"Our biggest challenge in the implementation," Bartok continues, "was making sure that our data were clean so that we could execute everything correctly, whether it be modeling, reporting or exceptions. We wanted to make sure that our customer codes and attributes that were unique to a vendor, unique to a customer or unique to an item were clean. That was our biggest task."
In addition, because Wells Lamont was implementing the new forecasting solution at a time of the year when plans were already set for many of its key customer accounts, Bartok's team opted to map the existing forecast data into the Galt engine for the affected SKUs, using those figures as the starting active forecast and making changes, as necessary, to that data immediately after the implementation. The other option was to map in the sales histories for each product and allow the engine to generate forecasts, but because the planners also incorporate so much outside business intelligence into their forecast figures, it made more sense to simply go with the existing projections.
As the process stands now, at the beginning of a new planning period Bartok and his team map sales history updates into the forecasting engine and then run all the exceptions for all their customers and products, giving them visibility into who's trending up and down, based on over- or under-performing sales versus current forecast. For example, Wells Lamont's planners run exceptions that identify seasonal customers whose products should be booked at that particular time and exceptions to identify new products without a forecast. They also ran exceptions to identify products that are trending toward obsolescence, or to determine which customer or territory has the highest percentage of ownership for a particular SKU over a 12-month period, which can give the planning team a heads-up to watch a particular customer to detect any trends, especially when key accounts are involved.
Once the planning team has completed running the exceptions, they run the statistical models for all the product lines in the two divisions and then make available to all the appropriate internal decision-makers — those with account or product line responsibility — all the demand data by putting the forecasts up on the company's intranet in the form of Excel pivot tables. The data available to the decision-makers include sales history, current forecast for all items in the customer program or product line, and statistical forecast changes (what the system is recommending), as well as any customer or collaborative data as another opinion line for the decision-makers to consider in making adjustments. Importantly, the system also highlights for the decision-makers those items that require action.
With those data in hand, at some point during the planning period the decision-makers can elect to accept the statistical model changes for their SKUs, or they can interject their own business intelligence or a customer's recommendation. Either way, they provide feedback on the pivot tables to the planners, and the system maps the pivot tables into a simulated scenario so that the planners can review the new projections, compare them to the active forecast and, where necessary, question changes, ultimately allowing demand planning to make a decision on which data streams to take as the forecast adjustment.
The demand planners make the forecast adjustments, based on the decision-makers' feedback, throughout the monthly period, but they also do updates back to Wells Lamont's ERP system once a week, incorporating all the latest forecast changes. In addition, four or five times a month the planners capture the net unit changes for production and purchasing to make those departments aware of any impact on current plans within the lead time for different SKUs. The company forecasts out up to 18 months from the middle of the year, going out through the end of the following year for the "rough cuts." Then the planners make changes during the course of the year as necessary.
Currently Wells Lamont has two user licenses for the Galt solution, but the Excel pivot tables generated by the planning team have an audience of about 50 people within the retail division and an equal number on the industrial side of the business, according to Bartok. "We're doing the pivot tables right now, because it's in line with what the sales guys were used to before," the demand planning manager adds. Starting in 2005, Wells Lamont plans to move to the Web form of the John Galt engine, which will allow the planners to share information with decision-makers more quickly, Bartok says.