Point of Sale Data – Category Analytics

If you haven’t already read the previous entries in this series, you may want to go back and check out [Point of Sale Data – the basics] to see why you really need a DSR to handle this data, and  [Point of Sale Data – Sales Analytics]  for some thoughts on analyzing sales drivers that are equally relevant to Category Management,

 As Category Manager you’re working with the retailer to help drive sales for the entire category.  You hopefully have access to the full data for your category (which could be substantially more than your account manager colleagues).  Let’s see how predictive analytics and modeling could help address some of your challenges:  How well are current planograms performing?  What is the best product assortment for each store?  How can you best balance customization of assortment by store with the work required to create that detail?

• Understanding which stores are truly similar in terms of what sells and why in terms of demographics, geographies and local competition helps you manage your store list as a small set of groups (“clusters”) rather than trying to deal with each store individually or abandoning store-customization as too time consuming. 
• Cluster analysis is a relatively simple statistical process (see Cluster Analysis - 101 ) and there are tools available from high-end statistical modeling packages to Excel Add-ins that can handle this process. Preparing the data appropriately, interpreting the results, and creating presentation ready output quickly is more of a challenge. I’m working on something to help with this…watch this space. 
• Tools that help you find individual products that are not currently listed in a store but sell well in local competitor stores (from syndicated data) or in the retailers own “similar stores” help correct the assortment list.
  It’s relatively easy to find such “missing stars” when you know where to look. For example, stores in areas with lots of Hispanic households should probably stock every one of your top 10 Hispanic-oriented products. But this approach relies on you knowing these associations up front: the analytic approach scans your database to find all possible associations. 
• Understanding product substitution and the shopper’s decision tree ensure that you add the right products onto the assortment without excessive cannibalization of existing products.Tools that handle assortment optimization are becoming more common now. Quite how well they work should depend heavily on how well you clustered stores into “like” stores and how accurate your decision tree is. 
• Automating the assortment-selection and planogram-build processes can allow you to work at lower levels of detail and provide more finely tuned customization.  In many ways, this is not the area likely to generate the best return. The next best option is actually to bring in more temporary labor and it just does not cost that much to do so. However, what I see in reality is that category managers work extraordinary hours to try and get it done. They can’t work 10x normal hours so the work cannot get done to the depth that is possible. Neither can they handle changes late in the project with the same diligence and structure they brought to it initially – there is simply not enough time. There is at least one tool on the market for building planograms now and I may be tempted to build another J

Predictive analytics like these may seem complex and may only drive a few percentage points of incremental sales.  But then, what’s even 1% of incremental sales worth to you?  Worth handling a little complexity?

All posts in this series

• Point of Sale Data –Basic Analytics
• Point of Sale Data – Sales Analytics
• Point of Sale Data – Category Analytics
• Point of Sale Data – Supply Chain Analytics