DC Velocity - May 2013

however, it has been using actual retail sales data to drive both replenishment and manufacturing.

Three retailers, which account for one-third of Kimberly-Clark’s consumer products business in North America, currently provide point-of-sale data. That information is fed daily into the solution’s engine, which then recalibrates the shipment forecast for each of those retailers. Each day, the software evaluates any new data inputs from the retailers along with open orders and the legacy demand planning forecast to generate a new shipment forecast. That forecast, in daily buckets, covers the current week plus the next four weeks. Kimberly-Clark then uses that forecast to guide internal deployment decisions and tactical planning.

The software contains algorithms that process data provided by the retailers, such as point-of-sale information, inventory in the distribution channel, shipments from warehouses, and the retailer’s own forecast. It reconciles all of the data to create a daily operational forecast. The software also identifies patterns in the historical data to determine which inputs are the most predictive in forecasting shipments from Kimberly-Clark’s facilities. The inputs are re-evaluated weekly, and how much influence each input has on the forecast can change. For example, POS might be the best predictor of a shipment forecast on a three-week horizon, but actual orders could be the best predictor for the current week.

By using actual demand—that is, the point-of-sale data—to recalculate its operational forecasts, Kimberly-Clark can better ensure that it has the products consumers want to buy in stores at the right time. Although only three companies at the moment are providing POS data, Kimberly-Clark is also using the Terra solution to create forecasts for its other retail customers. For that customer group, the manufacturer relies on historical shipment data to develop its forecast.

LOWER FORECAST ERROR RATES The incorporation of demand signals into Kimberly-Clark’s shipment forecasting has resulted in substantial improvements in several respects. For one thing, the compa-

ny has been able to develop a more granular metric for forecast errors. In the past, it measured forecasts by product categories; the metric it uses now tracks stock-keeping units (SKUs) and stocking locations. This metric is defined as the absolute difference between shipments and forecast, and it’s reported as a percentage of shipments.

Since it began using that particular metric to evaluate its daily forecast, Kimberly-Clark has seen a reduction in forecast errors of as much as 35 percent for a one-week planning horizon and 20 percent for a two-week horizon. “What we’ve noticed is that as you go farther out in the [planning] horizon, the inputs are less predictive and the amount of forecast-error reduction tends to erode,” says Jared Hanson, a demand planning specialist.

Thanks to the reduction in fore-
cast errors, there is less need for
safety stock. In fact, Hanson says,
more accurate forecasts have
allowed Kimberly-Clark to take out
one to three days’ worth of safety
stock, depending on the SKU.
“From a dollars or return on invest-
ment perspective, that’s the most
tangible benefit,” he says.