Data Enhancement
Status: On request. Writes to the CRM.
Operating model, before anything else
These are configured operations that run when you set them up and run them. There is no cron entry and no continuous service behind them. A pass is scoped, agreed and executed against a defined set of records; it does not watch the CRM and it is not running on any portal between passes. The audit is what makes a pass worth running: it reports the duplicate pairs and the inconsistent values, and the pass acts on the set you approve.
Enhancement is not enrichment
Enrichment adds information that was not in the CRM, from an external provider. Enhancement works on data already in the CRM: standardising it, reconciling it, deduplicating it, or deriving a value from it. No external provider is involved in any operation on this page.
Normalization
Standardises categorical values, country and region among them, to a target vocabulary defined before the pass runs. The mapping is the client's rule, not an inferred one: a value not covered by the mapping is reported, not guessed at. The pass is idempotent, so running it twice over the same records changes nothing the second time.
Deduplication and merge
Candidate pairs come from the audit's two-pass duplicate detection: an exact pass over normalised values and a corroborated near-match pass. Matching for contacts considers identity fields rather than name and company alone, which was a real blind spot and is now covered. Merges are executed through the platform merge API against an approved set, and every merge is recorded with both record identifiers.
Derived fields, including ACV
Some values the business needs are not stored anywhere directly. Annual Contract Value is the standing example: the contract exists across line items, billing periods, dates and terms, while reporting needs one figure.
The calculation recomputes ACV from the line items actually attached to a deal, including recurring billing terms, and handles the calendar-versus-fiscal year boundary explicitly rather than assuming they coincide. Where the stored figure has drifted from what the line items support, the deal is corrected. It is arithmetic over dates and amounts: nothing is inferred, and once the organisation has defined how ACV is calculated there is no reason to reinterpret that definition per deal.
Every write is logged and reversible
A pass records what it changed, on which record and from what prior value, before the write. Scope is agreed in advance, so a pass cannot widen itself mid-run.
Determinism
No part of this calls a language model. Same inputs, same calculation, every time.