Customer insight specialist Dunnhumby has taken its existing service, which uses data from Tesco's Clubcard, to the next step. It has launched a tool to accurately predict future sales of potential new products, which can track even the lowest volume products that are launched.
Dunnhumby's traditional service analyses the data of 10% of its Clubcard holders, equating to a sample of around one million people.
In a typical week it gives information on 600,000 customers making one million visits to the group's stores and buying 18.2 million individual items.
This has helped Tesco and its suppliers tailor ranges more appropriately, as Sarah Archer, wine customer manager for Tesco, says: "The Clubcard data is central to the wine category in that it tells us about all the customers' likes and dislikes.
"This enables us to tailor the breadth and choice of our range to meet their needs. It is an integral part of the way we work nowadays."
Indeed, so beneficial has this data been to the business that Dunnhumby is now taking its skills to other, non-competitive, retailers. This week it signed a deal with US grocery retailer Kroger and The Grocer understands it is poised to sign a similar deal with an Australian grocer.
Weekly benchmark
The new tool builds on this to provide a weekly benchmark for any new product against all launches in that category over the past three years. These figures can be provided three weeks after launch and then on a weekly basis.
The real benefit of the service is the accurate prediction of future sales. It enables companies to predict sales six months to one year ahead.
It uses the initial growth curve of a new product and compares this with similar launches in the same category, then takes an average of these historical launches to predict sales of the new product.
Using this technique Dunnhumby claims to have improved its predictive results from +/-3-5% to nearer +/-1% accuracy, and this, it says, is set to improve as it includes further variables in its calculations.
A key benefit of using the Tesco sample is its size. It is this that enables Tesco and Dunnhumby to track even the lowest volume products that are launched. Even a product with a customer penetration figure (the percentage of people who buy that product against the others in its category) of just 1% would be bought by as many as 6,000 customers a week within its million-strong sample.
Andy Hill, product and services director at Dunnhumby, claims tracking these low volume products is a problem for the panels which derive data from customers at home, scanning barcodes of all the products they buy. These tend to have too small a sample.
"If it is too small then the panel simply won't track it," he says.
For these products, panels will only use a raw data sample and will not weight it for the whole of the UK.
This adds some unpredictability to the results which is a deficiency, says Paul Newberry, sales and marketing director at Stream Foods, which produces dried fruit snacks and has a modest turnover of £5m per year. "With such a small sample, if Mrs Jones [who buys the product] goes on holiday, our sales could drop by as much as 10%, according to panel data," he says. In order to get a robust set of data from a consumer panel it may take a small manufacturer up to half a year, according to Newberry.
Where ACNielsen, for example, delivers data from its New Product Launch Monitor evaluation tool at eight, 16 and 24 weeks, Dunnhumby data gives Newberry a "robust base in four to five weeks".
Advantages from predictive tools
It is therefore able to quickly spot the flops and then either de-list them or take positive action such as changing the marketing mix, or the ingredients if it is a foodstuff.
Conversely, a success can be quickly followed by wider distribution.
Hill says: "A retailer might have chopped a product if sales were not high enough but, if the early data had shown that it had some regular purchasers, it might instead have given it less space or only put it in certain stores."
These regular purchases show up in the all-important repeat-purchase figures, which are a true indicator of whether a product is performing well rather than being something that is only ever bought once.
Dunnhumby believes there are undoubted advantages to be had from the use of predictive tools.
Hill says: "These predictions are far greater, faster and more accurate than have been available before. They will give manufacturers data that they can act on. If this means they can reduce the number of failed launches to only eight out of 10, it will make a big difference."
Newberry believes that with accurate forecasts on a new launch he could confidently gear up the company accordingly. "We could get real commercial benefit from this as we could decide what is needed in terms of marketing and machinery and have the confidence to go to the bank and borrow money to do a marketing campaign."
Localised launches
The Dunnhumby data also gives manufacturers the ability to perform localised launches.
Manufacturers could launch a product in a variety of ways in different types of stores or different areas.
The number of stores used to launch a product could also be reduced from the average of 50-100 in Tesco to as few as a single outlet, according to Hill.
In this case, rather than relying on the million-strong sample, it would use the entire base of Clubcard holders who shop in that particular store.
However, for Mark Baillache, partner and head of the food sector at KPMG, big is not necessarily best. "It is a large sample but it is not necessarily representative."
In contrast, the TNS and ACNielsen panels accrue data from across all the supermarkets, he points out.
And Chris Morgan, customer relations director at Cadbury Trebor Bassett, says the wait for numbers from the panels is not that great a problem for larger manufacturers. When the likes of Cadbury launches a mainstream product, it can take a full four months to fully distribute its first batch.
This is because the first is likely to be 750,000 cases of either 36 or 48 units, and by that time it will start to receive detailed numbers from TNS. "The stats might be small but they are fairly robust for us," says Morgan.
Whichever launch method is chosen, Baillache advises manufacturers to use all forms of data available to them. He suggests the "broad view" data from the panels which is good at highlighting trends should be combined with the more detailed data from Dunnhumby.
He says: "If you are willing to invest 1% or 2% of your sales in research and development, you can go that extra step and put confidence behind your launches."
{{ANALYSIS }}
Dunnhumby's traditional service analyses the data of 10% of its Clubcard holders, equating to a sample of around one million people.
In a typical week it gives information on 600,000 customers making one million visits to the group's stores and buying 18.2 million individual items.
This has helped Tesco and its suppliers tailor ranges more appropriately, as Sarah Archer, wine customer manager for Tesco, says: "The Clubcard data is central to the wine category in that it tells us about all the customers' likes and dislikes.
"This enables us to tailor the breadth and choice of our range to meet their needs. It is an integral part of the way we work nowadays."
Indeed, so beneficial has this data been to the business that Dunnhumby is now taking its skills to other, non-competitive, retailers. This week it signed a deal with US grocery retailer Kroger and The Grocer understands it is poised to sign a similar deal with an Australian grocer.
Weekly benchmark
The new tool builds on this to provide a weekly benchmark for any new product against all launches in that category over the past three years. These figures can be provided three weeks after launch and then on a weekly basis.
The real benefit of the service is the accurate prediction of future sales. It enables companies to predict sales six months to one year ahead.
It uses the initial growth curve of a new product and compares this with similar launches in the same category, then takes an average of these historical launches to predict sales of the new product.
Using this technique Dunnhumby claims to have improved its predictive results from +/-3-5% to nearer +/-1% accuracy, and this, it says, is set to improve as it includes further variables in its calculations.
A key benefit of using the Tesco sample is its size. It is this that enables Tesco and Dunnhumby to track even the lowest volume products that are launched. Even a product with a customer penetration figure (the percentage of people who buy that product against the others in its category) of just 1% would be bought by as many as 6,000 customers a week within its million-strong sample.
Andy Hill, product and services director at Dunnhumby, claims tracking these low volume products is a problem for the panels which derive data from customers at home, scanning barcodes of all the products they buy. These tend to have too small a sample.
"If it is too small then the panel simply won't track it," he says.
For these products, panels will only use a raw data sample and will not weight it for the whole of the UK.
This adds some unpredictability to the results which is a deficiency, says Paul Newberry, sales and marketing director at Stream Foods, which produces dried fruit snacks and has a modest turnover of £5m per year. "With such a small sample, if Mrs Jones [who buys the product] goes on holiday, our sales could drop by as much as 10%, according to panel data," he says. In order to get a robust set of data from a consumer panel it may take a small manufacturer up to half a year, according to Newberry.
Where ACNielsen, for example, delivers data from its New Product Launch Monitor evaluation tool at eight, 16 and 24 weeks, Dunnhumby data gives Newberry a "robust base in four to five weeks".
Advantages from predictive tools
It is therefore able to quickly spot the flops and then either de-list them or take positive action such as changing the marketing mix, or the ingredients if it is a foodstuff.
Conversely, a success can be quickly followed by wider distribution.
Hill says: "A retailer might have chopped a product if sales were not high enough but, if the early data had shown that it had some regular purchasers, it might instead have given it less space or only put it in certain stores."
These regular purchases show up in the all-important repeat-purchase figures, which are a true indicator of whether a product is performing well rather than being something that is only ever bought once.
Dunnhumby believes there are undoubted advantages to be had from the use of predictive tools.
Hill says: "These predictions are far greater, faster and more accurate than have been available before. They will give manufacturers data that they can act on. If this means they can reduce the number of failed launches to only eight out of 10, it will make a big difference."
Newberry believes that with accurate forecasts on a new launch he could confidently gear up the company accordingly. "We could get real commercial benefit from this as we could decide what is needed in terms of marketing and machinery and have the confidence to go to the bank and borrow money to do a marketing campaign."
Localised launches
The Dunnhumby data also gives manufacturers the ability to perform localised launches.
Manufacturers could launch a product in a variety of ways in different types of stores or different areas.
The number of stores used to launch a product could also be reduced from the average of 50-100 in Tesco to as few as a single outlet, according to Hill.
In this case, rather than relying on the million-strong sample, it would use the entire base of Clubcard holders who shop in that particular store.
However, for Mark Baillache, partner and head of the food sector at KPMG, big is not necessarily best. "It is a large sample but it is not necessarily representative."
In contrast, the TNS and ACNielsen panels accrue data from across all the supermarkets, he points out.
And Chris Morgan, customer relations director at Cadbury Trebor Bassett, says the wait for numbers from the panels is not that great a problem for larger manufacturers. When the likes of Cadbury launches a mainstream product, it can take a full four months to fully distribute its first batch.
This is because the first is likely to be 750,000 cases of either 36 or 48 units, and by that time it will start to receive detailed numbers from TNS. "The stats might be small but they are fairly robust for us," says Morgan.
Whichever launch method is chosen, Baillache advises manufacturers to use all forms of data available to them. He suggests the "broad view" data from the panels which is good at highlighting trends should be combined with the more detailed data from Dunnhumby.
He says: "If you are willing to invest 1% or 2% of your sales in research and development, you can go that extra step and put confidence behind your launches."
{{ANALYSIS }}
No comments yet