Retailers have long recognised that geography is one of the most important factors in their business. The days of “if you build it, they will come” are long gone. Now the cutting edge is identifying where there is a sufficiently large catchment area to support a new outlet without cannibalising existing stores or being prone to excessive competition.
Added to store planning, the trend towards Efficient Consumer Response (ECR) has compounded the need to understand geographical issues. Aligning stocking with demand levels in order to manage distribution, ordering and inventory better can provide real cost benefits and potential competitive edge. But it does require the ability to marry up information about location with data about consumption patterns and sales.
Geographical information systems (GIS) have long been a mainstay of property portfolio man agement. The challenge now is to link these systems with marketing databases and use the total picture provided to support management decisions.
Tesco is an example of a retailer which has migrated to GIS and is now looking beyond purely site issues to wider business solutions. Before implementing a Smallworld GIS, its site research team worked with paper maps and grids, plus text-based systems, to predict where the best locations would be. “Although we did use computer models, we would end up with a map covered with sticky dots, which was not the best way to provide an accurate analysis,” says Chris Parkinson, computer division project manager at Tesco.
With the help of GIS, the store chain has been able to layer up data and run more complex analyses. The geographical building block is Ordnance Survey maps, to create a real-life picture of drive-times to stores. This is augmented with Department of Transport information on the number of lanes in a road, typical traffic flows and speed limits to improve accuracy.
“We also include information of our own, such as bad junctions and whether streets are one-way,” says Parkinson. Additional layers of data have been drawn from Experian, on how customers use shopping centres, while the drive-time model is also linked to geodemographic and lifestyle information.
“We can look at a week’s shopping patterns and illustrate it as a ‘hedgehog’ diagram around different stores. It’s much better than working it all out from a list of percentages, which is the same information, but presented in a non-graphical way,” says Parkinson. What Tesco is now looking towards is adding in Club Card data to improve the accuracy of its customer behaviour models.
It is the arrival of such large volumes of data on customer activity which some GIS specialists believe is bringing about a major change in the types of system which are appropriate to retail. In particular, high-end technologies which allow for complex modelling need to be adopted in order to ensure that decisions are only made on the latest data available.
“The biggest single change results from a long-term rise in data volumes. People in retail are relying on returns from EPOS and loyalty cards. The volumes of data every week are huge. The thing you need to be able to do is model data more accurately. That is where object-oriented applications come into their own,” says Dr Tim Hartnell, director, business development, at Laser-Scan.
With non-object oriented GIS, the drivetime model is static, based on whatever data was available at the time it was created. That means if something changes, this model then has to be taken out of the system and reworked. “The key to object-oriented systems is that the objects you define don’t just have static properties, but also behaviour which brings them to life. The objects look after themselves,” he says.
An example of where this could have a real benefit for a retailer is if the physical geography is changed. When a new by-pass is built, for example, drivetimes and gravity models for an outlet in that area will all be affected. With an object-oriented system like Laser-Scan’s, the new geography can be inputted, automatically updating the model, without requiring fresh external modelling.
“So each time you make a change, for example, if a competitor opens a store, if you want to model that situation, you don’t have to change all of the data, you just change the object,” adds Hartnell. Modelling local conditions using a different gravity model, for example, can be done without altering the assumptions used in the rest of the system.
Laser-Scan is the first vendor of GIS to offer this technology, which it claims is far superior to other desktop packages which are available. Not surprisingly, its competitors disagree that retailers need object-based systems. “It is too easy to say that current desktop mapping tools are not capable of engaging with complex business scenarios in retail and distrib-ution,” says Steve Morland, technical co-ordinator at Tactician UK.
He believes any problems with the way GIS is used might reside in factors other than the software’s functionality. The first of these is the collection and analysis of data in a spatial environment. Because this differs from conventional techniques, what data is collected and what type of analyses can be run may not have been clearly defined. “Often GIS packages are purchased to deliver solutions to problems that have not been thought through properly,” he says.
This can result in a system being swamped with data, or an insufficiently robust analytical engine having been specified. But the system alone is not a panacea – Morland estimates that 80 per cent of Tactician’s users solve 80 per cent of their marketing problems using just 20 per cent of the system’s functionality.
There is greater complexity in what is being demanded of GIS, however, and the problem is that the majority of mapping software has not been developed from a marketing perspective, nor has marketing thinking been attuned to the benefits of spatial analysis.
Whitbread is an example of a Tactician user which has progressed from site planning and analysis into wider marketing applications. Previous direct mail campaigns had proven effective in generating traffic in the highly competitive pub restaurant market. As a result, a propensity model was built of the likelihood of a customer visiting a Whitbread outlet. It took into account distance from the outlet and available competition. The resulting gravity model was overlaid on each outlet and applied to the 1.6 million UK postcodes, leading to improved targeting and more effective direct marketing.
The cost of acquiring an object-oriented GIS is likely to be a deterrent to many, even if they do have the data and capability to handle the outcomes it will produce. But according to one vendor, it might not be necessary to buy in to top-of-the-range software to get full functionality.
GMSL has developed Pro Travel and Pro Territory based on the ArcView GIS produced by GIS specialist ESRI. Two recent extension modules have considerably increased its capabilities. “Spatial Analyst and Network Analyst, since they were launched in 1996, have made a big difference to the sort of work you can do,” says director Jeanne Knowles.
A key difference of Pro Travel is that it calculates drivetimes from a grid, which can have a resolution of anything between one metre and one kilometre, rather than the link and node approach in most desktop mapping. This makes it much more accurate. Pro Territory adds in the ability to include impedances to how a network is planned, such as the density of customers in an area, the average length of a visit and how many sales people are available.
“We did some work for a food distributor which had five regional managers, each with eight sales people. It wanted to put that sales force into its business plans, looking at the density of customers, the call frequency and the fact that each sales person has two or three people who look after displays, stock, and so on. All of that can be used in this system on the desktop,” says Knowles.
But while the trend in GIS may be towards greater complexity of models, within ECR, benefits can be derived from much simpler approaches. Modelling customer behaviour data and overlaying this against postcode can yield an uplift in marketing results. “With a number of manufacturers, we have done supply and demand modelling based on a profile of who buys what type of product and who shops in which stores. Lifestyle and lifestage are also added,” says Chris Morris, database marketing director at Claritas.
Hotpoint used this approach for a trade marketing campaign through Northern Electric outlets last year. It wanted to align in-store promotional support for its “hero” products with local demand. The problem was that using Prizm, the classification system based on Claritas’ lifestyle database, did not provide enough discrimination.
Instead, a grid was drawn up of Prizm clusters within axes of lifestyle and lifestage, producing 19 segments. Lifestyle was driven by income, and lifestage by the combination of age and presence of children. High indices of Hotpoint buyers were looked for within this grid and then compared with which retail outlets they shopped at.
“We ended up with one chart showing a comparison of the competition, potential customers and TV region,” says Morris. From this process, Hotpoint picked which products to promote in each region. The impact of the profiling was a campaign which was five times more successful than any previous activity.
If simplicity can show such returns, then complexity should blow the roof off. For retailers especially, deciding which way to go may be driven as much by how many users and departments will be involved with a system as by what queries need to be run. Some people believe a complex answer can only come from a complex system – others disagree. Retailers may all have to work with the same geography, but they do not have to use the same GIS.
Digital Mapping Show Preview
The Digital Mapping Show and its GIS Seminars take place at the Novotel in Hammersmith, London, from May 12-13 and styles itself the UK’s only specialist desktop mapping for business event.
The Show was launched three years ago and although it was initially aimed at a London and South-east audience, visitors have pre-registered from Belgium, Ireland, Scotland and Wales.
New products being launched at the Show include the latest version of Beacon Dodsworth’s Prospex market analysis GIS; AA Kingswood will be showing its new version 3.6 Geo Concept information system for Windows; and Dataview Solutions will be demonstrating its latest RouteViewPro, Arrow and RiskView strategic mapping and routing tools.
Each day of the exhibition and its accompanying seminar is designed to attract delegates with specific interests. The first day, of the event is principally aimed at those interested in GIS for marketing purposes; the second day is targeted at executives from the utilities, local and central government. Speakers will be from both the vendor and user community.
The keynote presentation on the first day will be given by Peter Sleight of the Target Market Consultancy.