How to improve demand forecasts for new product families

Editor’s note: The following article is a joint project of professors in the University of Tennessee Knoxville’s Global Supply Chain Institute. For more on the Institute, click here.

Demand planners struggle with forecasting demand for new product families. Initial demand forecasts often deviate from actual sales, which can lead to disruptions in supply chain processes and major headaches. One of the biggest pain points is determining appropriate inventory levels and replenishment schedules once the initial product rollout is complete, as supply chain planners struggle to accurately predict individual SKU demand within the product family. Lance Saunders, Jason Merrick, Chad Autry, and Mary Holcolmb address this issue in a recent article.

Apple’s launch of the iPhone X product family illustrates just this point. Despite an accurate overall forecast at the family level, discrepancies emerged at the SKU level, with surpluses for some models and stockouts for others. This mismatch arose because planners underestimated the variability in demand between different SKUs. This led to inefficiencies in inventory, logistics, and transportation that negatively impacted Apple's returns on assets.

 A major reason for this was SKU-level spread bias, which refers to the judgmental error planners make when allocating family-level demand to individual SKUs. As a result of this bias, planners may underestimate the range between the highest and lowest demand SKUs, overestimating demand for lower-performing SKUs and underestimating demand for higher-performing ones. This type of forecast error is particularly harmful in new product family launches because it distorts inventory and logistics planning, leading to higher costs and missed sales opportunities.

What can managers do about this?  Planners should recognize that forecasting for new product families often involves this “SKU-level spread bias.”  For this reason, it is essential to differentiate the demand planning process for new product families from that of existing products. This recognition allows for a more tailored approach that can better account for the inherent uncertainties in new product launches. Using multiple estimates (low, medium, high) for SKU-level demand rather than relying on single-point estimates is key. This approach helps capture the variability in demand and reduces the likelihood of significant forecast errors.

Planners can then use sales cycles to refine their forecasts to define demand patterns. Early demand signals, such as initial replenishment orders from major retailers, can provide valuable insights into SKU-level performance. These signals can inform adjustments to forecasts and inventory levels as actual demand begins to materialize. Alternatively, planners may postpone planning decisions until more accurate SKU-level demand data is available. This could involve delaying final procurement and production decisions until the demand for individual SKUs becomes clearer. Planners may also consider using common components that allow for aggregated forecasting for supplier orders. This strategy can help mitigate the impact of SKU-level spread bias by leveraging shared demand forecasts across related products.

At the outset of a family product launch, planners should ensure that safety stock levels are aligned with their understanding of SKU-level demand distribution. Avoid over-padding forecasts for lower-demand SKUs, which can lead to excess inventory, while also ensuring that high-demand SKUs are sufficiently stocked to prevent stockouts.

Collaboration and training can increase information flow and understanding. When managers foster collaboration among forecasting, production, and marketing teams, they can align on demand expectations and share insights across various functions. This is likely to lead to a more comprehensive understanding of market dynamics and consumer behavior. Also, training planners and managers can help them understand the cognitive biases involved in forecasting and how to mitigate them.

According to lead researcher Lance Saunders, “this research illustrates a recurring theme that we have identified through collaborating with companies about improving their supply chain planning process……planning should not be one size fits all. While in this case it is differentiating the planning process for new products from that of existing products, being aware of how to segment the planning process in a way that makes sense for your supply chain is an opportunity for most companies. This is the last research project that we worked on with the late Mary Holcomb, who had a huge influence on our team and many others.”

The road to accurate planning for new product families is not easy to follow, but there are ways to mitigate the inaccuracies that can have an impact on the bottom line. Such planning requires a nuanced approach that addresses the unique challenges posed by SKU-level spread bias. By implementing the managerial strategies outlined above, firms can enhance their forecasting accuracy, reduce stockouts and excess inventory, and improve overall supply chain performance.

About the Global Supply Chain Institute

The University of Tennessee's Global Supply Chain Institute (GSCI) is the preeminent hub for supply chain thought leadership and talent development. The pinnacle of GSCI’s efforts is UT’s Supply Chain Forum, which brings together more than 80 of the world’s most innovative and impactful companies twice a year to learn, network, and recruit the country’s top supply chain talent. 

SC
MR