Demand forecasting over product life cycle stages

Demand forecasting over the product life cycle means changing the assumptions you review as an item moves through launch, growth, maturity and decline. Compare forecast misses within each stage before changing the whole portfolio plan. A balanced company total can conceal underforecast growth products and overforecast items being phased out.

Define stages before scoring the forecasts

The OpenStax product life-cycle framework describes introduction, growth, maturity and decline. It associates maturity with slowing sales growth and decline with a significant reduction in sales and profitability. Use that framework as a commercial classification, not a clock that gives every product a fixed lifespan.

For the export, label introduction as launch if that matches your team's vocabulary. Record a stage effective date and the evidence behind it. A product may have different positions in different channels, so classify at the level where the forecast and purchasing decision are made.

Sales or product management owns the commercial stage assessment. Demand planning owns the dated forecast and error calculation. Procurement reviews the resulting supply exposure. Someone should be able to explain a stage change without pointing only to the forecast miss that followed it.

Build one matched comparison file

Bring forecast issue date, target period, SKU, site, stage at issue, forecast units, actual demand units and availability flags. Choose a common forecast horizon, such as the buying window used for the decision, and compare forecasts that were equally far from their target period.

Join forecast and actual exports by product, site and target period. Check unmatched rows separately. A missing actual is not automatically zero, and a new product should not disappear simply because it lacks a prior-year match. Preserve stage at issue rather than replacing it with today's stage.

For launch assumptions without usable history, use new product demand forecasting. The recurring baseline workflow belongs in demand forecasting for existing products.

Work the same period across stages

A US industrial-supply distributor compares saved forecasts for four products over the same selling period and at the same forecast horizon. All products were available. Actual units therefore provide a comparable demand measure for this example; the stage labels were recorded before the period began.

The worksheet uses A:G, with data in rows 2:5. Error means forecast minus actual. Percentages are rounded to one decimal place; calculations use unrounded values.

In E2:G2, filled down:

=C2-D2
=ABS(E2)
=IF(D2=0,"",F2/D2)

Format G as a percentage. If actual demand is zero, retain the unit error and flag the row for review rather than presenting an undefined percentage as zero accuracy loss.

Prove the total, then inspect the stages

The portfolio signed error is 530 - 540 = -10 units, matching -20 - 30 + 10 + 30. Its bias is -10 / 540 = -1.9%. That near-balanced total does not describe the underlying problems.

Absolute errors sum to 20 + 30 + 10 + 30 = 90 units. Portfolio weighted absolute percentage error is 90 / 540 = 16.7%. Calculate it from the totals, not the average of the four row percentages:

=SUM(E2:E5)/SUM(D2:D5)
=SUM(F2:F5)/SUM(D2:D5)
=SUM(C2:C5)-SUM(D2:D5)-SUM(E2:E5)

The last formula must return zero. In a longer file, Microsoft's SUMIFS function adds rows meeting specified criteria. Calculate growth-stage absolute error divided by growth-stage actual units with:

=IF(SUMIFS(D2:D5,B2:B5,"Growth")=0,"",SUMIFS(F2:F5,B2:B5,"Growth")/SUMIFS(D2:D5,B2:B5,"Growth"))

Add site and forecast horizon as further criteria when the export contains mixed groups. For repeated directional misses, use forecast bias by SKU and site alongside the stage split.

Change the next decision, not the scored history

The growth item was underforecast by 30 units. Check whether additional locations opened sooner than expected or comparable locations ordered more. Those explanations imply different next-period assumptions. The observed miss alone does not justify applying the same uplift to every growing product.

The declining item was overforecast by 30 units. Check the replacement schedule and remaining customer commitments before approving another order. A faster withdrawal could support a lower future demand assumption, while a postponed replacement might support a temporary extension.

Choose review windows around these decisions. Launch and growth assumptions need revisiting when adoption or distribution evidence changes. Mature products need comparable seasonal and promotional reviews. Declining products need reviews around last-order dates and commitments. Record the evidence, owner and revised assumption beside the next forecast version.

Where it goes wrong

• Relabeling history. Today's decline label cannot replace the maturity label known when an earlier forecast was issued.
• Confusing stockouts with decline. Falling shipments during unavailable periods do not establish falling demand.
• Mixing horizons. A late update and a forecast made before purchasing should not receive the same accuracy interpretation.
• Using portfolio bias alone. Opposite misses cancel even when both create costly inventory decisions.

Covirage is not affiliated with Microsoft.

Review your product stages in Covirage

Upload a sample of your own dated forecasts, demand export and stage register. Covirage maps the columns, its tools compute and check the comparisons, and the external AI model explains the result with every figure traced to its rows. Create a free Covirage account and set up your business before bringing your files.

See forecast analysis in Covirage for the supporting workflow.