Blog · Forecast and pipeline · Supply chain
Build a demand range from confirmed orders and explicit assumptions when usable sales history is missing. Keep a dated estimate, then compare it with initial orders and actual sales.
Demand forecasting without historical data means building an explicit estimate from what you know now, then testing it as orders and sales arrive. Start with confirmed quantities and add a separate assumption for demand that is not yet committed. Keep low, base, and high scenarios visible instead of presenting one unsupported number as certainty.
Missing history can mean a new product, a new territory, an unavailable legacy file, or sales records that cannot be compared with the new business. Write down which problem applies. A forecast for a newly opened US distribution region might still use evidence from existing customers without pretending the region has its own sales history.
Define the item, location, selling unit, period, and cutoff date. Decide whether the forecast concerns customer orders or completed sales. This example forecasts completed unit sales; confirmed orders are scheduled to ship in the stated weeks, and the example assumes no supply restriction or cancellation.
Forecasting: Principles and Practice identifies judgmental forecasting as an option when historical observations are unavailable and recommends a documented, systematic process. That supports a transparent business estimate, not a claim that analytics can create the missing history.
For choosing comparable launches and measuring buyer adoption, see new product demand forecasting. Here the job is to keep evidence, commitments, and scenarios consistent as observations arrive.
Export order-line ID, product ID, location, requested ship date, status, and remaining quantity. Preserve the export's timestamp. Remove duplicate lines before totaling quantities and distinguish a new order from a later status update to the same order.
Use the requested ship period for this sales estimate, rather than the date the order was entered. Keep canceled lines and already fulfilled quantities out of remaining commitments. If an order can slip between weeks, record that risk in the assumptions instead of silently moving it after the period closes.
Suppose RawOrders contains product in B, requested ship date in C, status in D, and remaining units in E. This export is already restricted to the selected location. The selected product is in M1 of your summary. A2 contains the week's starting date. This Excel formula totals confirmed units requested during that week:
=SUMIFS(RawOrders!$E$2:$E$1000,RawOrders!$B$2:$B$1000,$M$1,RawOrders!$D$2:$D$1000,"Confirmed",RawOrders!$C$2:$C$1000,">="&A2,RawOrders!$C$2:$C$1000,"<"&A2+7)
Microsoft's SUMIFS documentation explains summing a range against multiple criteria and requires matching range dimensions. Extend every range together if the export is longer. Covirage is not affiliated with Microsoft.
The formula selects rows; it does not establish their quality. Tie the result back to the frozen order report and investigate excluded statuses. The demand forecasting data guide covers the broader field-selection question.
A US equipment distributor freezes its estimate before the first week below. Columns C through E contain additional demand beyond the confirmed orders in B. They exclude those orders and change only the assumed uncommitted volume.
Start this summary in A1:
| Week starting | Confirmed units | Additional: low | Additional: base | Additional: high |
|---|---|---|---|---|
| October 5, 2026 | 300 | 50 | 100 | 150 |
| October 12, 2026 | 200 | 100 | 200 | 300 |
| October 19, 2026 | 100 | 150 | 300 | 450 |
| Total | 600 | 300 | 600 | 900 |
The additional quantities are planning assumptions, not measured sales. Record their owner and evidence in adjacent notes: quoted opportunities, expected walk-in demand, or a comparable territory. Do not add a quoted opportunity again if it becomes one of the confirmed orders already included.
Put low, base, and high total demand in F, G, and H. In row 2:
=B2+C2
=B2+D2
=B2+E2
Fill through row 4. The totals reconcile to 600 confirmed units plus 300, 600, or 900 additional units. The scenario range is therefore 900 to 1,500 units, with a base estimate of 1,200. None of those figures is a promise or a statistical interval.
Add actual completed sales in I only after each week is complete. Preserve the original scenario columns and put revisions in a separately dated version.
| Week starting | Low units | Base units | High units | Actual units | Base minus actual | Absolute base error |
|---|---|---|---|---|---|---|
| October 5, 2026 | 350 | 400 | 450 | 360 | 40 | 40 |
| October 12, 2026 | 300 | 400 | 500 | 420 | -20 | 20 |
| October 19, 2026 | 250 | 400 | 550 | 350 | 50 | 50 |
| Total | 900 | 1,200 | 1,500 | 1,130 | 70 | 110 |
This display summarizes F through K of the worksheet. In J2 and K2:
=G2-I2
=ABS(J2)
The base forecast is 70 units high overall, but its absolute weekly misses total 110. Check that SUM(J2:J4) equals SUM(G2:G4)-SUM(I2:I4). Weighted absolute percentage error is:
=SUM(K2:K4)/SUM(I2:I4)
That is 110 / 1,130 = 9.7%, rounded to one decimal place. The low and high scenarios have absolute errors of 230 and 370 units respectively. These observations favor the base assumptions for this window; they do not prove that the range will contain future sales.
Use the comparison of sales forecasting methods to distinguish assumptions from other forecasting approaches. Keep this first validation focused on the evidence you actually possess.
Bring your dated scenarios, initial order export, and actual-sales file to Covirage. Covirage maps the columns; its tools compute and check comparisons from the supplied quantities, and the external AI model explains the results with every figure traced to its rows. Create a free Covirage account, then set up your business and bring a sample of your own data.
See forecast analysis in Covirage for the supporting planning context.
You can construct an assumption-based estimate using orders, customer research, comparable products, and business knowledge. Its uncertainty remains real. Record the evidence behind each input and define which observations will cause the team to revise the next estimate.
Initial orders show committed quantities under their stated terms, not every purchase that might occur. Separate them from uncommitted opportunities and incremental demand assumptions. Check cancellations, requested delivery dates, and duplicate lines before using the order book as a starting point.
A range built from management assumptions is a set of scenarios. It is not a statistical confidence or prediction interval unless a separate, justified statistical procedure produced it. Label what changes between scenarios and avoid assigning probabilities without supporting evidence.