A Weighted Forecast From Your Pipeline CSV
Turn a pipeline export into a forecast you can defend: commit vs. upside, best/likely/worst scenarios, and an honest gap-to-quota — instead of a number you sandbagged or hoped for.
A forecast isn't a number you feel good about. It's a number you can defend deal by deal.
Gut forecasts are either sandbagged or hopeful, and neither survives a forecast call. Store your stage weights and what 'commit' means, and a pipeline export becomes a forecast with the reasoning attached — best, likely, worst, and the gap you actually have to close.
The number you can't break down is the number nobody believes — including you, the week it's due and you're short.
A forecast you can't explain deal by deal is a guess wearing a decimal point.
Weighting is a model, not a crystal ball. The recipe makes your assumptions explicit so you can argue with them — it doesn't predict the future. Not financial advice.
| Complexity | Easy |
| Tools needed | Claude Pro or above, Claude Desktop → Cowork mode, on macOS or Windows, Claude Projects |
| Time to build | ~25 min first time · ~5 min per forecast after that |
| Best for | Founder or sales lead who owns the number — B2B services and software |
What this solves
Forecast numbers that are sandbagged or hopeful and fall apart the moment a manager asks how you got there.
The problem
Forecast day comes and you produce a number. You either sandbagged it so you can beat it, or you talked yourself into the optimistic version because the quarter’s tight. On the call, your manager asks how you got there, and you can’t break it down past “I feel good about these three.” The number isn’t wrong because you’re dishonest — it’s unreliable because it was never built from the deals up, with stated assumptions about what each stage is really worth. Without that, the forecast is a vibe, and everyone treats it like one.
The fix is building the forecast from the pipeline with explicit weights, so every number traces to a deal and an assumption you can defend.
Ingredients
- Claude Subscriptions — Cowork and Projects aren't available on the free plan
- Platforms & Modes — Cowork runs in the desktop app only — not on web or mobile. This recipe runs the math on your pipeline and builds scenarios — multi-step work that runs in Cowork
- Claude Projects — Your stage weights, your definition of commit vs. upside, and your quota are constant. A Project stores them so every forecast uses the same model
How it works
You forecast every period, and your weights, commit definition, and quota are constant. A Project stores them so every forecast uses the same model and you’re not reinventing the method under pressure.
- Open Claude Desktop and click Cowork in the mode selector across the top (Chat · Cowork · Code).
- In the left panel, find Projects and click the + button.
- Choose Start from scratch. Name the project “Forecast” and let Claude set up its folder.
- You’ll know it worked when the project appears in the left panel with its own folder and an instructions field.
Your project has a folder on your computer — that’s where the model file lives permanently. The pipeline export joins it each time you forecast.
- Ask Claude, right in the project: “Create a file in this project’s folder:
forecast-model.md. Leave it empty — I’ll fill it in.” (Or create it yourself in any text editor and save it into the project folder.) - Fill it in using the description below.
- Each forecast, drop a fresh CRM export into the same folder as
pipeline.csv(or connect the CRM and pull live — see Extra Credit). - Confirm Claude can see them: ask “List the files you can see in this project.” Both filenames should come back. If they don’t, see If It Doesn’t Work.
forecast-model.md
Your stage probabilities, what qualifies as commit vs. upside, and your quota — the assumptions the whole forecast rests on, written down so they can be examined.
Example: “Stage weights: discovery 10%, demo 30%, proposal 60%, negotiation 85%. Commit = negotiation stage with a confirmed next step and verbal. Upside = proposal with momentum. Quota this quarter: 250k. Discount anything with a pushed close date by one stage.”
pipeline.csv
Your open deals — stage, amount, close date, last activity. Export it, or connect the CRM and pull live.
Open your Project in Claude Cowork. Give Claude the specifics for this run, ask for the main output, then follow up for any additional pieces you want. The exact wording for each prompt — with what it’s asking for and why — is in What you actually type below.
Check three things before you commit to the number:
- The math is shown and traceable. A forecast number with no breakdown is one you can’t defend on the call. Confirm you can see which deals make up commit and how the weights produced the likely figure. If it’s a black-box total, ask for the deal-by-deal build.
- Commit is genuinely commit. The most common self-deception is calling hopeful deals commit. Check that the commit number only includes deals meeting your real criteria from the model. If a shaky deal snuck into commit, the whole forecast is soft.
- The scenarios are honest, not three versions of optimism. Worst case should reflect real downside — deals slipping, not just a slightly smaller best case. If worst and likely are nearly identical, the range isn’t telling you anything.
Before you submit: does the number match your read of the deals? You know which “negotiation” deals are actually stuck. The model is only as good as your honesty about each deal — and this is a planning tool, not financial advice. Override the weighting where you have real knowledge the data doesn’t show.
What you actually type
Name your files explicitly the first few runs, and ask Claude to show its work on anything that matters.
“Here's my open pipeline [attach CSV or pull from CRM]. Using `forecast-model.md`: give me a weighted forecast with commit and upside separated, best/likely/worst scenarios, total coverage against quota, and the gap I still have to close. Show the math — which deals make up commit, and how the weighting produced the likely number.”
“Which three deals would most change the likely number if they moved a stage? That's where I should focus.”
A forecast with the math attached: commit and upside separated, best/likely/worst scenarios, coverage against quota, and the gap left to close. Every number traces to named deals — you can see which deals make up commit and how the weights produced the likely figure. If it's a single total with no deal-by-deal build, or worst case is just a slightly smaller best case, it missed.
- The forecast is only as good as the stage and probability data in the CSV. Wrong stages, stale close dates, and missing amounts flow straight into the number — it doesn't fix dirty pipeline data. Scrub first.
- Stage weights are your assumptions, not measured truth. Until you've compared a few forecasts to actuals, treat the weights as a starting model that needs tuning.
- It's a planning tool, not a prediction — and not financial advice. The model can't see the deal that's quietly stuck in 'negotiation'; your honest read of each deal still corrects the number.
If it doesn’t work
- No Cowork tab in Claude Desktop — update the app to the latest version and confirm you’re on a paid plan; Cowork isn’t on the free tier. On Windows, Cowork also needs the Virtual Machine Platform feature enabled — if the tab still won’t appear, that’s the fix.
- Claude can’t see
forecast-model.mdorpipeline.csv— the files aren’t in the project’s folder, or the fresh export landed somewhere else. Open the project, check which folder it points to, and move the files there. Then re-run “list the files you can see.” - The forecast is a black-box total — a number with no build is exactly what you had before. The prompt’s “show the math” line does the work; if the breakdown is missing, ask: “List the deals that make up commit, and show how the weights produced the likely number.”
- Worst case is just a smaller best case — the scenarios are shading optimism, not modeling downside. Ask: “For worst case, assume every deal with a pushed close date slips out of the quarter — what’s the number then?” A range that can’t hurt you isn’t telling you anything.
Extra credit
Small additions that pay back the next time you run it.
- CRM connector — connect your CRM so the forecast pulls live and reflects this morning’s reality, not last week’s export. See the Connectors guide.
- Scrub first — run the Pipeline Review recipe before forecasting so hygiene issues don’t distort the math.
- Track accuracy — keep each forecast and compare it to actuals; over time you’ll learn whether your stage weights are right or need adjusting.
“The forecast that holds up is the one where every number traces back to a specific deal and a stated assumption.”
What this teaches you about Claude Cowork
The recipe is one application. The principles apply to everything you’d hand to Claude.
Explicit assumptions beat gut feel. A forecast’s credibility comes from being built deal-by-deal on stated weights. Storing the model forces those assumptions into the open, where they can be checked and argued — which is what makes the number defensible instead of a vibe.
Showing the math is the safeguard. For a number this consequential, traceability matters more than a clean total. Insisting Claude show which deals built commit lets you catch the shaky deal that snuck in before your manager does.
The model serves your judgment. Weighting produces a starting number; your honest read of each deal corrects it. The recipe is a planning aid — not a prediction — and its value is making your assumptions visible enough to improve, period over period.
Who this is for
Founder or sales lead who owns the number in B2B services and software (5–100 employees).
The pain: A forecast number that can't be defended deal by deal on the call
The outcome: A commit number that breaks down deal by deal, with the gap to quota known early enough to act
Published June 26, 2026 · 0 views