Systems and records
Revenue operations case completeness: audit a sales history for AI
By SourceX Editorial · Reviewed by Noah Loul ·
9 sections
Short answer
A sales history is a complete AI training case when every stage the CRM records is backed by dated evidence, the action taken, the authority behind it, and an outcome that can be checked outside the CRM, such as a signed contract, an amendment, a delivery or a renewal. Audit each opportunity stage by stage, and treat any stage the CRM asserts but nothing else confirms as incomplete.
Value drivers
Revenue operations case
Key takeaways
- Case completeness asks whether each stage is backed by evidence, action, authority and an outside outcome; field completeness does not.
- Stage and field history age out in some CRMs, so export them before rebuilding cases.
- Contradictions between the CRM and signed documents are the most valuable findings, and amendments outside the CRM are the commonest.
- Score each deal as complete, repairable or partial; shared entity identifiers decide how many can be repaired.
- Keep sellers' explanations, such as loss reasons and forecast categories, separate from contracted and delivered facts.
What makes a sales history a complete AI training case?
A sales history is a complete training case when every stage the CRM records is backed by four things: dated evidence that the stage happened, the action someone took, the authority behind that action, and an outcome that can be checked outside the CRM. Remove any of the four and the record still describes a deal, but it no longer shows how the decision was made or whether it was right.
Revenue operations case completeness is the share of opportunities that pass this test stage by stage. It is a different measure from field completeness. A deal can have every required CRM field filled and still be incomplete as a case, because the close date was typed in at quarter end, the amount differs from the signed order form, and the renewal that proved the customer stayed lives in a billing system nobody linked.
The audit below turns that test into a worksheet a RevOps team can run on its own opportunities before a sales history is used for forecasting models, agent training or licensing.
Which CRM records show how a deal actually moved?
Stage history, field history and closure records show movement; the opportunity record itself shows only where the deal ended. Each major CRM stores movement differently, and each has limits that decide how much history you can still recover.
Salesforce keeps stage changes in the OpportunityHistory object and tracked field changes in OpportunityFieldHistory. Salesforce's help documentation says that without Field Audit Trail, field history is retained for up to 18 months in the org and up to 24 months through the API, so older changes may already be gone. HubSpot creates date entered and date exited properties for each pipeline stage, plus latest and cumulative time in stage, and populates them retroactively from existing data. Dynamics 365 Sales records an Opportunity Close activity when a deal is closed as won or lost, holding the actual revenue, competitor, close date and description, and lets users reopen a closed opportunity.
Reopened deals deserve their own flag. A deal closed as lost in one quarter and reopened in the next is two decisions, and a history that keeps only the final state hides the first one.
The stage, evidence, action and outcome completeness audit
Run the audit per opportunity, one row per stage the deal passed through. Each row asks the same four questions; the last column says what makes the row incomplete. Stage names vary by CRM and company, so map your own stages onto these rows before starting and keep the mapping with the results.
| Stage | Evidence that confirms it | Action and authority to record | Outcome check | Incomplete when |
|---|---|---|---|---|
| Qualified | Dated first meeting, inbound request or qualification notes | Who qualified the deal and against which criteria | Deal progressed, or was disqualified with a reason | Stage date equals creation date for most deals, a sign of bulk entry |
| Discovery | Meeting records, call notes or a needs document | Requirements captured and by whom | Requirements appear in the later proposal | No activity exists between qualification and proposal |
| Proposal or quote | Quote versions with dates and status | Who issued each version; discount approvals | Final quote accepted, or closed as lost, canceled or revised | Several active quotes with no record of which one the customer accepted |
| Negotiation | Redlines, approval requests, revised terms | Approver role for any non-standard term or discount | Approved terms match the signed document | Discount or term changed with no approval record |
| Commit | Forecast category change with a date | Who set the category and on what basis | Deal closed in the committed period, or the slip is recorded | Category was set to won or lost by hand instead of through closure |
| Closed won | Signed contract or order form, with signature date | Signer authority on both sides | Amount, term and products match the CRM | Close date or amount differs from the signed document |
| Closed lost | Loss record with reason and competitor | Who recorded the loss and when | No later reopen, or the reopen is linked | Reason is a default picklist value with no supporting activity |
| Delivery or onboarding | Order, provisioning, shipment or kickoff record | Team that accepted the handoff | Delivery completed as contracted | No record links the won deal to delivery |
| Renewal, expansion or churn | Renewal order, amendment or cancellation | Who approved the change | Customer behavior over the following term | Amendments exist in billing or contract systems but not in the CRM |
Where do contracts, approvals and billing disagree with the CRM?
The most valuable part of the audit is the list of contradictions, because each one shows where the CRM's apparent answer was wrong. Look for these first; they are cheap to find with a join between the CRM export and the contract or billing system.
Amendments are the commonest blind spot. Under the FASB revenue standard, a contract modification is a change in scope or price approved by the parties, and approval can be in writing, by oral agreement or implied by customary business practice. A change like that can alter what was actually sold without any CRM opportunity being touched. This is general accounting context, not advice; your finance team or auditor decides how a specific change is treated.
- Close date earlier than the signature date, often a quarter-end entry for a deal signed days later.
- CRM amount different from the signed order form, or products on the opportunity that are missing from the contract.
- A won deal whose accepted quote was later closed as revised, with the revised version never linked.
- Forecast category set to won or lost by hand, which Microsoft warns can drop the deal from the forecast or distort rollups.
- Discounts above the approval threshold with no approval record in the CRM or the approval tool.
- Renewals, seat changes or cancellations recorded only in billing.
- Account names in the CRM that map to a different legal entity on the contract.
Complete, repairable or partial: scoring each opportunity
Score each opportunity once the stage rows are filled. Complete means every stage has evidence, action and outcome, and no contradiction is open. Repairable means the missing evidence exists in another system, such as a CPQ tool, e-signature archive or billing platform, and can be joined by a shared account, contract or order identifier. Partial means at least one stage rests only on the CRM's own assertion.
Entity identifiers decide how many deals are repairable. Record the CRM account ID, the legal entity on the contract, any parent account and the billing customer ID side by side. Where those do not resolve to one customer, the deal stays partial, however rich its activity history.
Score lost deals with the same rules. A lost deal with a dated competitor mention, a final quote and a closure note is a complete case of a loss, and is often more instructive than a won deal with no paper trail.
Attribution limits: what even a complete case cannot show
A complete case shows what happened and who decided; it does not prove why the customer bought. Loss reasons and win reasons are the seller's interpretation, recorded after the fact, and forecast categories are a judgment of confidence. Microsoft's own description of the commit category, for example, is a customer that has made a verbal or contractual commitment, which is a statement about the seller's belief.
Label these fields as opinion rather than outcome. A model trained to predict a loss reason learns how sellers explain losses, which is useful, but it is not the same as learning why customers leave. Keep the contracted and delivered facts in separate fields from the explanations.
Illustrative audit: a software renewal traced stage by stage
This example is illustrative, not a real deal. A mid-market software renewal shows closed won in the CRM. The audit finds the evidence for most stages, one contradiction on dates, and an expansion that never reached the CRM.
| Stage | What the CRM says | Evidence found | Result |
|---|---|---|---|
| Proposal | One quote, accepted | Two quote versions; the first closed as revised after 9 days | Repairable: link both versions |
| Negotiation | No note | Approval email from the sales director for a non-standard payment term | Repairable: attach approval with approver role |
| Commit | Forecast category committed for 6 weeks | Category change history present | Complete |
| Closed won | Closed on the last day of the quarter | Signature dated 3 days later | Contradiction: record both dates, use the signature date as the outcome |
| Onboarding | Not recorded | Kickoff completed 21 days after signature in the project tool | Repairable through the shared account ID |
| Renewal term | No later activity | Amendment adding 15% more seats in month 7, recorded only in billing | Partial until the amendment is linked |
Deals to leave out, and those that need a contract or privacy reviewer
Leave a deal out when its stages cannot be reconstructed even with other systems, when the account cannot be resolved to one customer, or when the deal is still open or inside its first renewal term, so the outcome is not yet known.
Send a deal to a contract reviewer when the agreement restricts disclosure of its terms, pricing or the customer's identity, which many enterprise agreements do. Send it to a privacy reviewer when the evidence includes emails, call recordings or notes about named people at the customer. The audit can record that evidence exists and when, without copying its content into the case set until those reviews are done.
How complete cases change a sales history's licensing value
Companies licensing CRM and sales histories to AI developers can do so through SourceX, which handles rights review, preparation, contracting and delivery. Sales histories are judged against a guide of two or more years of consistent records in one system and tens of thousands of records, so a history where most deals score complete or repairable is a stronger starting point than a larger one that is mostly partial.
A RevOps lead can check whether the company qualifies through a short assessment, without uploading any opportunity records.
Frequently asked questions
What should the stage evidence action and outcome completeness audit include?
One row per stage per opportunity, recording the dated evidence that the stage happened, the action and the authority behind it, an outcome checked outside the CRM, and the condition that makes the row incomplete. Each deal then gets a score of complete, repairable or partial, plus a list of contradictions.
Can stage history be recovered if it was never exported?
Sometimes. HubSpot populates its stage date properties retroactively from existing data, while Salesforce retains field history for a limited period without Field Audit Trail. Export stage and field history now, and rebuild older stages from quotes, contracts and activity dates where the CRM history has aged out.
Are lost deals worth auditing?
Yes. A lost deal with a dated final quote, competitor information and a closure note is a complete case of a loss. Lost deals show where a sales process fails, which a set of only won deals cannot.
Do the signed contracts need to be part of the dataset?
Not necessarily. The audit needs to confirm that the contract exists and that its date, term and scope match the CRM. Whether contract text itself is shared is a separate decision for a contract reviewer, since many agreements restrict disclosure.
How many years of sales history make a useful case set?
There is no fixed minimum, but buyers generally look for consistent history over multiple years in one system. Two years of complete cases is usually more useful than five years in which stage history has aged out.
Sources
- Salesforce's OpportunityHistory object represents the stage history of an opportunity; OpportunityFieldHistory holds changes to tracked opportunity fields. Source
- Without Field Audit Trail, Salesforce retains field history data for up to 18 months, and up to 24 months through the API. Source
- HubSpot provides date entered, date exited, latest time and cumulative time properties for each pipeline stage, populated retroactively from existing data. Source
- Closing a Dynamics 365 opportunity as won or lost records actual revenue, competitor, close date and description in an Opportunity Close activity, and closed opportunities can be reopened. Source
- Dynamics 365 describes the committed forecast category as a verbal or contractual commitment and warns that manually selecting won or lost can remove a deal from the forecast or produce inaccurate rollups. Source
- In Dynamics 365 Sales a quote that is not accepted can be closed as lost, canceled or revised, optionally creating a revised quote. Source
- Dynamics 365 opportunities, quotes, sales orders and invoices share data such as products, prices, discounts and tax, and an opportunity can be converted into each. Source
- Under ASC 606, a contract modification is a change in scope or price approved by the parties, and approval may be written, oral or implied by customary business practice. Source
See if your company qualifies
A short company assessment. No data uploads are needed.