Demand forecasting for industrial goods

Demand forecasting for industrial goods separates repeat purchasing, replacement needs and project orders before comparing them with the supplied forecast. Use the same product, customer, period and order definition throughout. The result should explain which assumptions changed demand, rather than simply replace the previous total with a new one.

Choose the order measure before choosing assumptions

For an industrial distributor, a useful starting measure is accepted customer orders, net of cancellations, by order-entry month. Keep requested delivery dates as another field. An accepted order for future delivery contributes to current order intake but does not represent a current shipment.

The Census Bureau's manufacturing definitions describe new orders separately from unfilled orders and shipments. Its new-order definition also distinguishes documented buying commitments from informal interest. Use your company's booking policy consistently rather than treating every quotation as an order.

Start with the broader demand forecasting guide if the company has not defined its planning grain. This article focuses on reconciling industrial order drivers after that definition exists.

Assemble the workbook from existing exports

Use the forecast snapshot that was available before the planning period. From the order export, retain product, customer, order identifier, order-entry date, requested delivery date, ordered units and canceled units. Build actual net orders from those rows using the same cancellation timing policy each period.

Add a separate assumptions sheet for the commercial team's repeat purchases, replacements and projects. Each assumption needs a period, quantity, owner and supporting reference. A maintenance schedule is different evidence from an unaccepted quotation; preserve that distinction.

Set the categories before adding quantities:

• Repeat purchasing: expected recurring demand that excludes the separately listed replacements and projects.
• Replacement needs: maintenance or replacement quantities absent from the repeat baseline.
• Project orders: incremental quantities associated with a named customer project, excluding items already captured above.

Customer production activity may help challenge the repeat baseline. The Federal Reserve's G.17 methodology explains that its industrial production index measures real output. It is a useful external context series, but it is neither your customer's purchase order nor evidence of your share of its purchases.

Calculate a transparent demand scenario

Driver-based forecast = repeat purchases + replacement needs + project quantities

A fictional US distributor forecasts orders for the coming month. Its supplied driver assumptions exclude overlaps, and the products are counted in their selling units. This is a unit forecast; a revenue view would also require prices in USD.

With those columns in A:G, enter the scenario in E2 and fill down through row 4:

=SUM(B2:D2)

The component check is 960 + 260 + 340 = 1,560 units. The row check gives the same result: 600 + 540 + 420 = 1,560. Tie actual net orders of 1,540 units to the filtered order export, not to invoices or shipments.

The assumptions also need an overlap check. Match their supporting order or project references against the repeat baseline. A correct addition formula cannot detect a customer project already embedded in the recurring quantity.

Work the same actual rows against both forecasts

Use forecast minus actual as the sign convention. Positive means the forecast was too high; negative means it was too low.

In the first table, add H:K for supplied signed miss, supplied absolute miss, driver signed miss and driver absolute miss:

=F2-G2
=ABS(H2)
=E2-G2
=ABS(J2)

Fill down. The aggregate absolute-error formulas are:

=SUM(I2:I4)/SUM(G2:G4)
=SUM(K2:K4)/SUM(G2:G4)

The supplied forecast gives 80 / 1,540 = 5.2%; the driver scenario gives 20 / 1,540 = 1.3%, rounded to one decimal place. Both totals miss actual orders by 20 units in opposite directions, but the supplied forecast contains larger offsetting product misses.

This single month illustrates the calculation; it does not prove the driver approach will remain better. Compare repeated periods at the same horizon, as the forecast accuracy guide explains.

Where industrial forecasts go wrong

• Project demand is counted twice. Reconcile confirmed project lines with the repeat baseline before adding an incremental allowance.
• Maintenance timing becomes purchase timing. A planned outage may be delayed, and its parts may already be in the customer's stock. Check what the quantity actually represents.
• Cancellations disappear from actuals. Apply one documented net-order policy so a revised export does not silently change the historical comparison.
• Totals conceal the product problem. Review signed and absolute misses separately. Investigate persistent direction using demand forecast bias by SKU and site.

For spare parts with many periods without orders, use the separate intermittent-demand forecasting guide. A steady repeat-purchase baseline is not appropriate merely because the part belongs to an industrial catalog.

Compare your industrial demand file in Covirage

Upload a file or Google Sheet containing supplied forecasts, actual net orders and the documented driver scenario. Covirage maps the columns, and its tools compute and check the comparisons; the external AI model explains the results with every figure traced to its rows. Create a free Covirage account, which takes about a minute, then set up your business and bring a sample of your own data.

See analytics for industrial distributors for the supporting industry context.