By Jaime Reints | Director, Demand Planning Intelligence Consortium (DPIC) | Chainalytics
Asking these three questions below will help establish better forecast accuracy targets, prioritize improvements and ensure forecast accuracy goals are achieved.
Forecasting demand can be one of the more difficult parts of managing an effective supply chain. In fact, participants in a recent Gartner survey named forecast accuracy and demand variability as the top barriers preventing companies from reaching their broader supply chain goals.
If the organization is moving into the new year with an increasingly volatile or growing portfolio of products, ask these three questions below to help establish better forecast accuracy targets, prioritize areas of improvement opportunities and ensure those forecast accuracy goals are actually achievable.
1. What’s A “Good” Forecast Accuracy?
How the demand behaves within a given portfolio naturally affects a planner’s ability to predict and forecast future demand. For example, a slow-moving, low-velocity item with intermittent non-consecutive demand pattern that is highly variable should have a different accuracy expectation than say a fast-moving high-velocity, low-variability stable item.
In other words, the definition of “good” will fluctuate accordingly with each brand, category, so on and so forth. To establish what forecast accuracy is “good” relative to the difficulty of the demand’s behavior to predict it’s essential to first quantify and measure the forecastability – the rated difficulty of a given demand behavior to accurately forecast – for each product at the item-stocking location level, and where the other fulfillment actually happens.
However, to begin empirically defining and quantifying what “good” actually is requires (1) a data set that provides the correct basis of comparison and (2) an approach that controls for key differences between data points within the given data set. Advanced segmentation techniques can be used to control the different factors that influence a product’s forecastability – such as volume, velocity, variability and behavior.
Deriving the forecastability at this level of detail for each product within a segment, realistic, fact-based targets can then be aggregated and easily tailored specifically for each facet of the business, and product lifecycle stage.
Viewing forecast accuracy performance along the spectrum of forecastability, to control for the differences across various portfolios, products, groups, and geographic regions, enables more intelligent – apples-to-apples comparisons to be made while also bringing into alignment tailored, data-driven expectations for each segment. So before taking last year’s forecast accuracy metric and simply adding a few percentage points to establish next year’s goals, remember a portfolio’s overall forecast accuracy performance and expectations should be viewed in relation to its forecastability ranking.
2. Where Should We Focus Our Efforts?
Neither improvement opportunity nor benefits are evenly distributed across an entire product portfolio. Embracing this basic concept is critical to the success of driving actual performance improvements and stop spinning the wheels;
Fact: “Not all opportunities are created equal.”
At the heart of this is a core belief espoused at Chainalytics regarding portfolio segmentation and differentiation. Deep segmentation provides a more accurate capture of the true demand volatility and drivers of a product and portfolio’s demand forecastability. These setment-specific insights can also be used to create a differentiated and optimized process for planning at each relevant hierachy, grouping and channel if support with marketing intelligence around achievable and realistic performance expectactions.
Granularly understanding how demand is behaving, quantifying which behaviors are easier to forecast, identifying what potential accuracy is even achievable, and then measuring how current performance stacks up against this calculated potential will clearly expose the lowest hanging fruit that’s ripe and ready for the picking.
Given limited time in each planning cycle (and day), understanding current performance in relation to each segment’s respective maximum potential accuracy is one highly efficient method to expose specific product lines, items and item-locations which are truly underperforming in relation to a defined standard of “good” based on the forecastability.
3. What Needs to Change?
Forecasting is often a unique blend of art and science. A lot of time, energy and effort can be put into getting the final consensus number just right which can elicit strong emotions from all parties involved in forecasting exercises.
More often than not, effective change management hinges on a planning team’s ability to provide a solid business case at the start that clearly communicates bottom-line benefits and subsequent impacts any changes will have up and downstream to key stakeholders involved. This task requires having access to the right data, combined with the capability to easily perform investment-grade analytics to answer key questions and enable a culture of continuous improvement. In 2017 this shouldn’t be an unrealistic pipe dream for planning teams.
In conjunction with proven best-practices and available industry benchmarks, consider these tips below to fuel more informed, data-driven discussions which engender accountability with all involved in the consensus forecasting process to help drive tangible bottom-line results;
- Analyze error and bias across various critical planning lags, item-locations and market segments in detail to quickly reveal underlying root causes to ensure subsequent changes are actually targeted at the true source(s) of error.
- Take a second look at historical performance over the span of the entire planning cycle to pinpoint where and when the true source of the error was introduced into the process. That means if final consensus forecasts aren’t being saved and archived at each critical planning lag consistently over time, now’s a great time to start!
- Use well-designed data visualizations and dashboards, in tools such as Tableau, to effectively communicate to other non-planning stakeholders and enable teams to literally “see” how the behavior of forecasts vs. actuals trends over time, consequences of missed timing of promotion/events and impacts of bias, over- or under- reactive forecasting processes – to clarify assumptions and help to take some emotion out of otherwise highly-political sales and operation planning (S&OP) processes.
- Leverage Forecast Value Add (FVA) analyses properly can identify where to specific errors are being introduced into the consensus process. FVA can help planners and sales, marketing and customer service resources to better allocate their time and efforts by providing insight on where improved business information is needed most during the S&OP process while also revealing where manual adjustments and business inputs are enhancing or degrading the final forecast’s accuracy.
No doubt in today’s business environment of proliferating product portfolios and increased demand volatility, there is a great deal riding on getting the forecast right. Ben YoKell, vice president of Chainalytics’ Demand & Supply Planning (IDSP) consulting competency and I will elaborate on what’s needed to achieve 2017’s forecast accuracy goals and improve demand planning performance in the years beyond during our upcoming webinar, “Four Demand Planning Resolutions for 2017.” Click here to view the webinar replay.
As the director of Chainalytics’ Demand Planning Intelligence Consortium (DPIC), Jaime Reints oversees membership development product innovation, and delivery of the DPIC market intelligence solution for some of the largest CPG supply chains around the world.
Read more about how DPIC can help improve forecast accuracy and demand planning performance:
