Location Intelligence Conference : Location Technology & Business Intelligence

ESRI Business GeoInfo Summit

by Hal Reid
April 21, 2005

Business GIS is moving beyond the time honored applications such as site selection and target marketing, and into the arena of the visualization of Business Intelligence (BI).GIS is now a mainstream technology, and since everything happens somewhere, mapping software is obviously the right choice for the visualization of "where."

The tone for this conference, held in Chicago on April 18 and 19, was set by ESRI's David Huffman, director of commercial sales.He stated that the goals for the conference were to provide education, opportunities for interaction and demonstrations - but that the real purpose was to expand the use of GIS in business.Business GIS (BG) is the fastest growing sector within ESRI, he said, in part because of ESRI's integration with SAP, Oracle, SQL Server, DB2 and ERP (enterprise resource planning) systems.

Dave Busser, ESRI's director of corporate business development, and Steve Benner, director of strategic accounts, explained the scope of ESRI's offerings and strategic directions.But I thought the most interesting presentation about this shift of BG into the BI space was by Michael Gonzales, president of The Focus Group, a consulting firm that specialized in BI-related techniques and technologies.

Gonzales began by identifying several key issues.

BI holds a lot of promise - "The ability to push up actionable insight and knowledge" - but there's a huge gap from promise to reality
The shift in terminology from Data Warehousing to BI has actually meant building more disparate data structures
There's a certain momentum to continue with fragmented, disjointed systems, neither dynamic nor spatial.

Why the gap BI from the promise to reality? According to Gonzales, the systems are time-based, not spatial, so while the when could be determined, the where could not.

He continued by explaining that a successful BI system was would have these characteristics:

Offer seamless integration of technology (spatial and temporal)
Have simplified data delivery
Zero latency in the analytics.

Major database vendors are pushing the integration of these technology components:

ELT technology (extract, load, transform)
Spatial data
OLAP technology (online analytical processing)
Data mining technology
Portal support

This indicates that the database vendors may be closer to the truth of how to close the BIG gap goal than the BI vendors. Gonzales indicated that in his opinion, the technology was there, but what was needed was overcome the problem that BI is blind to anything outside the data space.The primary issues in overcoming the problems are fairly straight-forward.Gonzales made these points.

We think visually, therefore we need to see our logic
Maps are a technique for organizing and embedding knowledge that is easily understood
Spatial provides the "who" by binding new information content and enables the analysis of "where".

This presentation's key point was "without spatial visualization, we are only reporting the obvious."

James Akright of General Motors Powertain Division offered a practical application of BG and BI.General Motors has created a system to look at patterns of incidences in their vehicles based on a set of groupings, which were:

Fuel system analysis
Weather analysis
Manufacturing
Driving habits
Electrical system analysis
Fleet tracking
Warranty analysis

They tracked vehicle problems and looked at patterns within each grouping.For example, there was a spark plug problem that appeared in one Canadian province.Spark plugs were failing after about 5,000 mile in new vehicles.Their system found a pattern of failure and established a geographic framework for these incidences.Investigation found that it was a fuel supply issue in that market that was affecting the durability of the spark plugs.The solution was a recall to upgrade the spark plugs from a standard $.97 part to a $5 part that would allow for the reliable use of this local fuel.

The most interesting use of the General Motors BG/BI system was the link to OnStar.They could track vehicle conditions in those GM-owned vehicles that subscribed to a datalink service.Problems could be diagnosed automatically and in some cases, problems could be corrected by a download or a recalibration of electronic systems remotely. Customer notification of imminent failure could be implemented before the vehicle was disabled by the impending part or system failure.

Perhaps the most interesting from historic perspective was the session on modeling with Dr.David Huff and Dr.Tony Lea, as people of this calibre are rarely in the same presentation.These are two of the icons in the field of modeling the performance of retail locations, market planning, cannibalization and network optimization.

Dr. Tony Lea and Dr.David Huff spoke on modeling at the conference.

Dr.Huffs' presentation was purposely non-technical as he walked through a general checklist of the steps needed to create the data needed for a basic Huff model.It was not designed to create a model, but rather as a way of explaining how the model is created and what attributes are needed.These steps were given in conjunction with a visual example of a study area containing several supermarkets - the objective being to determine if an additional store would contribute enough sales to offset any impact to existing stores.

The steps were:

Delineate the study area
Divide the individual stores' sales areas
Identify the centroids of those sales areas
Identify the competition
Determine the drive time distances between stores (yours and theirs)
Identify the facility attributes that affect consumer preferences

Conduct a customer intercept survey about which stores they go to, how often, things they like, etc.
Look at the total dollars available in the study area.

Each of the steps brings you closer to an understanding of the study area and those factors that can influence the nature of the model.The base divisible number was the total dollars available in the study area.By knowing that and your current market share, it is easier to see what can be obtained by making network changes.

Dr.Huffs' objective was to show that the mechanics of building a reasonable model are well established, and if you followed the steps, didn't take any shortcuts, performance prediction was within reach.

Dr.Lea, while following similar steps to Dr.Huff, added to the discussion by pointing out several interesting complications with models.For example, you needed to know what was the problem you really want to solve.He listed several network management scenarios:

Add a store
Close a store
Move a store
Create a new network
Change store attributes
Change demand

Maps aren't all that important in the case of models, the output reports and tables were the most important.

Also, in the discussion of trade areas, Dr.Lea established that the trade area was a surface, not a polygon.This meant that the trade area should be viewed as a 3D entity rather than a 2D geography.

If you bring together the main themes I've discussed about the conference, you can start to think about the concept of BI visualization, the spatial interaction model you have created, the trade area 3D surface, etc., and you might still just be reporting the obvious!

Like many conferences, I enjoyed the interaction with people doing the same things - BG-ers, BI-ers, database aficionados and others.What made this conference different is that while the old standard applications were certainly present, the real focus was how to visualize Business Intelligence.

Every conference has a section on predicting future trends - what we will be doing five years hence.In this case, it was about fulfilling the promise of GIS, the visualization of where.But where also needs what and who in order to have actionable knowledge.Neither GIS or BI has really delivered as well as hoped on their promises.Perhaps togetheras BG-ers, we can get to a better place when spatial, temporal and action become one.Many of these ideas may well be expanded on at next week's Location Technology and Business Intelligence conference.
More about this author...