Skip to content 
Background
A leading manufacturer of residential ceiling fans experienced a surge of demand for some of its most popular models, leading to stockouts and backorders for its distributors.
They responded to demand by ramping up production, only to see an equally dramatic decline in demand six months later. They then had bloated inventory of ceiling fans and excess storage costs. Eventually, the company sold their excess inventory at a discounted price to normalize their stock levels.
Client Question:
What prevented us from anticipating changes in demand for ceiling fans? What changes should we make to our forecast model going forward?
Our Approach
First: We Review Data
The client provided actual SKU-level sales detail for past 36 months. They also provided historical forecasts from their demand planner.
We calculated the actual error for the demand plan, and we created a trend-line of the error bias over time.
Next: We Analyze
We applied statistical analysis to historical sales to find a better forecasting technique going forward.
We compared forecast models from popular demand planning software with our own in-house statistical modeling. Further, we researched external market and economic indicators to identify predictors of demand.
Our Learnings
The client’s demand plan typically averaged 70% accuracy. This level of accuracy was sufficient to avoid stockouts and excess inventory.
However, planning accuracy declined to below 50% by the middle of 2020, with a bias that understated sales by 30%. The client couldn’t keep up with demand, and had painful stockouts for most of its models.
Near the end of 2020, this reversed: Demand fell sharply, and the client’s demand plan overstated sales by 30%. As a result, the client ended the year with bloated inventory amid declining sales.
The client’s demand planning forecast relied upon a rolling twelve month average to predict future demand. This approach understates recent volatility in forecasting, as recent demand fluctuations are simply averaged with the prior year results.
Our Solution
New Forecast Model
We built a forecast model that gave higher weight to recent fluctuations in demand. We also modeled a precise seasonality index, to reflect the typical spring/summer selling seasons for ceiling fans.
New Market Predictors
Google search trends are a strong predictor of near-term demand for many consumer products, such as ceiling fans. We applied Google search trends for relevant keywords to our new forecast model, which yielded even higher forecast accuracy.
Reporting and Training
We built customized, repeatable tools for measuring and reporting on forecast accuracy. The client’s demand planner was less familiar with statistical approaches. We guided her through the new forecast model, and offered ongoing support for a duration of time.
Outcome
The client spent nearly a year trying to understand the problems with their planning cycle. It took us a total of three weeks to digest the data, perform statistical analysis, and provide a root cause for the planning errors.
We gave them a robust forecast model that better anticipates future demand for each model of their ceiling fans.
We also provided our custom-built reporting solution so they can monitor and correct for planning errors on a regular basis. And we stay in touch.
Our client now spends less time managing surplus stock, and more time building the latest models of ceiling fans.
