Demand planning is often described using examples from retail, FMCG, or consumer goods, where products sell continuously and historical sales provide a reasonably strong basis for forecasting future demand.

Industrial and B2B businesses are different.

Demand may be driven by projects, tenders, shutdowns, maintenance schedules, customer contracts, engineering requirements, or one-off orders. Many products sell only occasionally. Order quantities can vary sharply from one transaction to the next, and a large customer order can materially change the demand profile of a SKU.

For this type of business, demand planning cannot rely on statistical forecasting alone. It requires a broader planning framework that combines historical demand, demand-pattern classification, customer intelligence, project pipelines, inventory exposure, lead times, and purchasing commitments.

The objective is not simply to produce the most accurate forecast. The objective is to make better inventory and procurement decisions under uncertainty.

Why Industrial Demand Is Different

Consider a typical industrial distributor supplying fasteners, engineered components, spare parts, or project materials.

A standard product may sell every month. Another item may sell only three times per year. A specialised product may have no sales for 18 months and then suddenly receive a large project order.

All three items may be commercially important, but they require very different planning approaches.

Typical characteristics of industrial and B2B demand include:

  • intermittent customer ordering;
  • highly variable order quantities;
  • long periods of zero demand;
  • project-based purchasing;
  • customer-specific specifications;
  • long supplier lead times;
  • minimum order quantities;
  • large one-off orders;
  • tender and contract demand;
  • slow-moving but operationally critical items.

This creates a fundamental problem for traditional forecasting.

If a product sold 1,000 units last month and nothing for the previous six months, calculating a simple monthly average may produce a forecast, but it does not necessarily represent the true nature of future demand.

The planner therefore needs to understand not only how much demand occurred, but also how demand occurred.

1. Start with Demand Segmentation

A practical demand-planning process begins by segmenting the product portfolio.

Two useful approaches are ABC analysis and demand-pattern classification.

ABC Analysis

ABC analysis identifies the products with the greatest financial importance.

A traditional ABC analysis may use annual consumption value:

$$ \text{Annual Consumption Value} = \text{Annual Demand} \times \text{Unit Cost} $$

However, industrial businesses may also use alternative measures depending on the management objective, such as:

  • inventory value;
  • sales revenue;
  • gross margin;
  • outstanding purchase order value;
  • working-capital exposure.

A typical ABC structure is:

Class Typical Interpretation
A Highest-value products requiring close management
B Medium-value products requiring regular review
C Lower-value products suitable for simplified control

The purpose is not merely to label products. It is to decide where planning effort should be concentrated.

An A-class item with significant purchasing exposure should receive more scrutiny than a low-value C-class item, even if both have similar demand behaviour. (For a deeper look at combining value with predictability, see ABC-XYZ Analysis: A Practical Guide to Inventory Segmentation.)

2. Classify the Demand Pattern

ABC analysis tells us which products are financially important, but it does not tell us whether demand is predictable.

For industrial planning, demand-pattern classification is therefore equally important.

A useful classification is:

  • Smooth — frequent and stable demand
  • Erratic — frequent demand, highly variable quantities
  • Intermittent — many zero-demand periods, consistent order quantities
  • Lumpy — irregular timing combined with highly variable quantities
  • Sparse — very few demand occurrences over the historical period

Smooth Demand

Smooth items have relatively frequent and stable demand.

For example: 100, 95, 110, 102, 98, 105

These products are the closest to conventional demand-planning situations. Statistical forecasting methods such as moving averages, exponential smoothing, or ARIMA-type models may work reasonably well.

Erratic Demand

Erratic products have frequent demand, but order quantities vary significantly.

For example: 120, 450, 80, 300, 150, 500

Demand occurs regularly, but the magnitude is difficult to predict. These items usually require higher safety stock or more careful service-level management.

Intermittent Demand

Intermittent products have many zero-demand periods but relatively consistent order quantities when demand occurs.

For example: 0, 100, 0, 0, 120, 0, 110, 0

This is common in industrial distribution and spare-parts environments. Traditional forecasting models can struggle because the important question becomes not only how much will be ordered, but also when the next demand event will occur.

Lumpy Demand

Lumpy demand combines irregular timing with highly variable quantities.

For example: 0, 50, 0, 800, 0, 0, 120, 2,000

This is one of the most difficult demand patterns to forecast statistically. Industrial projects, large maintenance orders, and customer-specific requirements often create this type of pattern.

Sparse Demand

Sparse items have very few demand occurrences over the historical period.

For example: 0, 0, 0, 0, 500, 0, 0, 0, 0, 0, 0, 700

Historical data provides limited information about future demand. For these products, customer and project information may be much more valuable than statistical forecasting.

3. Combine Financial Importance and Demand Behaviour

The most useful planning framework is created by combining the two dimensions.

ABC classification answers: How financially important is this product?

Demand-pattern classification answers: How predictable is its demand?

Together, they create an exception-management framework:

Segment Typical Management Approach
A + Smooth Statistical forecasting and tight inventory control
A + Intermittent Intermittent-demand forecasting plus regular review
A + Lumpy Close planner review and customer collaboration
A + Sparse Project/customer-driven planning
C + Smooth Automated replenishment
C + Sparse Simplified control or make-to-order

This is particularly valuable because planning resources are limited. A planner should not spend the same amount of time reviewing every SKU. The highest-value and least-predictable products generally deserve the greatest management attention.

4. Forecast Differently by Demand Type

One of the biggest mistakes in industrial demand planning is applying the same forecasting method to every item.

Different demand patterns require different forecasting approaches.

Smooth Items

These are suitable for conventional time-series forecasting. Possible methods include:

  • seasonal naïve;
  • moving average;
  • exponential smoothing;
  • ARIMA/SARIMA;
  • regression with known demand drivers.

Forecast accuracy can be measured using metrics such as WAPE, MAE, RMSE, and bias. (See A Practical Guide to Forecast Error Metrics for how to use them.)

Intermittent Items

Intermittent-demand methods may be more appropriate. Examples include:

  • Croston’s method;
  • SBA;
  • TSB;
  • probabilistic demand models.

These methods separate the estimated demand size from the interval between demand events.

Lumpy and Sparse Items

For highly irregular products, statistical forecasting may provide limited value. Planning should rely more heavily on:

  • customer forecasts;
  • project schedules;
  • tender pipelines;
  • sales opportunities;
  • maintenance shutdown plans;
  • contract commitments;
  • engineering requirements.

The forecast becomes a combination of quantitative analysis and commercial intelligence.

5. Separate Forecast Demand from Known Demand

This distinction is especially important in project-driven businesses.

Not all future demand should be treated as forecast demand. There may be several levels of demand certainty:

Demand Type Example Confidence
Confirmed order Customer purchase order received Very high
Contract requirement Scheduled contract delivery High
Project forecast Customer advises likely requirement Medium
Sales opportunity Tender or quotation pipeline Low to Medium
Statistical forecast Based on historical behaviour Variable

Combining these blindly into one forecast can create confusion. A better approach is to maintain visibility of each demand source separately.

For example:

$$ \text{Future Demand} = \text{Confirmed Orders} + \text{Contract Demand} + \text{Project Forecast} + \text{Statistical Baseline} $$

The weighting or treatment of each component should reflect its level of certainty.

6. Connect Demand Planning with Procurement Exposure

Demand planning in industrial businesses cannot be separated from purchasing.

Long lead times, MOQs, international freight, and supplier production schedules mean that procurement decisions may need to be made months before demand materialises. This creates working-capital risk.

A useful management view therefore combines:

  • on-hand inventory;
  • outstanding purchase orders;
  • confirmed customer demand;
  • forecast demand;
  • supplier lead time;
  • MOQ;
  • historical demand pattern.

For example, consider an A-class sparse product with:

  • $80,000 inventory on hand;
  • $120,000 in outstanding POs;
  • no confirmed customer order;
  • only two demand events in the previous three years.

This should receive immediate management attention. The issue is no longer forecast accuracy. The real question is:

Why are we committing another $120,000 to a product with limited evidence of future demand?

This is where demand planning becomes a working-capital discipline rather than simply a forecasting exercise.

7. Use Exception-Based Planning

Industrial product portfolios may contain thousands or tens of thousands of SKUs. Manually reviewing every product every month is inefficient.

A better approach is exception-based planning. The system should identify products where management attention is required.

Possible exception rules include:

  • high-value product with sparse demand;
  • high outstanding PO value with no confirmed demand;
  • forecast significantly above historical demand;
  • inventory coverage above target;
  • repeated stockouts;
  • supplier lead-time increase;
  • large forecast bias;
  • new project demand;
  • large customer order outside normal demand;
  • obsolete or slow-moving stock.

This allows planners to focus on the minority of items where decisions can materially affect service level or working capital. (I’ve written a separate guide on designing these reports: Demand Exception Reports: Purpose, Review Process, and Practical Challenges.)

8. Forecast Accuracy Is Not the Only KPI

In retail demand planning, forecast accuracy is often one of the central KPIs. In industrial planning, it remains important, but it should not be viewed in isolation.

Other important measures include:

  • forecast bias;
  • service level;
  • fill rate;
  • inventory turnover;
  • days of inventory;
  • excess and obsolete stock;
  • stockout frequency;
  • working-capital exposure;
  • purchase-order coverage;
  • supplier lead-time performance.

For intermittent and project-driven items, poor forecast accuracy may sometimes be unavoidable. The more important question is whether the planning process makes good decisions despite uncertainty.

For example, a planner may not predict the exact month when a customer will place a project order, but can still ensure that the business understands the potential requirement, supplier lead time, inventory exposure, and procurement decision window.

9. Sales and Customer Collaboration Become Critical

For industrial demand, the sales team often holds information that cannot be found in historical data.

Examples include:

  • upcoming projects;
  • customer shutdowns;
  • tender outcomes;
  • competitor changes;
  • expected contract renewals;
  • engineering changes;
  • product substitutions.

A strong demand-planning process therefore requires structured collaboration between Sales, Procurement, Operations, and Finance. This is one reason why S&OP or IBP processes are particularly valuable.

The monthly demand review should not simply ask: What does the statistical forecast say?

It should ask: What has changed in the market, customer pipeline, projects, and supply environment since the last planning cycle?

10. A Practical Planning Framework

For an industrial or project-driven business, a practical demand-planning process can be structured as follows.

Step 1 — Segment the portfolio. Classify products using ABC financial importance, demand pattern, customer importance, lead time, and criticality where relevant.

Step 2 — Establish a baseline. Generate a statistical baseline for products with sufficient historical demand. For intermittent products, use appropriate intermittent-demand methods.

Step 3 — Add known future demand. Incorporate confirmed customer orders, contracts, projects, tenders, and maintenance requirements. Keep different demand sources visible rather than hiding them inside one number.

Step 4 — Review exceptions. Prioritise high-value products with unusual forecast changes, large outstanding POs, low demand frequency, excessive inventory, or customer risk.

Step 5 — Reconcile with supply. Compare demand against inventory, open purchase orders, supplier capacity, MOQ, production lead time, and freight lead time.

Step 6 — Make a management decision. Possible decisions include: place the PO, delay the PO, reduce quantity, expedite supply, transfer inventory, request customer confirmation, convert to make-to-order, or accept calculated safety-stock exposure.

Conclusion

Demand planning in industrial, B2B, and project-driven businesses is fundamentally different from forecasting high-frequency consumer demand.

Historical data remains valuable, but it is only one part of the planning process. The most effective approach combines:

  • financial segmentation;
  • demand-pattern analysis;
  • statistical forecasting;
  • customer and project intelligence;
  • inventory visibility;
  • procurement exposure;
  • exception-based management.

The central lesson is simple: do not use the same planning method for every SKU.

A stable, high-volume product can be managed statistically. A specialised project item with three demand events in four years cannot.

For industrial businesses, good demand planning is therefore less about generating a single perfect forecast and more about matching the planning method to the nature of demand, focusing management attention on financial risk, and making better procurement decisions under uncertainty.