Demand Planning Analyst, Supply Chain
Forecasting models are standard — and the best one tested still missed by 56%.
Research by Dominique Genovesse
Typical employers Retail · manufacturing · sales and operations functions
What the job is
A demand planning analyst uses data to forecast future product demand, guiding decisions about inventory, production, and supply. Day to day that means structuring daily, monthly, and yearly forecasts, analysing sales numbers, managing current inventory, and presenting findings to the operations unit.
How AI is used today
Forecasting has been substantially reshaped by machine learning, to the point that some practitioners treat it as standard practice in the field — via tools including Microsoft Power BI, Excel Copilot, and Blue Yonder. These systems combine a business's historical data with additional factors such as fluctuating prices, environmental conditions, and social signals. McKinsey's analysis describes demand planning as having "been revolutionized by integrating machine learning and harnessing new sources of data," with firms reporting improvements in inventory levels, speed and quality of service, and logistics costs.
The Mean Absolute Percentage Error for the XGBOOST was 56.18%.
Hossain & Parvin (2025) — the highest error of any model they tested, and a reminder that these systems are not uniformly better than people
What’s changing
The biggest change is a shift in the task mix: less manual input work like spreadsheet editing and SKU-level forecasting, and more oversight to validate what the models produce. That shift needs watching, because the pressure to adopt is partly a talent problem rather than a capability one — McKinsey's 2024 survey of companies pursuing this integration found that "90 percent say that their companies lack sufficient talent to meet their digitization goals." Looking forward, as models improve, forecasting accuracy may rise and in turn invite more dependence and less human oversight.
What a new entrant needs
Beyond baseline fluency with SQL and data interpretation, an entry-level analyst has to understand the blind spots of the models they are using — and be able to validate and in some cases override the forecasts those models produce. This connects directly to Critical Thinking and Judgement, which emphasizes applying routine evaluation methods to verify sources, letting analysts exercise better judgment when drawing conclusions from new data.
Where the evidence disagrees
The models are not uniformly better than human forecasting, and some carry high error rates. A peer-reviewed comparison of statistical and machine-learning models for predicting restaurant demand found "the Mean Absolute Percentage Error for the XGBOOST was 56.18%, which is the highest among the machine learning models," with the model struggling to capture fluctuations in sales of certain products. An analyst has to know which products a given model cannot handle and forecast those independently.
3 sources, located and verified by the contributor
- Alicke, K., Dilda, V., Görner, S., Mori, L., Rebuffel, P., Reitr, S., & Samek, R. (2021). Succeeding in the AI supply-chain revolution. McKinsey & Company.
- Alicke, K., Foster, T., & Trautwein, V. (2024). Supply chains: Still vulnerable. McKinsey Global Supply Chain Leader Survey 2024.
- Hossain, M. S., & Parvin, F. (2025). A comparative study of various statistical and machine learning models for predicting restaurant demand in Bangladesh. PLOS ONE, 20, e0325449.