How to build a sales forecast from CRM deal history, and what the data depth tells you

How do you build a sales forecast from CRM data?

Build it from stage-to-stage conversion and win rates measured on closed deal history, applied to today's open pipeline and adjusted for deal age and size. The method works only if the CRM holds enough closed deals, consistently staged, across enough quarters. That requirement is also a clue to how deep a client's sales records are, which a fractional CFO can note quietly during the work.

This guide walks through the forecast, the data checks that decide whether it can be trusted, and a permission-first way to raise a licensing introduction with the owner afterwards. It assumes a Salesforce, HubSpot or similar CRM and read access granted by the client's admin.

What data does a pipeline-based forecast need?

If stage history is not retained, you can compute win rate on closed deals but not conversion between stages. Say so in the forecast notes.

Steps to build the forecast

1. Define the population. Pick one segment, such as new-business deals above a minimum size, so rates are not blended across different motions.
2. Pull closed deals. Use at least eight quarters if volume is modest; more history is better when deal counts are low. Exclude duplicates and test records.
3. Calculate stage conversion. For each stage, divide deals that moved forward by deals that entered it.
4. Calculate cycle length. Measure median days from creation to close by segment.
5. Apply to open pipeline. Multiply each open deal's amount by the conversion from its current stage to close, then place expected close in the quarter implied by age and cycle length.
6. Back-test. Run the same method on a past quarter you already know and compare.
7. Set ranges, not one number. Present low, expected and high cases and the assumptions behind them.

How long does the forecast stay valid?

Refresh conversion rates every quarter and re-run the back-test whenever pricing, sales team structure or target segment changes. A forecast built before a CRM migration, a new product launch or a change in how stages are defined should be rebuilt from the point of change, not patched. Keep a short assumptions log with the date of each refresh so a successor can reproduce the numbers.

What are the common mistakes?

What does the work tell you about the client's records?

By the time the forecast runs, you know how many years of closed deals the CRM holds, whether stage history survives, whether notes, email and call logs are linked to deals, and whether any earlier CRM was retired. Years of linked deal histories with outcomes are the kind of structured record of real work AI buyers value. For why sales data matters, see why sales and CRM data is valuable for AI.

Keep your notes to the system list and year ranges. You do not need to read, export or describe any record for this purpose.

• The company had 50+ full-time employees at peak (contractors excluded).
• The CRM holds several years of deals with outcomes, ideally with stage history.
• Records span multiple systems, such as CRM, email, call tools, finance and support.
• The company owns the records and has no customer-contract barrier to licensing.
• An authorized sponsor (owner, CEO, CFO or authorized representative) exists.
• Previous CRMs were archived rather than deleted.

Run the same checklist through the company fit checker for a non-binding preliminary screen, and read who qualifies for the baseline.

How do you raise the introduction with the owner?

Ask first. Do it after the forecast is delivered and trusted.

If the answer is yes, share your referral link or submit basic company details through the referral form. The company works directly with SourceX on the inventory and rights review. Partners never export, upload or describe confidential records. Before the owner proceeds, it helps to know who can sign a data license. If the CRM is about to change vendors, preserve a full export first; the guide to switching to an AI-native ERP covers what happens to legacy history.

Illustrative scenario

Illustrative, fictional: a 90-person IT services company wants a quarterly forecast. The fractional CFO finds six years of closed deals in one CRM, with stage history only for the last three. She builds the forecast on three years and notes that the earlier data supports win rate but not stage conversion. The company also retired a second CRM in a merger and kept a full archive. She records both facts, finishes the forecast, and later asks the owner for permission to introduce.

How do rewards work, and what should a fractional CFO check?

Partners earn 25% of the eligible platform fees SourceX actually collects from the referred company's licensing deals, capped at $100,000 cumulative per referred company. The reward is payable only after the buyer pays and SourceX receives its fee, and no reward is guaranteed. It is never deducted from what the company receives. Before you register, read your engagement letters and the professional rules that apply to you on referral fees and disclosure, and read the program terms. The fractional CFO overview explains the role further.

When to skip it

• The CRM was implemented less than two years ago and holds few closed deals.
• Deals are logged inconsistently, with no outcomes recorded.
• Most customer information is consumer personal data without a clear licensing basis.
• The company has already licensed the same data for AI training.

Next step

Add three lines to your forecast workpaper: years of deal history, whether stage history exists, and whether earlier CRMs were archived. If the picture is strong and the owner agrees, register as a partner and introduce the company. Related reading: customer profitability analysis.