Skip to main content
MethodSalesSales operationsIntermediate

Build a sales forecast

A monthly or quarterly sales forecast built three ways (deal-by-deal categories, stage-weighted pipeline and historical run rate), reconciled into one number with a range, the deals that decide it, and the assumptions behind it.

Use this task 5 quick questions, then open it in ChatGPT, Claude or Gemini.

This is a practical, copy-paste process for you need a number for next month or quarter for the owner, board or bank. Use it when you want a clear structure, then add your own details.

Verified on 2026-10-282 sources cited

Quick start

1. Clean your pipeline first: mark dead deals as lost (do not delete them, because your win rates come from closed deals) and fix close dates that have passed.
2. Sort deals into three groups: very likely (commit), possible (best case) and early (pipeline).
3. Use your own history: of the deals that reached each stage last year, how many did you win?
4. Multiply each deal by that win rate and add them up. Then compare with what you usually sell in a period.
5. If the numbers disagree a lot, find out why instead of averaging them.
6. Give a range: low, likely and high, and name the few deals that decide it.
7. At the end of the period, compare the forecast with what really happened.

Before you start

  • You need to clean up or inspect the pipeline itself first (stale deals, missing steps). Run a sales pipeline review
  • You need a cash forecast (when money arrives and leaves), not a bookings forecast. Build a 13-week cash-flow forecast
  • You have no deal pipeline (a shop, website, walk-in or appointment business, or subscriptions that renew automatically): forecast units and demand instead. If only part of your sales goes through deals (for example a few wholesale accounts), use this method for that part and the demand method for the rest, then add the two. Forecast product demand
  • Other CRMs: the rules apply to any CRM, but category names and the stage-to-category mapping differ by platform. Map your CRM’s categories to commit, best case and pipeline before you start.

Full method

How to use this method

  1. 1. Copy or downloadCopy the prompt below, or download the method file and attach it to your chat.
  2. 2. Paste into any AIChatGPT, Claude, Gemini or Copilot. Replace each [PLACEHOLDER] with your details, or let the AI ask you.
  3. 3. Add your materialAttach your own files, data or brief so the AI works from your facts, not guesses.

The prompt

Paste it as-is. The AI will ask for anything still in [BRACKETS].

Download .md
# Build a sales forecast
Method by Deepstory Research: deepstoryresearch.com/method/sales-forecast

You are a revenue-operations analyst who builds sales forecasts for small and growing sales teams. Follow the method below step by step. Goal: A monthly or quarterly sales forecast built three ways (deal-by-deal categories, stage-weighted pipeline and historical run rate), reconciled into one number with a range, the deals that decide it, and the assumptions behind it.

## My inputs
- Pipeline data: [PIPELINE_DATA]  (example: CRM export of open deals with stage, amount and close date)
- Historical results: [HISTORICAL_RESULTS]  (example: win rates by stage and last 4 quarters of bookings)
- Period: [PERIOD]  (example: Q1 2027)
- Sales cycle and deal size: [SALES_CYCLE_AND_DEAL_SIZE]  (example: 45 days, INR 4 lakh average)
- Target: [TARGET]  (example: INR 1.2 crore new bookings)

Before you start: if any input above is still in [BRACKETS] or missing, ask me for it in one short numbered list and wait. Do not guess. If I attached files (data, briefs, Deepstory evidence files), treat them as the primary source.

## Steps to follow
1. Clean the pipeline: mark dead deals lost (never delete), fix past close dates and missing amounts.
2. Put each deal in a forecast category (commit, best case, pipeline) using written rules.
3. Calculate the stage-weighted pipeline from your own historical win rates.
4. Calculate a run-rate view from past periods and the time left in this one.
5. Compare the three views, explain the gaps and agree one number with a low and high range.
6. List the deals that decide the outcome, with the next step and owner for each.
7. State the assumptions and the date the forecast will be checked.

## Deliver the output in this structure
- Pipeline clean-up list
- Deal categories (commit, best case, pipeline)
- Stage-weighted view
- Run-rate view
- Forecast with low, likely and high range
- Deals that decide the period
- Assumptions and review date

## Quality rules
- Use your own historical win rates, not generic percentages; label any default as an ASSUMPTION.
- Do not count a deal twice across categories or periods.
- A deal with a close date in the past is not in the forecast until it is updated.
- Label every number as FACT (with source and date), ESTIMATE (show the calculation) or ASSUMPTION.
- If a Deepstory evidence file is attached, use its figures as the reference values and cite its source references; tag claims that are not from it as [UNVERIFIED].
- End with: the 3 assumptions that matter most, what would change the answer, and the single best next step.

What you need to give the AI

InputPlaceholderExample
Pipeline data[PIPELINE_DATA]CRM export of open deals with stage, amount and close date
Historical results[HISTORICAL_RESULTS]win rates by stage and last 4 quarters of bookings
Period[PERIOD]Q1 2027
Sales cycle and deal size[SALES_CYCLE_AND_DEAL_SIZE]45 days, INR 4 lakh average
Target[TARGET]INR 1.2 crore new bookings

Use it when

  • You need a number for next month or quarter for the owner, board or bank.
  • The pipeline looks big but the forecast is always missed.
  • Setting up forecast categories in a CRM such as HubSpot.

Best with: Any chat AI that can analyse a spreadsheet; attach the open deals export and last year’s closed deals, with customer names removed if needed.

What you get

  • Pipeline clean-up list
  • Deal categories (commit, best case, pipeline)
  • Stage-weighted view
  • Run-rate view
  • Forecast with low, likely and high range
  • Deals that decide the period
  • Assumptions and review date

The AI will ask you first

Answer these before the AI advises you; each one changes what it recommends.

  1. 1

    Do you track sales as deals in a CRM or spreadsheet, with a stage and amount for each?

    The category and weighted views need a deal pipeline.

    If unknown: No pipeline: use the run-rate view only and follow the routing in Before you start.

  2. 2

    What is being forecast: new bookings, revenue, or number of deals, and for which period?

    Bookings (the value of signed deals, not appointments or orders), invoiced revenue, recognised revenue and cash fall in different periods; annual subscriptions invoiced upfront are booked on signing, recognised monthly and paid after the payment term. Ask who will read the forecast: a bank usually wants revenue or cash.

    If unknown: Forecast new bookings (signed deals) for the period asked.

  3. 3

    Can you export open deals with stage, amount, close date, owner, created date and last activity?

    Stage, amount and dates drive every view; age and activity show stale deals.

    If unknown: No deal-level view: forecast from the run rate or the funnel, and list the export fields needed.

  4. 4

    What share of deals reaching each stage were eventually won in the last 12 months?

    Stage-weighted forecasts must use your own conversion, not generic percentages.

    If unknown: Calculate from closed deals if exported; otherwise use early (first meeting or qualified demo) 10%, middle (proposal or trial) 25%, late (negotiation or verbal yes) 50%, labelled ASSUMPTION (rule of thumb), and show low and high at half and double these rates.

  5. 5

    What were bookings in each of the last 4 to 6 periods, and how much is already closed this period?

    The run-rate view checks the pipeline against history.

    If unknown: Annual sales / 12 a month, flat, labelled ASSUMPTION, with known peaks named.

  6. 6

    What is the typical sales cycle from first meeting to signed deal?

    Deals created too late to close within the cycle are unlikely this period, and a period longer than the cycle will also close deals not yet in the CRM.

    If unknown: The midpoint of the range the user gives, or 30 days if none, labelled ASSUMPTION; do not confuse the sales cycle with payment terms.

  7. 7

    What is the target for the period, and what are your financial year and quarter dates?

    The forecast is judged against the target, and quarter names differ between April to March and calendar years.

    If unknown: No target; name the period by its months (for example January to March 2027), not by quarter number.

  8. 8

    Which CRM and plan do you use, and should new customers, expansions and renewals be forecast separately?

    Forecast tools need specific plans and settings, and renewals and expansions close at very different rates from new deals.

    If unknown: Work from the export; forecast new bookings only and list other deal types separately.

  9. 9

    How do reps decide that a deal is "commit"?

    Categories mean nothing without written rules.

    If unknown: Use: commit = buyer confirmed budget, decision maker and date, and the next step is booked.

The procedure, step by step

Read this before you run it, so you can check the AI actually followed the method.

  1. 1

    Clean

    Mark dead deals Closed lost with a reason (never delete them), update past close dates, fill missing amounts, merge duplicates only after checking with the owner.

  2. 2

    Categorise

    Commit, best case, pipeline using the written rules; note any rep overrides.

  3. 3

    Weighted view

    Sum of amount x your own stage win rate for deals closing in the period.

  4. 4

    Run-rate view

    Closed so far + (average bookings per week in past periods x weeks left), adjusted by (same period last year / average period last year); if last year is unknown, say "no seasonal adjustment" and name the known peaks or dips in the period.

  5. 5

    Big deals

    List the deals that make up most of the gap to target, each in or out with a reason.

  6. 6

    Reconcile

    Likely = closed + weighted (normal deals) + big deals called in + allowance for deals not yet in the CRM (scaled to the share of the period left); compare with run rate; if more than about 15% apart, explain and choose; low = closed + commit, high = closed + commit + best case.

  7. 7

    Write assumptions

    Win rates used, period of history, deals excluded and why.

  8. 8

    Review and learn

    Weekly check against the forecast; at period end record accuracy.

Platform settings(show)
PlatformPathSettingNote
HubSpotSettings > Objects > Forecast > PipelinesForecast categories per deal stage (Not forecasted, Pipeline, Best case, Commit, Closed won); Automate forecast categories
HubSpotSettings > Objects > ForecastForecast deal amount (Weighted amount by default, or Total amount); forecast period (monthly or quarterly)Changing the forecast period resets revenue goals and forecast submissions.
HubSpotSettings > Data Management > Objects > Deals > PipelinesDeal probability per stage; Won and Lost stagesWeighted amount = amount x deal probability.

Worked example

The figures below are illustrative. Use your own numbers and facts.

Scenario: An Indian B2B SaaS company forecasting new bookings for January to March 2027 (Q4 of its April to March financial year), on 5 January 2027 (week 1 of 13 complete).

Input: Target INR 1.2 crore. Open deals: 38, total INR 2.1 crore; 32 of them (INR 1.8 crore) have close dates from January to March 2027, and 6 (INR 0.3 crore) have close dates in December 2026 and no activity for 80 days. Own win rates from the last 12 months: demo 18%, proposal 35%, negotiation 62%. Median sales cycle 45 days. Last 4 quarters, oldest first (January to March 2026 to October to December 2026): INR 1.05, 0.95, 0.9 and 1.1 crore, so the same quarter last year was INR 1.05 crore. Deals created and won inside the same quarter averaged INR 0.25 crore a quarter. Already closed this quarter: INR 0.12 crore. One deal of INR 0.4 crore is in negotiation; its legal review is not booked.

Output excerpt: Clean-up: the 6 deals with December close dates and no activity for 80 days (more than a third of 45) are stale: moved out of the forecast and the owners asked to update or mark them lost, leaving 32 deals worth INR 1.8 crore, all closing this quarter. Categories: commit INR 0.45 crore (4 negotiation deals with budget, decision maker and signing date); best case INR 0.75 crore (proposal deals INR 0.35 crore plus the INR 0.4 crore deal); pipeline INR 0.6 crore (demo). Weighted view, big deal treated separately: 0.6 x 18% + 0.35 x 35% + 0.45 x 62% = INR 0.51 crore; plus closed INR 0.12 crore = INR 0.63 crore. Big deal: out until the legal review is booked. Deals not yet in the CRM: 0.25 x 12 / 13 weeks left = INR 0.23 crore. Run rate: average quarter 1.0, so 0.12 closed + (1.0 / 13 weeks x 12 weeks left) x seasonal factor (1.05 / 1.0) = INR 1.09 crore. Likely: 0.63 + 0.23 = INR 0.86 crore, 21% below the run rate; the gap is explained because the run rate assumes a typical large deal closes and this quarter’s only one is out, so likely stays at 0.86. Forecast: low INR 0.57 crore (closed + commit), likely INR 0.86 crore, high INR 1.32 crore (closed + commit + best case). The 1.2 crore target is reachable only if the INR 0.4 crore deal signs (0.86 + 0.4 = 1.26). Review every Monday; record actual / forecast at quarter end.

Check your result

Before you trust the AI's output, confirm each of these.

  • ☐ Stale and dead deals were taken out of the forecast (marked lost or updated, never deleted), with a list.
  • ☐ Win rates are your own, or labelled ASSUMPTION.
  • ☐ All three views are shown with their numbers.
  • ☐ The forecast has low, likely and high values, or says why low and high cannot be computed.
  • ☐ Large deals are forecast individually.
  • ☐ Assumptions and the review date are written down.
What you'll get(show template)
## Sales forecast: [PERIOD]

Target: [ ] | Closed so far: [ ] | Open deals in period: [n], [total]

### Clean-up
Deal | Issue | Action | Amount removed

### Categories
Commit: [ ] | Best case: [ ] | Pipeline: [ ]

### Views
Categories low (closed + commit): [ ] | Categories high (+ best case): [ ] | Weighted (own win rates, big deals separate): [ ] | Run rate: [ ] | Deals not yet in the CRM (history or funnel, and which): [ ] | Likely and how it was chosen: [ ]
If a section does not apply, write "Not applicable: [reason]".

### Deals that decide the period
Deal | Amount | Stage | In or out | Next step | Owner | Date

### Forecast
Low: [ ] | Likely: [ ] | High: [ ] | Gap to target: [ ]

### Assumptions and review
[win rates, history used, exclusions] | Review: [dates]

Quality checks built into the prompt

  • • Use your own historical win rates, not generic percentages; label any default as an ASSUMPTION.
  • • Do not count a deal twice across categories or periods.
  • • A deal with a close date in the past is not in the forecast until it is updated.
  • • Every number labelled FACT, ESTIMATE or ASSUMPTION; claims outside attached evidence tagged [UNVERIFIED].

Decision rules

If/then rules the AI should apply. Heuristic rules are a practitioner's rule of thumb, not a platform rule.

IfThenType
You set up forecast categories in HubSpot.HubSpot groups deal stages into Not forecasted, Pipeline (low likelihood), Best case (moderate likelihood), Commit (high likelihood, committed to the forecast) and Closed won. Categories can update automatically when a deal changes stage, and users can override a deal’s category on the deal record. Category meaning follows the order of the values in the Forecast category property, not their names.Platform rule
You use HubSpot’s forecast tool.It needs Sales Hub or Service Hub Professional or Enterprise, and Super Admin to set up. Its Forecast deal amount defaults to Weighted amount (amount x deal probability); the other options include Total amount. Check which one is set before reading category totals, and export the plain Amount field when you apply your own win rates. Changing the forecast period resets revenue goals and forecast submissions.Platform rule
You use a weighted pipeline in HubSpot.Weighted amount = deal amount x deal probability for the stage. Set each stage’s probability in the pipeline settings (the default Sales Pipeline comes with generic probabilities), and include both Won and Lost stages so reports process deals correctly.Platform rule
You set stage probabilities.Replace default probabilities with your own: for each stage, won deals that passed through it divided by all closed deals that passed through it, over the last 12 months.Heuristic
A deal’s close date is in the past, its age is more than about 1.5 times the median sales cycle of won deals, or it has had no logged activity for more than about a third of the cycle (30 days at most).Treat it as stale: move it out of commit and best case until the owner updates the date and books the next step.Threshold: 1.5 x cycle age; one-third of cycle inactivity (rules of thumb)Heuristic
A deal is marked commit.Require written evidence: confirmed budget, decision maker involved, agreed close date, and a booked next step. Without these, it is best case.Heuristic
You set the likely number.Likely = closed + weighted view of normal deals + big deals you call in + the allowance for deals not yet in the CRM. Compare likely with the run-rate view. If they differ by more than about 15% of the run-rate view, do not average; explain the gap (a big deal called out, a slow start, stale deals) and keep the view whose assumptions you can defend. Once you have 3 or more periods of forecast accuracy, prefer the view that was closest.Threshold: 15% (rule of thumb)Heuristic
The period has not started, the period is longer than your median sales cycle, or a deal was created recently.Add one allowance, shown as its own line, for deals not yet in the CRM: the average value of deals created and won inside the same period over the last 4 periods, multiplied by the share of the period still to come (weeks left / weeks in the period). If that history is unknown, use the funnel instead: expected new qualified opportunities in the rest of the period x your own win rate from that stage x average deal amount (without a known amount, forecast the number of deals). Use one of the two, never both. Keep deals created later than (period end minus median cycle) out of commit unless the buyer has confirmed a signing date.Threshold: median cycle; share of period left (rules of thumb)Heuristic
A few deals make up most of the forecast.Forecast those deals one by one (in or out) and apply weighting only to the rest; a probability on a single large deal is misleading.Heuristic
You report the forecast.Give a likely number with a low (closed plus commit) and high (plus best case) range, and track forecast accuracy each period (actual / forecast) to improve the rules. If there is no commit category (no pipeline, or the period has not started), set low and high from the same calculation with the lowest and highest win rate or monthly sales of the last 12 months; if those are unknown, write "Low and high: not computable" and never use a fixed percentage band.Heuristic

Failure modes

Common mistakes with this method, how to spot them, and how to fix them.

MistakeSymptomFix
Weighting with default stage percentages.Forecast consistently too high.Use your own win rates by stage.
Leaving stale deals in the pipeline.Big pipeline, missed forecast.Take deals with past dates or long inactivity out of the forecast: mark them lost or ask the owner to update them; never delete them.
Commit based on optimism.Commit deals slip every month.Written commit rules with evidence.
Applying a probability to one huge deal.Forecast swings on a single decision.Forecast large deals in or out individually.
Averaging views that disagree.A number nobody can explain.Explain the gap and choose.
Counting the same deal in two periods or categories.Totals add up to more than the pipeline.One deal, one period, one category.
Applying your own win rates to an already weighted amount.Weighted view far below every other view.Export the plain Amount field, or set Forecast deal amount to Total amount, before weighting.
Using "ever won" stage rates for a single period.Volume is right but deals land a period late.Measure what share of deals in each stage at the start of past periods closed won within that period, or discount by your usual slip rate.
Deleting dead deals instead of marking them lost.Win rates look higher than they are.Mark them Closed lost with a reason.
Never measuring accuracy.The same mistakes every quarter.Record actual / forecast each period.

Agent mode: if an AI agent runs this unattended

May do without asking

  • • Analyse exported deal data and calculate the three views.
  • • Draft the forecast summary and the list of deals that decide the period.
  • • Read CRM pipeline and forecast settings in a read-only session.

Must stop before (the user decides and does these)

  • • Editing deal stages, amounts, close dates, owners or forecast categories in the CRM.
  • • Deleting, merging or marking deals Closed lost.
  • • Changing pipeline stages, stage probabilities or the Forecast deal amount setting.
  • • Changing the forecast period (in HubSpot this resets all forecastable revenue goals and forecast submissions).
  • • Turning Automate forecast categories on or off, or creating or editing workflows.
  • • Creating, editing or deleting forecast goals, or submitting or deleting a forecast in the CRM forecast tool.
  • • Sending the forecast to investors, lenders or the board.
  • • Exporting customer data outside the business.
  • • Buying, upgrading or starting a trial of a CRM plan, add-on or seat, or accepting any terms.
  • • Assigning seats or changing user permissions.
  • • Importing, bulk-updating or uploading deal data, or connecting a new app or integration.
  • • Editing the Forecast category property (adding, renaming or reordering its values).

Evidence for this method

This method runs on your own material: your files, brief or code. If your task needs market or economic numbers, attach a sourced context file from the evidence library alongside the prompt.

Browse evidence

Run next

Methods that chain well after this one.

Deepstory methods are process guidance for AI tools. Review AI output before relying on it; legal, tax, medical and investment decisions need a qualified professional.