"If you don't know your customers as individuals, it's like trying to play the piano in boxing gloves." So said Terry Hunt, co-founder of Tesco's direct-marketing firm Evans Hunt Scott (now EHS Brann), during the first Clubcard trials in 1993. "You make a lot of noise, but it's not always pretty."
Back then, the standard retail technique was to send out untargeted promotions, hoping former customers got them. Without the name and address of the Clubcard member or details of their household, the scheme was no more useful than Tesco's daily till-roll and no more motivating to customers than a discount card.
But with these details, Tesco could accumulate the rewards earned by a customer and send them directly in the form of money-off vouchers.
That was the start. From the first days of the trial, Grant Harrison -- a New Zealander chosen by then marketing operations director Tim Mason to head the development of Clubcard -- realised the value of the data he was collecting. He knew that unless it was used to change the way Tesco did business, Clubcard wouldn't succeed.
"Do the stores want to run differently? Do the retail directors? Do the buyers?" asked Harrison. "We know you can find interesting stuff in the data. The challenge is how you get the business to engage and be prepared to change its processes or decisions, based on a more detailed source of customer understanding."
Using data to run the stores better was an admirable idea, but there was one major practical problem: when Clubcard launched in February 1995, there was too much data. The comparatively simple task of transmitting data between Tesco's IT department (where it was held to run the points-accounting database) and data consultancy dunnhumby (where it was analysed for marketing and business information use) took 30 hours using the highest bandwidth then available.
As owner of one of the most complex customer databases in the world, Tesco handled the output of more than 50 million shopping trips in the first three months, comprising more than two billion items from the shopping of more than five million Clubcard members. The challenge of analysing every purchase was beyond the technology of the time. Whereas Tesco launched Clubcard nationally 'because we can', it didn't try to collect and analyse all the data for the opposite reason: 'because we can't'.
Today, Tesco has the capacity to take information from every shopping basket processed by its checkouts and use it to make marketing and management decisions. But in 1995 the idea of breaking down the logs from every till on a product by product basis and analysing every scrap of information from every transaction was a fantasy. The experts' advice to Tesco was don't even try it.
The supermarket was willing to let dunnhumby - in which it took a 53% stake two years ago - take the lead in creating the strategy for data analysis. Tesco had never looked at mass volumes of individual customer data as it didn't have the skills in-house to do the job. As Mason points out, the people who can successfully analyse customer loyalty data to create useful insight weren't readily available in 1995. "It took us two years to identify the sort of people who were good at analysing Clubcard data. You have to use intuition and creativity as well as statistical know-how, and identify the right things to test."
In effect, he is talking about analysing data as grocers analyse their business and improve by experience. From Jack Cohen to Sir Terry Leahy, grocers have used their instincts, backed hunches, stayed close to customers and responded swiftly to what they need. Extending this approach to the use of data was natural to Tesco.
Trying to achieve perfection with transactional data is almost impossible; far better to have a good idea based on experience and instinct, and then go looking for the data to prove it, or at least strongly support it.
The first realisation was to not struggle to make sense of all the data on every customer, but to use a matrix of data samples to gain a statistically valid picture of behaviour. Analysts took 10% of the data once a week, processed it and applied the findings back to the other 90%. The plan was to use intellect to speed learning and use the insights as soon as possible.
Tesco's managers were happy with this approach. At the time, Dunnhumby employed 25 people and Tesco agreed to underwrite the cost of the hardware needed to deliver the analysis at the speed required. Investing in the computers to handle 10% of the activity once a week suited Tesco's thrifty instinct.
Making secure assumptions
One of the quirks of statistics is that not every record needs to be analysed to be able to make reasonably secure assumptions. If there is a trend to observe, analysing a small set of data offers a result with 90% certainty.To achieve 100%, every piece of data has to be analysed, and that is bought at a disproportionate cost.
Proof that Tesco recognised these analytical skills as critical to its business future arrived in 1997 when it set up the Customer Insight Unit.
The idea was to combine the skills of its site research team, who made proposals for store development based on local information, with skills from the marketing and commercial side.
But no transactional data can ever be perfect. There will always be errors or unknown factors to upset analysis.
For example, card holders move address, one card may have multiple users displaying different buying patterns, or someone may open a new account on a frequent basis because they constantly lose their card.
Results can also be distorted by unique occurrences or local effects; a store may run out of stock of a key product, for example. And, taking a broader view, if Clubcard usage only identifies 60% to 75% of shoppers at any one store, then 25% to 40% of customer behaviour is not traceable and individual data isn't being collected.
In developing Tesco's strategy, the data analysts were backing an educated hunch. Having worked with geodemographic data for years, they knew its power and accuracy were very limited, and over-use by marketers had reduced its efficiency. The major geodemographic databases had been compiled and marketed for years with thousands of brands using them to target offers as accurately as possible. The trouble with a geodemographic model of customer targeting is that it is based on the premise of you are where you live. It's clearly not true - who is exactly the same as their neighbour?
The Tesco model, by contrast, is based on individual particularities: you are what you buy. Rather than forcing a profile onto households using generalised data from their postcode, the customer behavioural approach starts with their distinctive actions. It creates like-minded groups of people with similar tastes and activities, according to their purchases.
When Safeway abandoned its ABC card, it said that trying to generate useful insights from the data was "like drinking from a fire hose". It was the same for Tesco, but it siphoned off what it could deal with at any one time and got on with selling to customers.
Dynamic data store
Working with dunnhumby, Tesco learned as it went along. Instead of building the biggest data store it could, it aimed to build the smallest store of data that would give useful information. It was a useable resource to understand better what customers did and a predictor of what they might do in the future.
Where Tesco started from was not asking what it would like to do, but what it could realistically do and whether it would make a profit. It reasoned that any new information was progress and built from there. It didn't agonise over information it couldn't extract at first or delay the process until it could get it. It built on each discovery, identifying its most profitable customer profile and driving later research in useful directions. While its early systems were sometimes awkwardly stitched together, they formed the basis for better techniques.
Clubcard wasn't about observing trends passively, it was a massive laboratory of customer behaviour. Retailers had always experimented with prices and ranges to see what worked, but now Tesco could measure exactly what worked in any store. When it was doing something wrong, it knew in days. When it was doing something right, it could implement it nationwide in weeks. Tesco could quickly identify lapsed shoppers or those who ignored certain departments, and incentivise them. It found ways to defend itself against rivals.
Segmenting users enabled Tesco to target its vouchers and coupons at people who really wanted them. Redemption rates for coupons varied from 3% (the sort of rate Tesco expected from a pre-Clubcard, untargeted mailing) to 70%. But, as it became better at modelling customer behaviour, the target reached an unprecedented 20% within a year.
In 1996, Tesco crossed a bridge. Previously it had used data when it was possible and available, but now it had data for almost every occasion. However, it didn't have the means to compare who drank Coke against who drank Pepsi, for example, so it turned to bigger, broader questions. Even at this stage, Clubcard data had been an influential adviser for every major Tesco initiative.
Richard Brasher, the then operations marketing director who is today director in charge of all non-food business, began five years of work with his Clubcard team. In two years since launch, Tesco had successfully recruited most of its regular customers to Clubcard - about £4 out of every £5 spent in Tesco came from Clubcard holders. Tesco had refined the scheme to maintain public interest and customer participation. It continued to create excitement and controversy in the press, and had lifted sales on a tactical basis.
Four times a year, the statement mailing recreated the Christmas sales effect and coupon redemptions hit levels Tesco barely dreamed of before Clubcard. To build on this momentum meant a new strategy. It was time to start serious work on the data.
Behavioural breakdown
Already, a lot of progress had been made establishing a platform to manage and harvest customer data. Customers were segmented by life-stage and sent specific offers, messages and magazines. The sales potential was becoming clearer, thanks to the basic behavioural categorisation work by recency, frequency and value. Customers could be identified by the store they used, down to the particular departments in which they shopped.
But by early 1997, Tesco's satisfaction at a job well done was giving way to frustration. It knew that, as a corporate asset, the data promised more. With Sainsbury's finally responding with its Reward card, the pressure was on to maintain the massive lead Clubcard had achieved.
Tesco needed to make fundamental business decisions, which could be made with greater confidence if the data could be made to reveal more of real customer behaviour.
As the years passed, Clubcard data analysis was able to describe groups of customers to whom staff in the stores could relate. It had peered into consumers' shopping baskets and made sense of what it saw.
TIMELINE - TESCO CLUBCARD
1993: A £1m research budget is earmarked for the Clubcard project. Appoints direct marketing agency Evans Hunt Scott (now EHS Brann) in May. Clubcard is tested at stores in Dartford, Sidcup and Wisbech in November, with take-up of 50%; 11 more stores slated for trials.
1994: In the three test stores, £6 out of every £10 is spent by Clubcard holders. In November, a Tesco spokesperson talks of "the potential to operate the scheme on a nationwide basis", but launch is near.
1995: On 13 February, Clubcard launches amid media frenzy. Two weeks on, seven million cards hit stores and supplies are nearly exhausted. Within days, more than 70% of all sales are matched to card holders. Market share surpasses Sainsbury's for first time and stays ahead. In October, Safeway unveils ABC card, backed by a £7m ad campaign, only to withdraw it five years later.
1996: In May, the first Clubcard magazine is mailed in five versions (young adults, students, families, older adults and the over-60s) according to the information shoppers gave when they signed up.
1997: Tesco Personal Finance launches on the back of the Clubcard data and proposition. By 2003, it has 2.5 million customers and £96m profits, and it has also lent £1bn in personal loans; 500,000 customers insure their cars and some 250,000 their pets with it. In August, Sainsbury's launches its Reward scheme.
1998: Datamonitor reports that UK retailers spend £400m a year on loyalty schemes.
1999: Tesco unveils Clubcard Deals. Card holders can spend their points with partners such as KLM, EuroDisney and Virgin Trains.
2000: Tesco Direct becomes Tesco.com. It can track online performance via Clubcard and target those likely to shop from home.
2001: Tesco takes 53% stake in dunnhumby, the data consultancy. A new subsidiary, dunnhumby Retail, gives manufacturers access to Clubcard information about anonymous consumers' buying habits.
2002: Air Miles, until now a partner of Sainsbury's Reward programme, becomes a partner of Tesco's Clubcard Deals scheme. In September, Loyalty Management UK launches Nectar, attracting brands such as Sainsbury's, Debenhams, BP, Vodafone and Ford. The scheme has more than 11 million collectors.
2003: Clive Humby and Terry Hunt celebrate ten years of Clubcard with Scoring Points.
READER OFFER
You can read the rest of the Tesco Clubcard story in Scoring Points: How Tesco is Winning Customer Loyalty, a book co-authored by Clive Humby and Terry Hunt, with Tim Philips. Marketing is offering readers the chance to pre-order the book, published October 16, to get a 20% discount on the retail price of £25. Readers should contact Littlehampton Book Services, distributors for Kogan Page, and quote reference MF32 in order to obtain the discount (p&p not included). Tel: 01903 828 503, or e-mail mailorders@lbsltd.co.uk. The offer is valid until December 31 2003.
.jpg)
.jpg)