Blog post

Demand forecast error NEVER WAS the biggest driver of inventory

By Paul Lord | December 07, 2015 | 11 Comments

Some have recently expressed surprise to learn that demand forecast error (DFE) might not be the significant driver of inventory they previously thought. After all, DFE is cited as the primary root cause for nearly every supply chain challenge, including inventory, in most of Gartner’s research surveys. Other commonly-cited causes are product complexity and risk.

The beauty of these simplistic ‘inventory sound bytes’ is that they are partially rooted in truth and usually sufficient to shift the conversation away from the topic. However, the issue never really goes away and the CFO eventually acts out of frustration by mandating a 10 or 20% reduction in inventory.

Positioning and managing inventory in the supply chain is a complex topic beyond what can be determined with average benchmark comparisons or correlations against DFE. While statistically undeniable, single-variable correlations have contributed to an oversimplified understanding of the topic and led to disappointment with technology investments and other reduction efforts.

Does improved forecasting benefit inventory performance? It would be unwise to say that it plays no role, but the truth extends far beyond a single driver. The need for inventory starts with the relationship between supply and demand lead times, compounded by risk, variability, forecast error and bias. It should also depend on the economics of your supply chain, including the risk-adjusted cost of carrying inventory relative to the opportunity cost of a service failure. Blend in some other supply constraints (both physical and economic) relative to event-based demand peaks, and you’ve got a wicked problem to solve.

Some of the more persistent problems we are seeing relates to a gap between stand-alone financial analysis of inventory versus the interdependence of inventory operating drivers with other supply chain performance metrics.

– Most supply networks have been designed for low unit cost, which impairs the supply response with longer supply lead times and other constraints.
– In an attempt to remain competitive and responsive, companies are compensating for long lead times by making speculative supply commitments (well in advance of firm demand).
– Alternatively, a focus on book profitability (rather than cash flow) contributes to over-production for fixed-cost-absorption purposes to reduce the unit cost of supply.

The result of these complex dynamics tends to be dismay with inventory levels, yet no clear path to resolution given the combination of design choices and operating incentives in play. What is the answer to all of this complexity?

– Supply chain leaders must embrace their roles as the stewards of inventory health by becoming ambassadors of inventory knowledge.
– They must advocate and lead the analysis, governance and implementation of decisions that impact all three categories of inventory: structural, operating and situational.

This will be the topic of future writing. In the meantime, feel free to contact me if you’d like to discuss or share your views.

The Gartner Blog Network provides an opportunity for Gartner analysts to test ideas and move research forward. Because the content posted by Gartner analysts on this site does not undergo our standard editorial review, all comments or opinions expressed hereunder are those of the individual contributors and do not represent the views of Gartner, Inc. or its management.

Leave a Comment


  • SIMON EAGLE says:

    Look forward to the rest of your articles Paul.

    Agree that DFE per se is not a driver of incorrect inventories – unless the forecast is being used to directly drive replenishment in which case its the under forecasts that cause the backorders, the unplanned overtime and, weirdly, generally excessive inventory levels.

    And if the forecast has persistent bias, as opposed to just error, it will cause performance issues even if only used for Planning.

  • Great insight, Paul. The unit cost focus coupled with long lead times (among other issues) are the perfect recipe for poor OTIF and large inventory excesses.

    As Simon said, look forward to your other articles.

    Thanks for sharing.

  • David McPhetrige says:

    Very refreshing article. Keep them coming! IMHO, random variations in demand and supply – not DFE – are the biggest challenge to optimal inventory position. Random variation’s unpredictable timing simply is not in a demand forecast. (Better forecasts can assign timing to some previously random variations, making them events.)

    I agree – financial metrics can inadvertently drive suboptimal inventory performance. You stressed the importance of metrics acknowledging the complicated interplay of drivers affecting every item in every location. These drivers go beyond random demand and supply variations, including MOQ, order multiple, reorder-review cycle, target service level (SL), actual-SL measurement cycle, desired confidence of target SL in any SL cycle, probability of demand cancellation vs past-due backlog, and replenishment method. As Simon Eagle’s comment pointed out, DFE is typically an inventory driver only when there’s significant bias. (IMHO, planning-BOMs mix assumptions often cause bias. They need the same S&OP scrutiny as forecast.)

    Actual inventory vs optimal is a key metric of multiple inventory drivers, avoiding the oversimplification dangers you mentioned. For an item in a location, “optimal” is the minimum inventory that achieves the target SL with the desired confidence, without expediting. “Optimal” can use as constraints each item-location’s current set of drivers, or it can instead use future, ideal or “what-if” driver values (e.g., smaller MOQ, shorter lead time and/or less lead-time variation, etc.).

    • Paul Lord says:

      Thanks David. Random deviations in demand manifest themselves as forecast error, especially if forecasting based on a rolling average. But safety stock is only one of six different functions for inventory and it’s the only one that is mathematically impacted by DFE or normal variations. My view is that lead times and other constraints are responsible for a much larger proportion of inventory, on average. Shrinking lead time is a way to reduce the impact of demand volatility. Ultimately economics and risks determine how we respond to constraints, uncertainty and variability.

      • David McPhetrige says:

        Agreed 100%, Paul, other factors such as lead-time variation, large MOQ / lot size and long reorder cycle are often big culprits. Though DFE is typically not a major contributor to inventory position, demand variation is often a significant inventory driver. Absolutely, as you pointed out, shorter lead times reduce the impact of demand variation.

        I’d like to emphasize that DFE is an inadequate measure of an inventory item’s demand variation. To the extent that forecast is a guess, DFE measures only how good is the guess. DFE, especially at the inventory-item level, is also clouded by bias, as you and others have mentioned. Also, forecast period, such as a month or even a week, is unlikely to represent customers’ sensitivity to lateness. To a customer, a day can mean late instead of on-time. But to DFE, demand that’s late but still fulfilled in the same forecast period doesn’t even count as a “miss”.

        I look forward to your further insightful articles!

      • SIMON EAGLE says:

        Why not expand the idea Paul?

        Given that variability in supply chains is what causes generation of buffer (excess inventory, lead-time, capacity) why not make the case for variability reduction as being the route to inventory reduction, greater responsiveness and higher capacity utilisation?

        How to reduce variability? – Lean and multi echelon Pull and NOT using the forecast to directly drive replenishment through MRP (as most companies are doing)

        • Paul Lord says:

          Thanks for the comment Simon. This is what makes supply chain different from manufacturing, where variability is more within control of the operator. Variability in demand (and even supply) is generally beyond the control of the supply chain operator. The answer is to reduce exposure to variability, not by reducing variability but by increasing responsiveness through lead time reduction. The most direct way to improve both service and inventory productivity is to reduce supply lead time. But there is a cost trade-off, which is the essence of supply chain.

          • SIMON EAGLE says:

            Agree re reducing lead-times Paul but surely any variability within the control of SC people should be eliminated if possible for the reasons I give. And ceasing to use forecast push MRP is the biggest source of such “special case” variability?

  • Excellent article with some key points we normally do not see in inventory discussions.

    Typically, when a company has inventory issues (almost all do), it is not one single area that is a problem. Rather, it is a multitude of issues that interact and that I put into four broad categories: organizational, cultural, strategic and tactical.

    Years ago, I identified over 15 contributors that affect inventory performance. I went back and looked at my list. It listed the forecast process as “goofy”, but did specifically mention forecast error. I guess there is room for another entry on the list.