William McKnight writes an interesting article in DM Review this month looking at business intelligence trends in retail.  As he says, retailers ideally need to track promotions and customer buying trends, yet are often poorly served by management information systems.  One reason for this is the sheer volume of data: if you consider every item on a till receipt as a transaction, then each store will be putting through many thousands of transactions per day. A convenience store might have 20,000 stock keeping units (SKUs) but a department store might have over 200,000.  In addition a retailer will want to keep track of customers who have loyalty cards, will be concerned about space optimisation and stock control.  The more astute retailers vary the stock on their shelves in accordance with buying patterns they have observed: clearly the customer profile in mid morning is different to just after schools finish, for example.  One Japanese chain reckons to change the stock profile on its shelves seven times per day.

To get this kind of insight you need a high quality, robust data warehouse that is able to handle large data volumes and keep up with rapid change. However one thing that I learnt when working on a Shell retail project a few years ago was that you can make life easier for yourself by quickly archiving the transaction detail.  A category manager may want to do basket analysis for a few days, but beyond that is interested in trends, which can be satisfied by aggregate information.  This allows you to rapidly archive the high volume transaction data and so keep the data in the warehouse to manageable levels. 

Good BI on retail can make a major effect on profits, as I have written about before.  The dawn of RFID, though still in its infancy, will further extend the possibilities for more elaborate analysis, though my suspicion is that most retailers are barely scratching the surface of what can be done today using the current technology.

William McKnight writes an interesting article in DM Review this month looking at business intelligence trends in retail.  As he says, retailers ideally need to track promotions and customer buying trends, yet are often poorly served by management information systems.  One reason for this is the sheer volume of data: if you consider every item on a till receipt as a transaction, then each store will be putting through many thousands of transactions per day. A convenience store might have 20,000 stock keeping units (SKUs) but a department store might have over 200,000.  In addition a retailer will want to keep track of customers who have loyalty cards, will be concerned about space optimisation and stock control.  The more astute retailers vary the stock on their shelves in accordance with buying patterns they have observed: clearly the customer profile in mid morning is different to just after schools finish, for example.  One Japanese chain reckons to change the stock profile on its shelves seven times per day.

To get this kind of insight you need a high quality, robust data warehouse that is able to handle large data volumes and keep up with rapid change. However one thing that I learnt when working on a Shell retail project a few years ago was that you can make life easier for yourself by quickly archiving the transaction detail.  A category manager may want to do basket analysis for a few days, but beyond that is interested in trends, which can be satisfied by aggregate information.  This allows you to rapidly archive the high volume transaction data and so keep the data in the warehouse to manageable levels. 

Good BI on retail can make a major effect on profits, as I have written about before.  The dawn of RFID, though still in its infancy, will further extend the possibilities for more elaborate analysis, though my suspicion is that most retailers are barely scratching the surface of what can be done today using the current technology.