CRM reconciliation and backfill behind trustworthy reporting

A founder-led B2B company running paid inbound into a newly built CRM, with a lead engine, scoring model, phone system and attribution all going live within a few months of each other. Leadership wanted a funnel report it could act on.

History had arrived through several doors. An agency-owned automation had silently dropped a share of leads. A form change had turned a text field into numeric ranges, invalidating every score computed before it. Time zones were missing on most records. Campaign attribution was blank on a large share of early deals. Calls logged to the contact instead of the deal. Several annual-revenue values had been keyed a thousand times too low.

Any report built on that data would have been wrong in ways nobody could see, and the first time leadership caught one wrong number, every other number would lose its credibility.

Hygiene is not a cleanup task at the end; it is the precondition for reporting. Each fault had a mechanical fix, and each fix could be verified by query rather than by eye.

A reconciliation of the agency's records against the CRM that recovered the lost leads with their original dates. A time-zone engine that back-filled hundreds of leads from area code and state. Campaign, ad set and ad name restored on every converted deal. A flag that separates records scored on the old form from the new, and a regression run that re-scored the history and confirmed zero disagreements with the live model. A before-save flow that binds phone-system calls to the account's open deal, plus a phone-number match for calls that had already been orphaned, which cut unattached calls from roughly three in ten to about one in twenty. Deal names standardized, legacy notes migrated onto the right records, and amount values restored after a field-permission fault had blanked them.

Every fix was logged with the count of records touched and the query that proved it. The remaining hygiene the system cannot do for the owner, such as logging approvals and loss reasons and keeping the open-deal count honest, was written up as the owner's standing list rather than left implicit.

When the funnel review ran, its numbers survived scrutiny. One draft metric did not: a connect rate that looked flattering because it counted thirty-second calls as connections. We checked it against the reps' notes, found the true figure far lower, and shipped the report without it. That is the standard the reconciliation made possible.

With the history trustworthy, the next build was reporting the owner could run weekly: speed to lead, conversion by rep, pipeline movement, and a full lead change log.