CRM Rescue · Local-first internal tool
CRM Rescue takes a messy CRM export and walks an operator through mapping, validation, and duplicate review — producing clean data plus an audit trail, without ever contacting a lead or touching the source CRM.
The problem
Every operations team eventually inherits a CRM export full of duplicates, missing owners, stale deals, and half-filled fields. The usual options are bad: clean it by hand for days, or run a tool that "fixes" things silently so nobody can say why two contacts became one.
Silent merges and opaque "lead quality" scores don't make data trustworthy — they just move the doubt somewhere you can't see it.
Who it's for
Sales operations and small-business teams using Salesforce, HubSpot, or spreadsheet-based tracking who need to repair an export before a migration, a campaign, or a report — and need to show their work afterward.
The build
A local-first web application that guides one operator through the whole cleanup: import a CSV, map its columns to a standard schema, run rule-based validation, review duplicate candidates with evidence side by side, and export clean data plus a complete change history.
Salesforce and HubSpot are supported as mapping presets. The deterministic core is complete without AI — there is no model, no scoring black box, and no network dependency.
The workflow
A six-step operator journey. The three screenshots below show the main screens used at steps 1, 3–4, and 5–6.
Pick the export format, load the file, and map each source column to the standard schema. A preview shows sample values so mapping is a decision, not a guess.
Plain, predictable rules check emails, phones, dates, owners, and stages. Invalid rows stay visible with rule codes — nothing is silently dropped or corrected.
The tool generates candidate pairs with the matching evidence spelled out: which fields matched, and how confident the rule is. Scores create review items — they never merge anything.
Each pair needs an explicit confirmation, and each decision lands in the audit trail. In the demo, a person resolved all six candidates: two merges, four kept separate.
Attention flags highlight records with missing owners, stale activity, or no next step — described as operational next actions, not a mystery score.
Cleaned CSV, change audit, and a quality report — reproducible, so rerunning the same inputs gives the same outputs. A local run can be deleted entirely when you're done.
Worth a closer look
The real screens
These three views show the operator's main touchpoints: import and mapping, duplicate decisions, and final review with exports and an audit trail.
Technical verification
Release evidence from the local verification run — synthetic fixtures, not customer data.
The repeatable demo processes 24 invented rows across Salesforce-style, HubSpot-style, and generic fixtures. A Chromium critical-flow test passed, dependency audits (npm and pinned Python) reported no known vulnerabilities, and two independent demo generations matched SHA-256 exactly. A backend test also sends a contact-data canary through the pipeline and asserts it never appears in application logs.