Archive for the ‘Forecasting’ Category

Getting buyers to agree to a test

February 24, 2010

Getting your buyer to agree to push order recommendations, modular changes, SKU assortment changes, etc can be a challenge.  Here is some practical experience on how to make it happen.

Running a test with your buyer can be a very effective way to ‘sell them’ on new ideas.  Many times our clients want to use our forecasting tools to push recommended orders to replenishment managers but the replenishment manager is not receptive to adding extra work to their day and they certainly don’t want to risk overloading stores with too much inventory.   Many times our clients want to change the modular assigned to a store or SKU assortment within an existing modular but again the buyer is reluctant to make a change that could have negative results.  Proposing a test is a good way to limit their risk and overcome their concerns.  If the test is properly designed and the control group is selected to provide a proper comparison your idea should receive a fair vetting.

I spent considerable time today helping a client build a list of stores for a modular test at Wal-Mart.  My client has modular’s at Wal-Mart in the following widths: 40’, 36’, 32’ 20’ 12’.  The SKU assortment grows based on the width of the modular so a 20’ modular has all the SKUs of a 12’ modular plus some extras.  They have gained the agreement of their buyer to test 25 stores with a larger modular than the store would otherwise qualify for to see if the demand for their products is deserving of more square feet.   The test stores were identified by the buyer and are in close proximity to my client’s office so they can easily visit the stores.  The test stores have all been promoted to 36’ modular’s which is larger than they had in 2009 and larger than they would otherwise be traited for based on their profile.  The task today was to identify 25 control stores so we can test the sales lift over a 18 week period.  To identify the control stores the following information was pulled out of Retail Link: 2009 total units sold by store for all stores in my client’s home state.  The first thing we did was calculate the minimum, maximum, average, and median 2009 unit sales for the test stores.  We then eliminated all potential control stores which were not within the min/max, and then further narrowed our list by looking for stores that were +/- 20% of the median 2009 test store group unit sales.  All stores in the test group are Supercenters so we then eliminated all stores under consideration for the control group that were not Supercenters.   The next consideration was the demographics of the test store group compared to the potential control stores.  We pulled a list of demographics for test store group using the store zip code for each of the 25 test stores and looked at the following traits:  population density, median income, dominate race, and median age.   We created a profile using the averages for these traits.  We then cross referenced the possible control stores demographics along the same traits to identify the closest matches. 

The key is that by using UPC/store level EDI 852 or Retail Link data the vendor is often in a better position to analyze the demand of individual stores and make recommendations to a buyer on things like orders, modular’s, and SKU assortment.  Store level planning is the holy grail of maximizing sales but I’ve not met a buyer yet that has the time or resources to do that.   So the responsibility falls on the vendor to make it happen and proposing a test is often the way to get the ball rolling.  more information on retail analysis.

Our team built a social networking community for retail industry professionals to discuss best practices like this.  Membership is complimentary so we recommend you join in.


Retail Replenishment – How tuned in are you?

February 23, 2010

I spent about 9 hours yesterday analyzing sales, order, and forecast data for Wal-Mart, Home Depot, and Lowe’s vendors and I am somewhat surprised by my observations.  It’s pretty clear there are some min/max rules in place as I can see patters to the order quantities based on the OH inventory and the order case pack quantities.  However what surprises me is that I also see a large number of what I would guess are “manual overrides.” That is UPC/stores which clearly need inventory and fall under the minimum OH of other stores but which do not have open order, and UPC/stores that are clearly overstocked (e.g. high WOS) and yet have an open order.  It makes sense there would be automated replenishment rules in place and then some lead-way for the buyer/replenishment manager to make judgment calls so that leads me to my question….

What do you know about your key retail customers replenishment rules?

  • Simple min/max ordering?
  • Based on OTB dollars?
  • WOS trigger?
  • At what level is demand calculated?  e.g. Category/Region, Category/State, Sub-category/State
  • Under what situations will the buyer do a manual override?

Retail sales improvement requires careful forecasting

February 5, 2010

The WSJ reported retail sales Rose 3.3% showing signs consumers are returning to stores.  This is a great sign for the retail market as it seems a turnaround may be in the works.  Macy’s posted a 3.4% increase, Saks reported 7%, and Costco 8%.  As demand begins to increase vendors need to keep a careful eye on the supply chain.  Retail buyers have been operating on low open to buy for over a year so inventory levels may be below where they will need to be to satisfy demand.  Vendors using EDI 852 data for forecasting need to make some careful adjustments to their forecasting model to not be caught by surprise.  Here’s why.  Forecast models using historical demand as the foundation for current year predictions but last January was a terrible month for retail sales so a simple look at comp year demand will give a misleading result.  To correct for this vendors should be considering not only last year’s demand but also the prior year’s demand and the current period trend.  By combining these three numbers vendors will have a more accurate model and hopefully not get caught by surprise.  But even with a good forecast we expect sales to be unpredictable for the foreseeable future so vendors must carefully watch demand and inventory levels by analyzing the EDI 852 data weekly or even daily and making push order recommendations to their buyers.