for the family — and all of the sudden you’ve got four million variations. We’ve got six or seven segmentations working together, and by the time you apply those combinations you get to four million.

On top of that, some of the offers are conditional spend, so they’re directly tied to what you do personally. You might spend $12 a week on fruit, and we’ll make an offer for $3 off a $15 purchase. You still spend $12 on fruit, but you stretch the spend a bit. Which is all incremental sales for Tesco.

PROMO: This insight into shoppers can be a platform for short-term promotions. How does Tesco do that?

HUMBY: We have a number of channels to the customer — the statement, in-store promotions, other mailings and till talkers where we can have a dialogue with customers at the till as part of the transaction.

We use those to optimize the shopper’s experience. Manufacturers want to run a wide range of promotions, but in any one week, we make sure we’ve got the right mix of promotions to meet the promotional needs of the widest range of customers. We’re constantly optimizing what we’re being offered from all the different manufacturers to get the maximum uplift from the customer, rather than just taking the most dollars that we could.

By taking the most dollars, you might find you’re spending mostly against young families that are promotionally promiscuous. That might be a good thing for manufacturers, but it’s not such a good thing for the customers, because certain segments of the customer base don’t see the promotional image of the store, because they’re not getting promotions that are relevant to them.

Our job is to optimize the mix to get the best coverage against all customers, and get away from just the very promotionally focused customers driving the promotional behavior, which ends up with a self-serving model.

PROMO: At what level can Tesco staff use Clubcard data for promotions? Can a single store manager use it?

HUMBY: Yes, for local competitive activity. We try to make the store manger’s life as simple as possible with the mainstream promotional calendar. But a store that’s impacted by a competitor opening or a particularly aggressive piece of marketing often wants to enhance that. We treat the local marketing rather different than the general promotional calendar. About 80% of marketing to Clubcard holders is national.

PROMO: What happens in stores when the quarterly mailings go out? How do stores gear up for that?

HUMBY: We’ve got projections for what will happen in each store, based on the statements going out, so we can predict what will happen. The key is making sure the supply check can keep up.

We tend to see a lift nearly as big as a Christmas lift each time we do a statement. That puts a real test on the supply chain, and it does challenge the classic supply-chain model, which has the computer re-ordering products for the following week based on the previous week’s sales. When you do this sort of marketing, you have to correct for that. Otherwise you get in-store discounting because you get too much volume on the shelf.

PROMO: Tell us about the magazine that goes out with quarterly mailings.

HUMBY: We segment the magazine based on life stage — young families, empty nesters and so on. We tried the economies of overlaying segments — not a very detailed segmentation, but up-marketer versus price-conscious customers, distinctions like that.

Marketers can buy ads in the magazine and be in maybe 500,000 envelopes, or go into the copy of the letter that goes with the magazines and be in anything from 1,000 to two million envelopes because of the way it’s segmented. It becomes quite a complex channel.

PROMO: What do you want to know about the 80% of shoppers who only account for 20% of your business? Why would you even want to track them at all?

HUMBY: We try to win a little bit more of their shopping, and they’ve got more headroom [opportunity for incremental sales]. We estimate what they’re spending elsewhere. In the crudest terms, we count the calories we sell them, and then look at how many people are in the household to figure out how many calories they’re buying elsewhere.

PROMO: That’s a very specific share of stomach.

HUMBY: (laughs) We actually do it based on meal events, but that’s the basic logic. When we know what we serve to customers, and understand what we therefore don’t serve them, we have a good idea where they fulfill that. We try to understand what the headroom is and get trade-drivers into that group of customers. That’s why we want to know about as many customers as possible.

PROMO: Tesco uses data from Clubcard to look for other business opportunities. To use supermarket loyalty card data to launch a personal finance and telephone service is remarkable.

HUMBY: Yes, but it’s about the brand being very trusted. Tesco has a great deal more trust than a typical American grocery store. They’ve been very good at being consumers’ ally.

Part of the problem with the American grocery scene is that two-tier pricing has created the impression that customers can be ripped off, that if you forget your card you’ve got to pay an extra fifty cents.

Over here, we’ve never had that issue. The card has always been a valued part of the consumer’s wallet. As a result, the brands are trusted, so they have the ability to take on areas where trust is lower. In the U.K., banks are less trusted than grocers, so it would be easy for a bank to lose to a grocer. As a result, it’s easy to extend the brand.

PROMO: You’re doing some work with Kroger. How is the U.S. different?

HUMBY: You’ve got much more cross-shopping in America, so we start from a different base.

We’ve built most of the basic tools we’ve got in the U.K. for Tesco on the Kroger data. We’re learning about how to classify the products you sell in America. We’re learning about some of the complications Wal-Mart poses, because you’ve effectively got store-specific pricing in a lot of areas: If Wal-Mart’s down the road the price is different than if Wal-Mart’s not down the road.

Then there’s the additional complication that Kroger has multiple formats. Do we classify products in different ways, in different formats in different brands? Using the rolling ball methodology is a bit more challenging.

PROMO: Are there other hurdles in the U.S. that you didn’t have with Tesco?

HUMBY: The balance between manufacturer brands and the retail brand is very different. In the U.K., the retailer’s brand is as strong if not stronger than the consumer packaged goods companies’ brands. In America, you’ve still got the dominance of the Procter & Gambles, the Krafts and General Mills.

PROMO: That’s probably one reason we have so much segmentation in channels of trade: If you can get the same brand at Wal-Mart or Kroger, trust is with the brand on the package, not on the store door.

HUMBY: The interesting thing is: What does that do to the packaged goods brand by the end of it? If the brand relies on a discounter format, does that devalue the premium nature of the brand?

PROMO: Tell us about the Banana Man.

HUMBY: (laughs) There was a bit of a cock-up [mistake] on points issued on bananas, and someone spotted that he’d get more money back than he spent on bananas. He started buying bananas on bulk — he ended up giving them all away — but we spotted it and it got a huge amount of p.r. It was fun at the time, even though a bit of a mistake and we got the points value wrong. But the p.r. balanced it out.

It shows how switched-on [Clubcard shoppers] are and that they understand what points are all about.

Consumers believe that Tesco uses the data to do a better job for them. And therefore, they’re very happy. We never get Big Brother issues with Clubcard.

PROMO: Isn’t privacy a big concern, especially after Sept. 11?

HUMBY: Yes, but I think that’s about abuse. If you don’t abuse the trust of customers and you give them lots of chances to opt out if they don’t want to be part of it, then they’re happy. Our job is to sell you things you do want and help you understand how to get more out of Tesco. That’s the trick: to sell and be relevant at the same time.

PROMO: Parliament recently asked Tesco to use Clubcard mailings to track shoppers buying junk food and then suggest they buy healthier items.

HUMBY: That’s the consumers’ choice; it’s not our job. Tesco is a grocer, not a public health authority.

We do have a Healthy Living Club where people opt in. We’ve approached customers who are clearly trying to lead different dietary regimes and give those people an opt-in to a Web site-based club that lets members specify health needs. Consumers like that because it’s not being intrusive. If they opt in, then we’ll engage them in a health dialogue. But we wouldn’t do that unless we got their permission.

PROMO: Tell us about the other sub-clubs, like Baby Club.

HUMBY: The main club was fine, but there were certain things that are very big drivers to spending — babies, wine purchasing things like that. We tried to develop clubs around those. There are two motivations: Not all Tescos are full-footprint stores, so this allowed us to use home delivery to extend the brand. Plus, we recognize that events like a baby’s birth, especially your first baby, change the shopping patterns of the household. So the clubs give very relevant offers to people at that life stage or, in the case of Wine Club, have a particular interest. It lets us enter a much more direct dialogue with those people with lots more levers.

If we can do a sub-club with 250,000 to one million members, it becomes a very exciting channel for vendors because they can talk to a whole section of very relevant population at competitive prices.

I passionately believe that retailers are becoming the next media channel. We can talk to very large groups of customers who are a very specialized niche at the same time.

PROMO: How do you keep shoppers interested in a loyalty program once novelty has waned?

HUMBY: Be relevant. If what they get through the letterbox each quarter is more interesting than anyone else’s, then we’re always up to date. It’s not rocket science.

PROMO: Tesco invested in the program conservatively in the beginning.

HUMBY: We’ve always had a test-in-line philosophy: Do it in a small way to show you can make some big changes. We can do something with a sample of 50,000 or 100,000 customers in five stores, show it’s going to work, and then ask what it will be worth if we replicate it in the whole customer base. That’s a constant process.

The only way to make [loyalty programs] pay for themselves is to make mass-marketing very niche. A lot of senior management commitment is to dollars, not customers. It’s the trade off between bucks this week and bucks next year.

PROMO: How is a startup different now?

HUMBY: The technology allows us to do stuff today that we only dreamt about then. The other thing is, we’re getting closer integration of the channels. The retail channel and the Internet and direct communications are coming together in a much more natural way now, so I can see what shoppers do online and in the store and I can understand the trade-off I get when you shop online instead of the store.

All the data in the world won’t solve your problem if you don’t recognize that people buying something on promotion is completely different from buying it not on promotion. You’d be amazed how many retailers don’t know if a product’s on promotion or not.

PROMO: How are you working with packaged goods companies in the U.S.?

HUMBY: When you see what can be done in Britain — and hopefully we can replicate the success in America with Kroger — the question for packaged goods becomes, “Where else in the world would you like to go to, and can we help you get there?”

PROMO: Does it take a retailer to do a loyalty program? Can a brand do its own?

HUMBY: I don’t think a consumer packaged goods brand can do a loyalty program the way a retailer can. Ultimately, it’s where you have the retail experience that really matters, and while consumer packaged goods brands may be stronger brands in America, they still don’t control the retail experience.

Packaged goods brands suffer [the problem] that any one of them represents such a small portion of consumers in a shopping basket.

It’s also about understanding what role the retailer plays as the medium, and what role as the retail outlet and what role as the loyalty partner. Retail is the biggest medium in the world. What other channel has your undivided attention for an hour a week, when you’re doing your grocery shopping? That’s one hell of a big media channel and if you actually got all of that to work as a cohesive force that helps the customer in a sensible way, it would be bigger than television.

PROMO: Can you do a loyalty program for any demographic?

HUMBY: Oh yes, definitely. We’ve got everything from lords and ladies down to people on a very limited budget and all of them like the mechanic, provided it’s relevant.

PROMO: What are you most proud of with Clubcard?

HUMBY: The fact that the average British consumer has one in their shopping basket and still uses it nine years on. It’s a program that most consumers look forward to getting their letters about. That’s a pretty rare thing these days.

PROMO: What do you like best about the work you’re doing now?

HUMBY: I still love the data. I love finding out something about consumers that no one else knows. Putting the insight into commercial terms and seeing their faces when they realize that if they were just to do this, it would make another $50 million. That’s great fun.

From Scoring Points

ON SEGMENTING SHOPPERS: The trick would be to create new customer groupings that were large enough to be cost-efficient to use, but with a richness of common interest to be truly meaningful. [We took] each product, and attached to it a series of appropriate attributes, describing what that product implicitly represented to Tesco customers. Then by scoring those attributes for each customer based on their consistent shopping behavior, and building those scores into an aggregate measurement per individual, a series of clusters should appear.

We set about imagining 50 things that our shopping baskets might say about customers. What does it mean if we buy a lot of ready meals? A lot of fresh produce? No meat?

Measuring customers on a number of these criteria could start to create distinct profiles. Such analysis would reveal distinct patterns in shopping, so it might be possible to identify the busy urban couple who shopped for ready meals but loved to be adventurous and spend a little extra. Or the health-conscious shopper who buys fresh fruit and vegetables, and avoids red meat but sometimes eats chicken.

ON GETTING THE BALL ROLLING: By creating 20 scales on which to judge the attributes of every product in the store, [we] could create 20 numerical measures. … But what scales to choose? “Low fat” against “high fat,” “big carton” against “small carton,” “needs preparation” against “ready to eat,” and “low price” against “high price” are just a few of the two-tailed Likert scales used. There were also single-tailed measures, such as “Is it a promotion?” and “Is it a major brand product?”

Judging 45,000 on 20 different scales would mean agreement on 1.2 million individual ratings before the segmentation could be used.

Some of the 1.2 million decisions would be easy to make: It’s not hard to decide if a product is “own-label” or not. Others were much harder. No one, as far as we know, had ever tried to distinguish how “adventurous” every product in a supermarket is. Tinned fish probably isn’t; extra virgin olive oil is. Is Brie adventurous? How adventurous is it? More than decaffeinated coffee? Less that a red pepper?

[We devised the Rolling Ball process] to allocate attributes for every item. [The team] started with a small set of products that definitely have the quality you seek. So if you want to find out which products are adventurous, start with extra virgin olive oil and ingredients for Malaysian curries, and see which customers bought those products.

Then look at what else these customers have in their shopping basket. Discard items that show up in everyone’s basket (bananas or milk, for example), and keep looking, building bigger and bigger groups of products. When can the process stop? This is where the rolling ball idea came in. The products that are picked up early will have a high “adventurous” rating. As the ball gets bigger, those ratings are probably lower, and certainly less reliable. … You might start off trying to predict adventurous products, but after 400 or 500 products are coded, you start to find a lot of products that are more “fresh” than “adventurous,” and so the ball has started to roll into an adjacent [category]. Eventually the rolling ball process produced an effective grading mechanism for all 20 attributes for every product, based on the contents of 10 million shoppers’ baskets. … [Tesco was] now able to use these 20 scores to create clusters of shoppers with shared motivations and preferences. … After six months, 13 well-defined and tested groups had been identified.

From “Scoring Points: How Tesco is Winning Customer Loyalty,” by Clive Humby & Terry Hunt with Tim Phillips. Published in the U.S. by Kogan Page Ltd., Sterling, VA.