Why CPG Demand Forecasting Has Hit a Ceiling
Consumer package goods (CPG) companies are looking for growth and scouring markets for opportunity. Their product portfolios grow with new product introductions, fresh takes on existing merchandise, and multiplying sales channels. Boston Consulting Group talks about how the “endless aisles” of the internet and omni-channel sales increase product assortment. Trade promotion spending aimed at pumping up sales continues at a dizzying pace.
The goal of this revenue chase is to grow demand— but this path to growth comes at a cost— the cost of demand complexity. And complexity creates a challenge of how to forecast accurately when faced with new items, new channels and demand shaping. Recent evidence strongly suggests that traditional forecasting techniques in this environment have reached their limits and hit a ceiling.
- The Boston Consulting Group said at the 2015 Grocery Manufacturers Association/Food Marketing Institute (GMA/FMI) supply chain conference that 22% of polled respondents said lessaccurate forecasts were one of their top hurdles to improving service.
- Across the CPG industry, demand forecast errors increased in 2014 and were unchanged in 2015, according to Gartner analyst Steve Steutermann. He adds, “Using a traditional, 30-day lag period and a mean absolute percent error (MAPE) unit/location forecasting measure, Gartner often hears from companies with forecast accuracy percentages in the 50s and 60s, with far fewer reporting forecast accuracy in the 70s.”
- The problem is even more acute with long-tail items, where demand can range from less than half to more than twice the original forecast. Many CPG companies are at a loss on what to work on to improve these forecasts. They don’t know how to segregate, forecast or inventory these “extreme error” items.
A consensus is forming that to break through the ceiling and forecast more accurately, CPG companies need to go beyond traditional processes. Here are three approaches that have gained traction.
Sense Demand
“Demand Sensing” involves processing demand signals closer to the actual sales event. It uses consumption data to generate daily statistical forecasts with shorter lag. Advanced demand analytics like pattern recognition and machine learning can help to gain better insight from this downstream demand data to improve demand projections and near-term forecasts.
Steutermann calls this “shifting from sell-in to sell-through”. In Use Benchmarking to Target and Improve Key Capabilities in Consumer Products he says, “The ability to use customer data to reduce forecast error and measure accuracy in shorter lag periods is a distinguishing capability of demand management leaders and allows companies to reduce their demand latency (the time it takes to understand what happened at the shelf and translate those demand changes into a supply chain response).”
We have seen hundreds of companies benefit from leveraging downstream data. We have written about the concept in previous blogs and also profiled two firms, Amplifon (Miracle Ear) and Costa Coffee. Even the UK’s National Health Service (NHS) is now gathering blood demand data from downstream hospitals.
Leverage Inventory More Effectively
Sensing demand leads to a second common tactic that benefits directly from leveraging downstream demand. Firms are incorporating a pull-based replenishment approach or an inventory “pull strategy” that can operate with less inventory, more stable supply operations and improved cash flow.
With forecasts stuck in neutral (or reverse) it’s understandable that many CPG companies hold excess safety stock, often with the wrong mix, to make sure growth won’t be stymied by unfilled demand. But this moves KPIs in the wrong direction. Add in capacity constraints, transportation bottlenecks, and promotional uplift and there is a need for better inventory planning and optimization across the supply chain network. A pull strategy combined with inventory optimization that focuses on the safety buffers needed to meet customer service goals, helps achieve corporate targets.
The Costa and the NHS examples cited above both employ pull-driven supply networks.
Establish a Baseline Forecast
Another growing tactic is establishing a statistical baseline forecast that understands and segregates baseline demand, illuminating the difference between nominal and promotional demand. The baseline forecast is augmented with inputs from sales, marketing, and channels to build a consensus demand plan that improves forecast accuracy and the ability to achieve the financial plan. We recently wrote about one company, RS Components, with a similar forecasting process.
Many firms benchmark and measure this plan to validate its accuracy and identify areas for improvement by measuring Forecast Value Added (FVA)—any change in forecast performance that can be ascribed to a process step or a participant’s input. Identifying the sources of error in the forecast builds greater accountability for the forecast.
Most CPG companies have hit a demand forecasting ceiling. The good ones are breaking through.
Click below a short whitepaper on leveraging supply chain data for improved forecasting.