Companies have used geographic information for years to help decide which ZIP codes to target in a mailing, which sites to pick for new stores and how to plan delivery routes. The geographic information systems (GIS) and tools to support those kinds of analyses are readily available and relatively mature.
But now the most advanced practitioners are integrating GIS with mainstream transaction-processing applications and databases, incorporating sophisticated location data in their business intelligence analytics and even tapping into pools of unstructured spatial information. And some companies are finding serendipitous uses of geographic information made possible by the marriage of BI and GIS.
Here's how four companies are pushing the edge of the BI/GIS envelope:
Site Selection at Staples Inc.
Staples plans to open 95 new stores this year after considering as many as 5,000 sites. Mistakes are costly -- closing a failed store can set the Framingham, Mass.-based company back $500,000 to $1 million.
The office supply retailer uses GIS tools from Tactician Corp. in Andover, Mass., combined with analytic tools from SAS Institute Inc. in Cary, N.C., to help it select store sites. The process all begins with a real estate model that forecasts weekly sales or potential sales by ZIP code. The forecasts drive activities such as site selection, budgeting, labor scheduling and marketing programs such as direct-mail campaigns, says Alan Gordon, director of sales forecasting at Staples, which now has GIS tools in a half-dozen departments.
The model considers some 30 factors that affect site selection, including obvious ones such as the presence of competitors and the demographics of the local population. "And there are things we put into our model that other people haven't learned of yet," Gordon says.
He says Staples hones its site-selection acumen by using SAS routines to correct and enhance the geographic data that it buys from external parties.
"The more we work in this area, the more we find problems and correct them," Gordon says. "We have explicitly tried to make that a competitive advantage."
For example, Gordon says, commercial databases of driving times between locations allow users to vary speeds by road type, but the databases don't take into account actual local traffic densities. Staples has written software that incorporates local conditions, so it knows how long it takes to drive from one ZIP code to another location through intervening ZIP codes of varying traffic density.
GIS and BI tool vendors are collaborating to integrate their products, so users don't have to. But the Tactician and SAS tools aren't yet integrated, and Staples passes files back and forth between the two companies' tools via FTP. But Gordon says Staples is building its own interface to allow both SAS and Tactician to access common DB2 or Oracle tables.
Targeted Ads at The Arizona Republic
The Phoenix newspaper has raised the insertion of advertising supplements to an art form. It can, for example, specify that copies of the paper going to any single ZIP code, street or circumference around an advertiser's store contain a particular advertiser's insert. Or an advertiser might specify that it wants inserts to go to just the 500 subscribers closest to each of its five stores. Or it might ask the paper to do a mailing to certain nonsubscribers near its stores.
A seller of swimming pool supplies, for example, might ask for its ad circular to go just to houses near his store that have pools, says Karen Parrilla, a principal analyst at the newspaper. "Or we might say that this is a high-end pool maintenance company, so we'll only hit pool owners with a household income of $175,000-plus. It's very powerful. Before, it was mostly at just a ZIP code level," she says.
The newspaper uses ArcGIS tools from Environmental Systems Research Institute Inc. in Redlands, Calif., to load, manage and query a spatial database. It also uses an ESRI product for delivering geographic information to end users as a Web service.
The key application, called Market Focused II, imports subscriber information from the circulation system and creates new carrier routes every day and then displays them on a map. Sales reps can use the maps to, for example, suggest to a grocery chain where it might best target its inserts. The results are sent electronically back to the circulation system and to the advertising order-entry system.
These kinds of analyses are computationally arduous, Parrilla warns. Market Focused originally ran on desktop PCs, and a complicated query -- examining a three-mile radius around each of 85 stores, for example -- could run for more than eight hours. Now the system is hosted on a server and accessed by a Web interface, and a user can run the same query in 35 minutes, she says.
Beer and BI at Hensley & Co.
The Phoenix-based beer distributor has had such success with location-enhanced BI that it plans to offer the information to its customers, along with the brew.
Hensley has for some years used Margin Minder from Salient Corp. in Horseheads, N.Y., to analyze sales and profit margins against variables such as in-store signage, displays, shelf-space utilization and special promotions. These analyses include data on competitors, as well as supply chain data such as inventory levels and delivery costs.
Recently, Hensley installed a new Salient product called Geo Minder and will use it to answer questions such as, How many customers are in this area, and what are the delivery costs and margins for that area? What customers are near a baseball field, and how many sales are there on game days? How do prices in this neighborhood compare with those of competitors?
Mark Miller, Hensley's sales vice president, said the company will share the insights gleaned from the system with its customers and will be able to offer stores advice about what to buy and how to price it to beat local competition.
"We could say to a convenience store, 'You're averaging $2 a case [margin], but within five miles the average is $1.80 per case. But their case [volume] is 20% higher and they are taking more money to the bank. So you are not being competitive enough to bring customers into your store,'" says Miller.
"If we can provide better information to our customers to help them, that will keep us a step ahead," says Karla Dooley, a category analyst at Hensley. "I don't think anyone else is doing this kind of geographic stuff."
Structuring Data at an Oil Company
A Fortune 100 oil company has been doing conventional BI and GIS analysis on structured data for years. "The thing that's more of a challenge is the information that's embedded in unstructured information -- in work products such as Word documents, spreadsheets, PDF documents and so on -- on servers used by workgroups," says a manager at the company, who asked not to be identified.
He estimates that his company has 150TB of such information, "and it's growing rapidly." In fact, there's more of it than conventional structured data, he says.
The company gets at the unstructured data using GeoTagger GIS tools from MetaCarta Inc. GeoTagger trolls though documents, identifies location-specific names and tags them with standard geographic references such as latitude and longitude. Then the oil company can produce a map of an oil field that displays wells with the documents associated with each, even if the well has been named and described inconsistently across systems and over time.
"We can search across multiple repositories that come from different sources," says the manager. "That could be technical memos written by our R&D labs, external databases or Web content. Being able to search all those things at the same time allows you to [get] geographical information from across the world and filter it down to your specific area of interest."
The oil company also found that GeoTagger could help it bridge its own data with data at a company it acquired. "It can crawl each companies' information independently, regardless of its structure, and use that as a unifying framework," he says. "Then you may not have to do transformations into standard things, you can leave it unstructured."
The Future
Claudine Bianchi, marketing vice president at MetaCarta, says companies will find gems when they extend their data mining to the unstructured information that makes up 80% of all corporate data, most of which contains geographic information.
More subtly and perhaps more profoundly, Bianchi says BI/GIS systems will eventually be refined and extended to make sense of data not normally thought of as geographical - the locations of electrical plugs in a facilities management system, for example.
"Even the human genome," she says. "Every gene is location-oriented, and you could have a data module around each thing in the genome."
Meanwhile David Sonnen, a consultant at Framingham, Mass.-based research company IDC, has a more somber and down-to-earth prediction. "Data quality will be the Achilles' heel for the industry," he says, "especially as we start to see spatial data integrated with other data."
Correctly matching location data from different sources -- street and building data from external sources with customer addresses from an internal system, for example -- can be tricky. Worse, detecting errors can be extremely difficult, never mind correcting them, says Sonnen.
Staples' Gordon seems to know that already. He warns, "There's a limit to what you can do with GIS technology and how much you want to trust maps."
|
Smarter BI
Stories in this report:
- Editor's Note: Smarter BI
- Better BI Decisions
- Tame the BI Jungle
- Beyond ZIP Codes
- BI Backup Breach
- Quickstudy: Enterprise Information Integration (EII)
- Where's BI?
- Executive Bulletin: BI and the Data Matrix
- Sorting Out the BI Players
- The Future of BI
- Data Points: BI
- BI Resources