By Ben YoKell | Principal, Supply Chain Analytics | Chainalytics
In 2014, Gartner put forecast accuracy at the very top of its hierarchy of supply chain metrics pyramid, highlighting the importance of pursuing excellence in forecasting and planning in order to help drive profitability and growth in the coming decade.
Fast forward two years, and a variety of reports, articles, blogs, and my own firsthand experience suggest that many companies are beginning to hit a limit when it comes to improving demand planning performance, raising questions about whether it makes sense to continue trying to “Chuck Yeager” their way through the uncertainty barrier.
As a supply chain advisor, I am sometimes asked, “Should we continue trying to improve our forecast accuracy and demand planning performance, or is it all just a waste of our time?”
In my opinion, there is still plenty of juice worth the squeeze. Not only are both demand and supply planning processes still ripe for harvest, the integration of supply chain planning with supply chain visibility and optimization is now on the horizon, and I’m hopeful this change will unlock new potential in the coming years. Also worth noting, the planning technology space has been busy with acquisitions and major investments, and machine learning is already in play and is here to stay. As I see it, most organizations haven’t even begun to hit Mach 1 yet.
“So then why are companies stalling now when it comes to delivering ongoing improvements in forecasting and forecast accuracy?”
Four Reasons Why Your Forecast Accuracy Isn’t Improving
Over my tenure as oversight for the Demand Planning Intelligence Consortium, there have been several common themes worth mentioning on this topic. For this post, I’m calling out four factors that I keep hearing, which hinder progress in achieving forecast accuracy improvements:
1. Our ability to plan and manage has not kept up with our capability to innovate
New item innovations and event- or promotion-driven sales now constitute a majority of demand for many companies, creating additional “self-imposed” volatility. As a prime example, take a look at how this has played out in the beer industry in this article, “Crafting Complexity”. Business strategies have changed without any data-driven understanding into how supply chain planning performance is being impacted.
2. Management and performance expectations have not evolved
Forecast accuracy targets are still being set by negotiation from historical performance rather than any empirical data on what is realistic or possible (we hit X last year, let’s improve by Y this year). This approach does not take into consideration the nuances of your portfolio’s unique product mix or forecastability (the predictability of demand). Further, this historical view neglects broad market trends, such as how market volatility changes through time. If volatility in the market has increased by 3 percent, is it realistic to expect the same – or better – forecast accuracy? Not without an investment in people, processes and/or technology. If your demand gets harder to predict, and you achieve the same forecast accuracy, is that an improvement? In absolute terms, no; but relative to the predictability of your demand? I would say yes.
3. Resources are constrained and time is still limited
As Lora Cecere points out in a related post, “Household Products Industry Stuck in Neutral and Going Backwards,” while many of the consumer product goods (CPG) companies have moved towards a more centralized approach to gain additional resources to leverage and better manage by exception, most have failed to provide a senior management career track required to retain and grow skillsets. As a result, many have fallen behind in the race for supply chain talent.
4. Teams lack quantitative performance insights, benchmarks and analytics
Despite technology advancements, most demand planning teams still do not have the bandwidth to manage their growing portfolios with thousands of demand forecasting units (DFUs), while also efficiently supporting the robust analytics required to produce corrective action quickly. For example, do you know what the maximum forecast accuracy potential of a new product introduction is today for a specific item-location one month ahead of demand? Even for those planning teams blessed with ample time and analytical prowess, insights are still limited to their own internal data, and lack visibility or any comparison basis around how to define “good” forecast accuracy for the specific behavior of their demand.
Despite all of the buzz around Big Data and advanced analytics today, the level of sophistication around understanding forecastability, forecast performance, and each item location’s expected and maximum possible forecast accuracy has been surprisingly low. If we don’t advance our thinking about how best to measure and improve forecast accuracy, there is a very real risk of becoming data rich and insight poor.
A Smarter Approach to Improving Forecast Accuracy
As I see it, part of the problem is having asked the wrong question to begin with. It’s not a matter of whether we should invest in demand planning; it’s a question of how and where to invest to maximize our return on effort.
Forecast accuracy does have real limits. These limits vary by business, go-to-market strategy, product line, geographies, and even by distribution channel – precisely because the behavior of the demand differs across these very dimensions. Even further, these limits and expected performance levels also change over time as market, technology and customer behavior trend and progress.
What forecast accuracy is reasonable to expect for a new product launch? What is the best that can be done – for my kind of business? What is a “good” forecast accuracy for a slow-moving, moderately volatile, somewhat seasonal item, two months ahead of demand? How are we supposed to know? What is good?
Using data-visualization tools, advanced demand planning and benchmarks available from the Demand Planning Intelligence Consortium (DPIC), many have begun to push to new limits by making data-driven adjustments to how they measure, segment and manage their demand planning process.
For those interested, I’m hosting a live session on August 26 that will explore this simple, scientific approach in more detail. If you’re hitting the sound barrier and looking for new ways to improve forecast accuracy, check out this session and please join us as we continue the discussion on these topics.
Ten years ago, there were no touchscreens, no WiFi on planes, and no real time augmented reality games sending people all over in search of anime. Yet this post was written from 35k feet above the eastern seaboard, edited and cited online, and sent for publication in the airport after landing, from my iPhone while playing Pokémon Go. Point being, let’s not write off progress in demand planning just yet, it’s a quickly changing landscape out there! We just need to get a little more scientific about how and where we invest our efforts to break through the barrier.
Ben YoKell is a principal at Chainalytics and leads the firm’s Demand Planning Intelligence Consortium (DPIC). He is passionate about the use of quantitative techniques in supply chain management and thrives on helping companies leverage optimization and analytics in planning environments with an emphasis on actionable results. In conjunction with the application of analytics, Ben brings to bear significant knowledge of supply planning practices and processes for a well-rounded approach to improvement initiatives.