IT BEARS REPEATING…HOW TO FORECAST SLOW SELLING ITEMS

By Demand Clarity on December 4, 2013, Supplychainbeyond.com

“It sells less than 10 units per store per year! Who cares?!?”

Slow selling items is a topic we cover periodically in our newsletter (The Right Flow) and with good reason. For most retailers 50-70% of the items they sell don’t exactly fly off the shelves.

As supply chain folks, it’s easy to throw our hands up in the air and say things like:

  • “We can’t make an accurate forecast at item/store/week level”
  • “The forecast doesn’t matter anyhow – the safety stock and inventory level is what’s going to drive store replenishment”
  • “We should be spending our time, energy and resources on the items that are actually selling”

If supply chain people had their way, these items would be delisted from the assortment faster than you can say
“Who cares?”

But Category Managers don’t think in terms of individual items. They’re trying to juggle a delicate balance between assortment breadth and inventory investment. If a hardware store doesn’t have every possible size machine screw (including the ones that don’t sell very well), then a customer who doesn’t want to waste time shopping around for the right screw will never walk in the door.

Broad assortments (not necessarily across the board, but at least in certain categories) are needed to attract and retain customers and they’re here to stay. And with assortment breadth comes slow selling items. In this regard, the slow sellers are even more important than fast sellers (which are probably readily available elsewhere) for long term customer retention.

So, given that slow selling items can’t be swept under the rug, how do we deal with them in the planning process?

We can’t make an accurate forecast at item/store/week level

It’s true that predicting the exact location and timing of sales when volumes are low is virtually impossible. But this level of accuracy isn’t really necessary, either. When selling volumes are low, the weekly calculated forecasts are primarily decimal numbers (and sometimes very small decimal numbers) and we all know that the actual sales will be in discrete units. However, this fact does not make the forecast useless in the planning process. In these cases, the forecast represents a prediction of the probability of a sale at a location rather than a direct prediction of sales. These probabilities can be “post-processed” into a sale expectation by starting with a random seed value between 0.00 and 1.00 and accumulating the probabilities across time. A “final forecast” of 1 unit is created in weeks when the accumulated probability becomes greater than 1.00. This results in an item/store/week forecast that is realistic and can be used for planning purposes.

The forecast doesn’t matter – safety stock will drive store replenishment

This may be true – for store replenishment only. But the world of supply chain planning is moving inexorably toward a demand-driven approach and gone are the days when we could think of “store replenishment” and “supply chain planning” as being separate and independent from each other. Stores serve customers. DCs serve stores. Suppliers serve DCs. Inventories and constraints exist at and among stores, DCs and suppliers.

There’s no getting away from the fact that planning requires forecasting. In order for the plans to be truly demand driven, the forecasts need to be created when and where sales are expected to occur – by item by store by week. In the interest of postponement and to allow for the fact that forecasts are not always reliable at item/store/week level, It’s not always necessary to trigger actual stock movement to the stores based on these plans. The plans themselves are, however, necessary to ensure that sufficient upstream inventory will be deployed at DCs and suppliers in preparation for store orders that will be triggered when actual sales drive store inventories below safety stock.

We need to focus our effort on the items that are actually selling

This is a throwback to the days when computer systems were not very sophisticated and items were segmented by selling volume (“A” items, “B” items, “C” items, etc.) as a way of driving workflow. In our early days of implementing time-phased planning in retail, we’ve found the exact opposite to be true – items that sell in large volumes are easier to forecast and much more stable. As a result, high in-stock rates and inventory turns can be maintained without very much effort. When was the last time you went to the grocery store and they were out of milk?

Nowadays, systems can do a lot more of the drudgery that was previously done by people, so the need to use volume classes to group items has greatly diminished. Instead, we are turning toward using true supply chain exceptions as a way of determining the priority of people’s work. What’s truly more important for a person to be dealing with, a top seller that has a 99.5% in-stock rate and 80 turns or a slow seller that’s chronically out of stock in 50 locations? Today’s planning systems are making it possible to be more customer focused (rather than product focused) and find needles in haystacks that require human intervention.

So who cares about slow selling items?

You do… right?