Systems and records
CRM opportunity reversal dataset: preserving why a deal moved backward
By SourceX Editorial · Reviewed by Noah Loul ·
9 sections
Short answer
Preserve why a sales deal moved backward by writing one record per reversal: the opportunity and account IDs, the field that moved and its before and after values from CRM history, the date, who made the change, a reason code with a short note, and links to the evidence that explains it, such as an amended order, a lapsed approval or a stalled security review. Export the history before retention limits erase it.
System export
CRM opportunity reversal
Key takeaways
- Five movements count as a deal moving backward: stage regression, forecast downgrade, close date slip past the period, amount reduction and reopen.
- CRM history shows that a field changed, by whom and when, but never why, so the reason and evidence must be recorded separately.
- Export stage, forecast, date and amount history on a schedule, because tracking limits and retention windows erase it.
- Signed agreements, approvals and delivery records outrank CRM fields when they disagree; flag the conflict instead of overwriting.
- Every reason needs a source and an evidence link, and an attribution limit when more than one reason applied.
- A reversal record states a reason and its evidence; it does not prove cause.
What counts as a deal moving backward
A deal moves backward whenever a change lowers the seller's confidence or pushes value out of the period. The obvious case is a stage regression, such as an opportunity going from negotiation back to evaluation. Four other movements matter just as much: a forecast category downgrade, a close date pushed past the period end, an amount reduction and a closed deal reopened.
Each of these is visible in a CRM's history if history was being kept, and none of them carries its reason. Salesforce stores stage history for opportunities in the OpportunityHistory object and changes to tracked opportunity fields in OpportunityFieldHistory. HubSpot keeps date entered and date exited properties for every deal stage, plus a cumulative time in stage that adds up repeat visits when a deal re-enters a stage. Dynamics 365 Sales lets a user reopen a closed opportunity. All three show that something moved. A CRM opportunity reversal dataset adds what they leave out: why.
Not every backward-looking change belongs in the set. A stage corrected within minutes because a rep clicked the wrong value, a close date nudged by a day inside the same month, or a bulk update run by an administrator during a pipeline cleanup are noise. Write those exclusion rules down, apply them in code before anyone codes reasons, and keep a count of what was excluded so the filtering can be checked later.
| Reversal type | How it shows in the CRM | Usual trigger to confirm | Evidence to attach |
|---|---|---|---|
| Stage regression | Stage history entry to an earlier stage; HubSpot date entered for an earlier stage after a later one | New stakeholder, failed evaluation, scope rework | Meeting notes, revised proposal, evaluation feedback |
| Forecast downgrade | Forecast category change in field history or audit log, for example committed to best case | Lost commitment, budget hold, approval not obtained | Customer email, budget notice, internal deal review note |
| Close date slip past period | Close date change crossing a quarter or month boundary | Procurement or legal delay, security review, signer unavailable | Procurement ticket, redline status, security questionnaire log |
| Amount reduction | Amount change in field history | Scope cut, fewer seats, product removed | Revised quote or order form version |
| Reopen after close | Closed won or lost record set back to open | Signature voided, order rejected, lost deal revived | E-signature log, order rejection, new inbound request |
Why the history has to be exported before anyone needs it
Field history in most CRMs is limited by what was tracked and how long it is kept, so the evidence of a reversal can expire before anyone studies it. In Salesforce, standard field history tracking covers up to 20 standard and custom fields per object. Without Field Audit Trail, Salesforce keeps field history for up to 18 months, and up to 24 months through the API, after which older entries are removed.
Dynamics 365 runs on Dataverse auditing, which answers who changed a record, which fields changed and what the previous value was. Its retention period is a setting, and Microsoft notes that changing the retention period does not change it for audit records that already exist. HubSpot's stage properties store dates per stage, which shows that a deal re-entered a stage but not necessarily every intermediate value of every other field.
The practical rule is to confirm that stage, forecast category, close date and amount are tracked, then export the history tables on a fixed schedule into storage the company controls. A reversal dataset built later can only reach back as far as the oldest export.
A monthly export is usually enough for history tables, as long as each export is kept rather than overwritten and carries its extraction date. Keep the related activities, emails and quote versions on the same schedule, because a reason note that points to an email the company no longer holds becomes an unverifiable claim. Note any period when tracking was switched off or a field was not yet tracked, so gaps are explained rather than read as quiet quarters.
The stage reversal rationale and evidence schema
One record describes one backward movement on one opportunity. The fields below are a minimum. The movement fields come straight from CRM history, so they are facts; the reason fields are judgments, so they carry an author, a date and a source; the evidence fields are links rather than copies, so the record stays small and the originals keep their own access controls.
At a glance
The stage reversal rationale and evidence schema
Use stable pseudonymous IDs for accounts, contacts and people from the first export, and keep the mapping key inside the company. That keeps records joinable after customer names and contact details are removed.
| Field | Allowed values | What it records | Validation rule |
|---|---|---|---|
| reversal_id, opportunity_id, account_id | Stable pseudonymous IDs | Which deal and which customer entity moved | All required; opportunity must exist in the export |
| reversal_type | stage_regression, forecast_downgrade, close_date_slip, amount_reduction, reopen | Which kind of backward movement | Required; must match the field that changed |
| field_name, old_value, new_value | Values exactly as stored in CRM history | The movement itself | Must match a row in the history export |
| changed_at, changed_by_role | Timestamp from history; role ID of the user | When and by whom the change was recorded | changed_at required; role resolves to the role table |
| reason_code | Closed list agreed by revenue operations, plus unknown | The primary stated reason | Required; unknown is allowed but must be explicit |
| reason_note and reason_source | Short note; rep_entry, manager_review, deal_desk, customer_statement | The reason in words and who gave it | If reason_code is not unknown, both are required |
| evidence_links | IDs of emails, activities, quotes, orders, approvals, tickets | Records that support or contradict the reason | At least one link unless reason_code is unknown |
| contract_ref | Agreement and amendment IDs in force | Which signed terms governed the deal at that moment | Required for reopen and amount_reduction on closed deals |
| outcome_after | recovered_same_period, recovered_later, closed_lost, still_open, closed_won_reduced | What happened next | Filled only after the deal closes or the window ends |
| attribution_limit | single_reason, multiple_reasons, reason_disputed | How confident the record is about the cause | reason_disputed requires a second reason_note |
Choosing reason codes that hold up over years
Reason codes are only useful if the same reason gets the same code across reps, teams and years. Many teams require a reason when a deal closes lost, but few require one when a deal moves backward while still open, so the reversal list usually has to be designed from scratch. Start from the reasons that already appear in notes and deal reviews, not from an ideal taxonomy.
Keep the list short, mutually exclusive and tied to something a reviewer can check. A code such as customer_budget_hold can be confirmed against a customer email; a code such as lost_momentum cannot be confirmed against anything and tends to absorb every hard case. When the list changes, version it and record the date, because records coded under different lists cannot be compared directly.
Separate the customer side from the seller side. A reversal caused by the customer's procurement process and one caused by the seller's own approval delay look the same in the stage history but teach opposite lessons. One field for which side caused the movement, set before the reason code, makes that distinction explicit.
- customer_budget_hold: the customer stated in writing that funds are frozen or moved.
- customer_process_delay: procurement, legal, security or signature steps on the customer side ran late.
- scope_reduced: the customer asked for fewer products, seats or services than quoted.
- seller_approval_pending: an internal discount, terms or credit approval was not obtained in time.
- competitor_reengaged: the customer reopened evaluation of another vendor, confirmed in notes or email.
- unknown: no evidence of a reason exists; allowed, but counted and reviewed.
Which evidence outranks what the CRM says
When the CRM and a signed document disagree, the signed document wins. A stage field is a person's entry at a moment in time; an executed order form, an amendment or an approval record is the thing the stage was supposed to describe. The record should cite the stronger evidence and flag the conflict rather than silently correct the CRM value.
Forecast categories are a good example. Microsoft's guidance for Dynamics 365 Sales describes committed as a verbal or contractual commitment from the customer and tells users not to set won or lost by hand, because closing through the close dialog updates the category automatically. A deal that sat in committed with no customer commitment anywhere in the evidence was never committed, and its later downgrade is a correction, not a reversal. Tag it that way.
- Executed agreement, order form or amendment: controls what was actually bought, in what scope and from when.
- Internal approval records, such as discount or non-standard terms approvals: control whether the seller was allowed to offer what the CRM shows.
- Customer statements in writing: control the customer's stated position at a date.
- Delivery and usage records, such as provisioning, invoices issued or seats activated: control what happened after the paper was signed.
- CRM fields and rep notes: the starting point, valid unless one of the records above contradicts them.
Worked example: a renewal that slipped from committed to pipeline
Illustrative: a fictional software company tracks a renewal with an expansion in its CRM. Over six weeks the deal moves backward twice and then closes smaller than forecast. Every value below is invented, and amounts are shown only as a percentage of the original forecast.
The resulting records: reversal R1 is a forecast_downgrade with reason_code customer_process_delay, reason_source deal_desk, evidence linking the questionnaire ticket, attribution_limit single_reason. Reversal R2 is an amount_reduction with reason_code scope_reduced, reason_source customer_statement, evidence linking the revised quote and the customer email, contract_ref the existing agreement, and attribution_limit multiple_reasons because the customer email also mentions a budget review. Both carry outcome_after closed_won_reduced once the order is signed.
| Week | CRM change | What the evidence shows | Record created |
|---|---|---|---|
| 1 | Forecast category set to committed at 100 percent of forecast | Customer email confirms intent to renew and expand | None; forward movement |
| 3 | Forecast category committed to pipeline; close date pushed one month | Customer security team opens a new questionnaire for the added product | R1 forecast_downgrade, linked to the questionnaire ticket |
| 5 | Amount reduced to 70 percent of forecast | Revised quote removes the added product; customer email cites budget review | R2 amount_reduction, multiple_reasons |
| 6 | Closed won | Executed order form matches the revised quote | outcome_after filled on R1 and R2 |
A failing record and the checks that reject it
A record fails when it asserts more than the evidence supports or cannot be traced back to history. Take R2 recorded badly: reversal_type stage_regression, although only the amount changed; reason_code price with no reason_note and no evidence links; contract_ref empty; outcome_after filled in the week the change happened. Each of those is a separate check, and all of them can run automatically against the CRM export.
At a glance
A failing record and the checks that reject it
Run the checks on every record before the set is shared internally or reviewed for any outside use. Record the count of failures by check, because a high failure rate on one check usually points at a process gap, such as reasons never being required, rather than at individual reps.
Keep failing records in a quarantine table instead of deleting them. Many can be repaired from the evidence once the check names the problem, and the ones that cannot are a measure of how much of the archive can support decision-level work at all.
- History match: field_name, old_value, new_value and changed_at match a row in the history export exactly.
- Type match: reversal_type is consistent with field_name, for example amount_reduction only when the amount field changed.
- Reason support: any reason_code other than unknown has a reason_note, a reason_source and at least one evidence link.
- Contract check: reopen and post-close amount_reduction records cite the agreement or amendment in force.
- Outcome timing: outcome_after is empty until the deal is closed or the observation window has ended.
- Conflict flag: when evidence contradicts the CRM value, the record keeps the CRM value and adds a conflict note instead of overwriting it.
What a reversal record cannot claim
A reversal record states a reason someone gave and the evidence around it; it does not prove cause. Reps tend to choose reasons that reflect well on them, managers may rewrite reasons in a deal review, and many deals move backward for two or three reasons at once. The attribution_limit field exists so that a record can say so.
Keep three limits explicit. Do not infer a reason from the timing alone, such as assuming every slip near quarter end was budget related. Do not merge reasons written at different times into one, because a reason recorded the day of the change and one recorded in a post-mortem are different evidence. And do not compare reason frequencies across periods where the reason list changed without noting the change date, since a new picklist value can look like a new trend.
Customer data, contracts and the decision to share reversal records
Reversal records carry contact names, customer statements and commercial terms, so preparation has to remove personal data and confidential customer details before any outside use. Customer agreements may also restrict how information shared during a sale can be used. Access to the CRM, privacy permission for contact data, ownership of the records and a contractual right to license them are separate questions, and removing names answers none of them alone. Counsel reviews the specific agreements and laws that apply.
If a revenue team decides to go further, SourceX manages the rights review, contracting, preparation and delivery for companies that license selected records to AI developers. The team's company signs off on scope and terms before anything leaves its hands, and the process opens with a short company assessment rather than a data upload.
Frequently asked questions
Is a pushed close date really a reversal?
It is when the new date crosses the period the deal was forecast in, because value leaves that period. A close date moved by a few days inside the same period is usually housekeeping and can be left out, as long as the rule is written down and applied consistently.
How far back can Salesforce show opportunity changes?
Without Field Audit Trail, Salesforce keeps field history for up to 18 months, and up to 24 months through the API, and only for fields that were being tracked. Stage history is kept in its own object. Exporting history on a schedule is the only way to keep older changes.
Should reps be forced to enter a reason every time a deal moves back?
Requiring a reason code on backward stage and forecast changes is the single most useful change, as long as unknown is an allowed value. Forcing a long note tends to produce copied text. A short code plus a link to the email or activity that explains it is usually better evidence.
Can AI tools use reversal records without customer names?
Often yes, because the useful signal is the sequence of changes, the stated reason and the evidence type, not the identity of the customer. Stable pseudonymous IDs keep records joinable after names are removed. Whether the records can be used outside the company is a separate rights question.
What if the stated reason contradicts the evidence?
Keep both. Record the stated reason with its source, add the contradicting evidence link and mark the record reason_disputed. The disagreement is itself informative, and rewriting the CRM value would destroy the original record.
Sources
- Salesforce's OpportunityHistory object represents the stage history of an opportunity. Source
- Salesforce's OpportunityFieldHistory object represents the history of changes to the values in the fields of an opportunity. Source
- Salesforce field history tracking covers up to 20 fields per object and, without Field Audit Trail, keeps history up to 18 months, or 24 months through the API. Source
- HubSpot stage calculated properties record the date a deal entered and exited each stage and a cumulative time in stage for deals that re-enter it. Source
- Dynamics 365 Sales records details entered when closing an opportunity and lets users reopen a closed opportunity. Source
- Dynamics 365 Sales describes the committed forecast category as a verbal or contractual commitment and says won or lost should not be set manually. Source
- Dataverse auditing shows who changed a record, which fields changed and the previous value, and a retention change does not apply to existing audit records. Source
See if your company qualifies
A short company assessment. No data uploads are needed.