A few weeks ago, I posted a blog outlining key themes from the 2018 Forecasting and Inventory Benchmark Study that will be published later this month. As the “State of the Nation” for forecasting performance, supply chain complexity and the actual inventory carried by large manufacturers in industries including food & beverage, household & personal care, chemicals, oil & gas and industrial manufacturing, it contains a treasure trove of information.
One central theme relates to item proliferation and the true cost of the long tail. While growth through innovation is a popular strategy, it turns out that innovation grows complexity twice as fast as it grows sales. Despite high expectations, the vast majority of new introductions end up in the tail, fueling proliferation. In the last blog, I pointed out the inherent expense of maintaining a larger portfolio and how this also impacts planner productivity and performance. Why? Not only does the workload increase in terms of items per planner, but these slow-moving items are by definition harder to forecast.
Everyone intuitively knows that fast-moving items are easier to forecast – but just how much easier? To quantify this, we looked at the aggregate data across all companies in the study. It turns out that your top moving A and super-A items are twice as forecastable as items in the tail, with a naïve forecast error of 27% compared to 50%. This means that coming out of the gate, items in the tail are twice as hard to forecast.
The combination of increased workload in terms of the items per planner and a portfolio mix skewing towards more slow-moving and harder-to-forecast items creates a dilemma for planners. Given that forecast accuracy for top-selling items is much better to start with and technically needs less attention than those in the tail, the question for planners is “Do you focus limited time on the handful of super-A items that are essential to get right or on hundreds or even thousands of items in the tail?”. The pragmatic answer is to always focus on the most important items for the business and this comes across loud and clear in the data. The study reveals that the forecast value added – a measure of the value achieved through planners’ efforts and a company’s investment in demand planning processes and technology – is more than 4X greater for top-selling items than for items in the tail.
So what can be done to tame the tail? For starters, consider a stronger SKU rationalization governance to reduce the size of the tail and check proliferation. Good news is that this doesn’t require any new software. Bad news is that convincing stakeholders to cut items requires cross-functional (and often highly political) change. If this seems like a heavy lift, two easier approaches that stay clear of political waters are to sense demand and automate baseline statistical forecasts.
Data from the study reveals that demand sensing boosts forecast value-added by 5X for items in the tail. Unlike the traditional approach where planners need to carefully focus limited time to a few important items, there are never too many items for an algorithm. The demand sensing algorithm provides the same care and attention to the thousands of slow-moving items in the tail as the top selling product. Likewise, manually selecting and tuning statistical models – especially for the thousands of items in the tail – is a low-value task that is better suited for an algorithm. Leaders have fully automated the statistical forecasting process, providing planners with a better starting point and freeing them to focus on strategic activities like accounting for promotions and market intelligence. Both demand sensing and baseline forecast automation are both best practices within e2open’s planning solutions.
Stay tuned for the release of the benchmark study. In the meantime, give us a shout if you want to share your story of taming the tail… or are interested in learning more about how demand sensing or baseline forecast automation might help your organization.
3 Best Practices
- SKU Rationalization Governance
- Demand Sensing
- Baseline Statistical Forecast Automation