Consulting · Demand planning
Demand planning: forecasts your planning can rely on
We structure demand planning and sales forecasting so that statistical methods and market knowledge come together. Forecast quality becomes measurable, assumptions become traceable.
Demand planning
What it is about
Demand planning is the structured forecast of future customer demand. It is based on sales history, market knowledge and input from sales and marketing. The result, the agreed demand plan, drives production, procurement and inventory and is the starting point of every S&OP cycle.
Good demand planning is not about the one right method. It emerges when the portfolio is segmented sensibly, the data basis is sound, each method fits its demand pattern and manual adjustments demonstrably add value.
Challenges
Typical starting points
Optimistic sales forecast
Numbers from sales are often targets rather than forecasts. The systematic surplus ends up in the warehouse.
Unclean history
Promotions, outliers, supply shortages and successor items distort the data every forecast builds on.
Intermittent demand
Spare parts and product variety create demand with many zero periods. Standard methods regularly miss here.
Accuracy is not measured
Without a fixed metric, level and lag it remains unclear whether the forecast is getting better or worse.
Planners as data collectors
Most of the time goes into gathering numbers instead of analysis and alignment.
High expectations of AI
Machine learning is supposed to rescue the forecast, but data and process are not ready for it yet.
Project
How we work
- 01
Diagnosis and segmentation
We analyse sales history and forecast quality and segment the portfolio, for example using ABC/XYZ logic. Today's forecast accuracy becomes the baseline.
- 02
Cleanse the data
Outliers, promotions, supply failures and product successions are cleansed or flagged so that the methods see true demand.
- 03
Methods per segment
For each segment we select suitable methods and backtest them against actual historical sales.
- 04
Consensus process
Sales and marketing enrich the statistical forecast with market knowledge in a structured, documented way. Forecast value added shows which adjustments improve the plan.
- 05
Automation and system
We anchor methods and process in the planning system, from a structured solution to SAP IBP for Demand.
- 06
Monitoring and enablement
KPIs, exception lists and training keep quality high in day-to-day work.
Methods
Forecasting methods at a glance
Exponential smoothing
Robust for stable items; the Holt-Winters variant also handles trend and seasonality.
Croston and TSB
For intermittent demand with many zero periods, such as spare parts.
Regression with indicators
Explains sales through drivers such as prices, promotions or leading economic indicators.
Machine learning
Gradient boosting and related models, when enough data and drivers are available.
Demand sensing
Corrects the forecast for the next days to weeks based on current signals such as incoming orders.
Best-fit selection
Methods compete in backtesting. The one that demonstrably performs best for the segment is used.
Impact
What can be measured
Forecast quality can only improve if it is measured consistently: at a fixed level, with a fixed lag and over a fixed period. These are the KPIs we typically use:
- Forecast accuracy as 1 minus WMAPE
- Bias or tracking signal for systematic over- or underestimation
- Forecast value added (FVA) per process step compared with a naive forecast
- Service level, delivery readiness and days of supply
Research shows why FVA matters: in an evaluation of more than 60,000 forecasts from four companies, small manual adjustments often reduced accuracy, and upward adjustments were wrong more often than downward ones (Fildes et al., 2009).
FAQ
Frequently asked questions about demand planning
What is demand planning?
Demand planning is the structured forecast of future customer demand from sales history, market knowledge and sales input. It is the basis for production, procurement and inventory planning.
What is the difference between sales planning and demand planning?
Sales planning describes the volumes and revenue sales expects. Demand planning translates this expectation into the requirements that production and procurement have to cover. In practice the terms are often used interchangeably.
How is forecast accuracy calculated?
A common measure is 1 minus WMAPE: the sum of absolute deviations divided by actual sales. A fixed planning level, a fixed lag between forecast and actuals and a fixed period are essential.
What is good forecast accuracy?
That depends on industry, planning level and horizon. Comparing against a naive forecast is more meaningful than a flat target: every step in the process should beat it.
Which forecasting methods are there?
Moving averages, exponential smoothing, Holt-Winters for trend and seasonality, Croston for intermittent demand, and regression and machine learning methods. The demand pattern decides which one fits.
What is demand sensing?
Demand sensing corrects the forecast in the short term, for days to a few weeks, based on current signals such as incoming orders or sell-out data.
Does AI improve sales forecasts?
Often it does: McKinsey reports 20 to 50 percent lower forecast errors with AI-based methods. Prerequisites are clean data and items whose demand can be explained at all.
More consulting topics