What this module covers
This is the first module of the Planning Foundations for Manufacturing course. It covers the basics: what demand planning is, how it differs from forecasting, who the stakeholders are, and the vocabulary that the rest of the course uses. If you are new to planning, start here. If you have been planning for years, skim this module to confirm we are using the same words the same way, then move to Module 2.
What demand planning actually is
Demand planning is the process of predicting what customers will buy, in what quantities, at what locations, over what time horizon, and turning that prediction into a number that the rest of the business can plan around. It is not the same as forecasting, although forecasting is the core technical activity within demand planning. Forecasting produces a statistical prediction. Demand planning wraps that prediction in overrides, consensus, approval, and publication, producing a single number that finance, operations, supply chain, and sales all agree to use.
The distinction matters because the forecast is usually right (within its confidence band) and the published demand plan is often wrong (because of the overrides and consensus process). When the business misses a target, the demand plan gets blamed, not the forecast. Understanding the gap between the two is the first step in diagnosing planning failures.
The stakeholders
Demand planning sits at the intersection of four functions, each with a different interest in the number.
Sales wants the demand plan to be high enough that inventory will be available to sell, but not so high that finance flags revenue risk. Sales is closest to the customer and has the best intelligence on upcoming promotions, customer pipeline, and competitive activity. Sales is also the most optimistic function, which is why their inputs are valuable but need to be checked.
Operations wants the demand plan to be stable, because stability makes manufacturing efficient. Operations cares about the rate of change in the plan more than the absolute level, because changeovers, capacity adjustments, and material procurement all depend on plan stability. Operations will push for smoother plans even when the underlying demand is volatile.
Finance wants the demand plan to be accurate, because accuracy drives revenue forecasting, working capital planning, and investor guidance. Finance is the function that will measure forecast accuracy after the fact and hold the demand planner accountable for it. Finance cares about bias (systematic over- or under-forecasting) more than random error.
Supply chain wants the demand plan to be granular (per SKU, per location, per period), because supply planning operates at that granularity. Supply chain will consume the demand plan as the input to inventory policy, replenishment, and distribution planning. Aggregated demand plans (per product family, per month) are not useful for supply chain, even if they are easier to forecast accurately.
The demand planner's job is to produce a single number that all four functions can plan around, despite their different interests. This requires both technical skill (producing a good statistical forecast) and political skill (navigating the consensus process). The rest of this course covers both.
The vocabulary
These terms are used throughout the course. Definitions vary slightly across organizations, so we are stating ours.
Forecast. The statistical prediction of future demand, produced by a model (statistical, machine learning, or ensemble) applied to historical data. The forecast is the starting point, not the final number.
Demand plan. The published, approved prediction of demand that the business will plan around. The demand plan is the forecast after overrides, consensus, and approval. It is the number that operations builds to and finance recognizes.
Override. A manual change to the statistical forecast, applied by a planner or stakeholder, with a stated reason. Overrides can improve or degrade accuracy. They are tracked for forecast value added analysis.
Consensus. The process of reconciling differing demand views (sales, marketing, operations) into a single demand plan. Consensus is a process, not a number. The output of consensus is the demand plan.
Baseline. The unmodified statistical forecast, before overrides. The baseline is the reference point for measuring whether overrides helped or hurt.
MAPE. Mean Absolute Percentage Error. The most common forecast accuracy metric. Covered in depth in Module 3.
Bias. The systematic tendency to over- or under-forecast. A forecast with low MAPE but high bias is less useful than a forecast with higher MAPE but zero bias, because bias can be corrected and random error cannot.
Forecast horizon. The number of periods ahead the forecast covers. A 13-week horizon covers one quarter. A 12-month horizon covers a year. Accuracy degrades with horizon.
Forecast granularity. The level at which the forecast is computed. SKU-location-week is the finest practical granularity. Product family-region-month is coarser. Finer granularity is more useful for planning but less accurate, because there is less data per series.
How demand planning connects to the rest of the business
Demand planning is the first link in the supply chain planning chain. The demand plan feeds supply planning (what do we need to make or buy), which feeds production scheduling (when do we make it), which feeds distribution planning (where do we send it). Each link uses the demand plan as input, and each link's effectiveness depends on the demand plan's quality.
This is why demand planning matters more than its size as an activity would suggest. A demand planning team might be 5 people in a 500-person operations organization. But if the demand plan is wrong, the production schedule is wrong, the inventory is wrong, and the customer service level is wrong. The leverage of demand planning on the rest of the business is enormous.
The demand plan also connects upward, to finance. Revenue forecasts, working capital projections, and investor guidance all start from the demand plan. A demand plan that is systematically biased high will produce revenue forecasts that are systematically missed, which damages credibility with investors and the board. A demand plan that is systematically biased low will produce conservative revenue forecasts that are beaten, which is better than missing but still indicates a planning problem.
What good demand planning looks like
By the end of this course, you will know how to evaluate a demand planning function. For now, here are the four properties of a good demand planning process.
The statistical baseline is the default. Overrides are applied only where there is a stated reason, and the override rate is tracked. A process where 80% of SKUs are overridden is a process that does not trust the forecast, which usually means the forecast methodology needs improvement, not more overrides.
Accuracy is measured at the right level. MAPE at SKU-location-week for a 4-week-ahead horizon is the standard measurement. Comparing accuracy at different levels (product family vs SKU) or different horizons (1 week vs 6 months) is meaningless. Always state the level and horizon.
Bias is tracked and corrected. A good demand planning process knows its bias (overall, per category, per planner) and takes corrective action when bias persists. Ignoring bias is the most common planning failure.
The demand plan is published and locked. After consensus and approval, the demand plan is published and not changed until the next cycle. Post-publication edits destroy trust in the process and make accuracy measurement impossible, because you are measuring against a moving target.
What to do next
This module established the vocabulary and context. Module 2 covers statistical forecasting models: exponential smoothing, ARIMA, and when to use each. Module 3 covers accuracy metrics in depth. Module 4 covers the S&OP process that wraps the forecast.
Before moving to Module 2, take five minutes to write down, for your own organization, who owns the demand plan, what level of granularity it is published at, and what the last measured accuracy was. If you cannot answer these three questions, the rest of this course will help you figure out why, and what to do about it.
Self-assessment
Question 1. What is the difference between a forecast and a demand plan?
Question 2. Name the four stakeholder functions in demand planning and one thing each cares about.
Question 3. Why is bias a more actionable metric than MAPE?
Question 4. Why does forecast accuracy degrade with horizon?
Question 5. What does "published and locked" mean for a demand plan, and why does it matter?
Answers are in the next module's introduction. If you can answer all five without looking them up, you are ready for Module 2.