Sales Forecast Submission Template
A roll-up that applies a rep's bias only where their own history can support it, plus the note that carries the number upward.
Free download · No account needed
Forecast Submission Note
[Team] — [Period] Submission
1. The number and the band
| Submitted | — |
|---|---|
| Raw sum of rep commits | — |
| Adjustment | — percent of the raw sum |
| Standard uncertainty | — |
| Band | — to — |
2. Where the adjustment came from
| Common term, applied once | — grand mean of [n] cells |
|---|---|
| Rep corrections | — [n] of [n] cleared the test |
3. What is not in the number
The reps left at a zero correction, and the dollars their spread puts into the band instead.
4. Why the band is that width
Straight sum of the rep dispersions, the quadrature sum, and the correlated part that does not shrink.
5. Movement since the last submission
Net, then the two reps who account for most of it, named, with amounts.
6. What would make this wrong
Three register rows, each with an amount, an owner, a date, and whether it already sits inside the band.
Search for a sales forecast submission template and you get revenue projection spreadsheets built for founders: months across the top, products down the side, a growth rate in one cell. None of them is the artifact a sales manager actually files, which is a roll-up of rep commits with an argument attached. The advice that does address rep bias stops at ranking reps by signed error and multiplying the ones who look consistent, which is also where the arithmetic quietly breaks.
A rep's mean error over six quarters is an estimate built from six observations, and its standard error is their quarter-to-quarter spread over the square root of six. When the mean is smaller than that, there is nothing to correct. NIST's measurement handbook has run this test on instruments for decades: a consistent bias is corrected only where it can be reliably estimated. A bias that changes direction gets a zero correction, with its magnitude carried as uncertainty rather than thrown away.
The pack carries a worked eight-rep book. Culverton Systems, forty-four quarters, six reps whose mean error is at least five points from zero, three that clear the test. One term is +12.76 points and another is -13.01, and only the second earns a correction, because the first carries 19.03 points of spread against the second's 6.24. Apply all eight measured means instead of the three and the rep-level total moves $1,507, since optimists and sandbaggers cancel.
What's in the pack
How a Correction Earns Its Place
The four numbers per rep, the three-way split of the error matrix, and the whole Culverton book worked to a submitted figure.
Bias Test sheet
Rep term, spread, standard error and the t value against its critical point, both halves, and a verdict.
Roll-Up With Bias Adjustment
Raw commit beside every adjusted figure, the common term applied once, and each rep's contribution to the band in dollars.
Submission History sheet
One row per rep per period, with the common term and the period effect broken out before any rep term is computed.
Risk Register
Every risk typed rep-specific or common cause, because only the first kind shrinks when eight rep books are added.
Movement Since Last Submission
New business, value changes, slips and losses per rep, reconciled to the dollar against the prior commit.
Submission Guidance
What a rep sends, what happens to it, what they get back within a day, and the four dates the cycle runs on.
Commentary Format
The six sections of the note that goes upward, in the order the person defending the number needs them.
Submission Note
One filled in end to end for Culverton's Q7, including the adjustments considered and deliberately not made.
How to use it
- 1
Open in River, or download it
Open the pack in River and let the agent build the matrix, or download the blank Word and CSV files free.
- 2
Load every past submission and its actual
One row per rep per period, commit as submitted against what closed, with the sign on the error kept.
- 3
Split the matrix, then test each rep term
Common term at the top, named period effects removed, and every rep term measured against its own standard error.
- 4
Price the failures into the band, then write the note
Uncorrected spread combines in quadrature into a range, and the six section note carries both upward.
Frequently asked questions
Is this template free?
Yes. Four Word documents and five CSV sheets, downloaded as a zip, no signup and no card. River is the optional half: it builds the error matrix, runs the test and drafts the note. Single-shot jobs live in River's tool index, and the rest of the packs are in the template library.
What format are the downloaded files?
Documents come as .docx and sheets as .csv, so nothing needs converting. The documents open in Word, Pages and Google Docs; the sheets open in Excel, Numbers and Google Sheets. The CSV files carry a byte order mark and Windows line endings so Excel reads them without a prompt.
What does Edit with AI actually do?
It installs this exact pack as a private Space and primes the agent with your history. It reads the submissions and actuals, splits each error three ways, computes every rep term with its standard error, applies only the ones that clear it, and drafts the note with the band already worked out.
Is a bias adjustment not just a haircut on the rep's number?
No, and the sheet is built to make that visible. Every rep's raw commit stays beside the adjusted figure, the adjustment lands on the roll-up rather than on the rep, and most reps get zero. On the worked book three of eight are corrected and the largest term is common to everybody.
Our CRM already reports forecast accuracy by rep.
It reports accuracy without a sign, which cannot be corrected because it has no direction: twenty points high one quarter and twenty low the next averages to zero error and reports as twenty. Nothing in that report tests whether a mean is larger than the uncertainty in measuring it.
What happens to a rep with only two quarters of history?
Nothing, and that is the arithmetic rather than a policy about new hires. Two quarters is one degree of freedom, where the two-sided critical value of t is 12.706, so no realistic term clears it. At four quarters the threshold falls to 3.182 and the sheet recomputes their verdict automatically.
Does this replace our forecast process?
No. This prepares one number on one date and defends it. Which stage the deals die at, how the pipeline gets cleaned, and what happens inside a deal review are separate work with separate artifacts. So is what the comp plan behind these commits actually pays a rep to chase, and the pack says so rather than quietly widening its own scope.
Submit a number with the arithmetic attached
Download the blank pack as Word and CSV files, or open this exact pack in River and have your own submission history split, tested and rolled up.
Edit with AI