Pick 4–6 lightweight signals and turn them into a 0–100 score
Start small: pick signals you can calculate today from your CRM, email tool or spreadsheet. Aim for 4–6 signals that are easy to compute and explain to the team — completeness, recent activity, provenance, duplicate risk and consent are usually sufficient.
- Completeness (30): required fields filled — e.g. email, company, phone. Give full points only when essential fields are present.
- Recent activity (20): activity in last 12 months (opens, logged calls, invoices). Older records lose points.
- Provenance (20): source reliability (manual verified > form capture > scraped list). Use a simple ranked scale from known sources to unknown.
- Duplicate risk (15): E.g. matching email variations, similar names, or multiple company records — higher risk lowers the score.
- Consent (15): explicit marketing/business consent where needed (critical for email sends).
Total the weights to 100 and produce a single trust_score (0–100). Define three thresholds you can act on: 70+ = automation‑safe, 40–69 = review / human queue, <40 = quarantine. Keep the weighting and thresholds documented so people in Fareham or elsewhere in Hampshire understand decisions.
Backfill historical records safely and test in a corner
Don’t overwrite live fields or retrigger workflows while backfilling. Create a new staging property/column called trust_score_staging and populate that first. Work in small batches (500–2,000 rows) and run a short manual review for the first few batches.
Examples by platform:
- HubSpot / Salesforce: add a new custom property/field for the score and import or update in small batches. Avoid firing downstream automations by either pausing workflows temporarily or importing into the staging field and using a controlled workflow to copy values once you’ve validated samples.
- Marketo / Pardot (email tools): calculate consent and engagement portions outside the ESP when possible, then sync a single score field so mailings can reference the score without complex segmentation logic.
- Spreadsheet‑backed integration: compute the score in the sheet, keep a timestamped column and use Zapier/Make to push only changed rows to the CRM. Use a dry‑run column first (e.g. new_trust_score) so you can preview changes.
Have a rollback plan: keep the original score column untouched until step‑by‑step validation is complete, and keep a CSV of pre‑update values so you can restore if needed. Run the whole backfill during quieter hours and validate a random 1% sample before committing the next batch.
Gate automations, monitor the results and assign owners
Apply the score as a gating condition in your usual automations. Use conservative thresholds at first and expand once you see low error rates.
- Lead assignment: only assign automatically when trust_score >= 70; route 40–69 into an "uncertain lead" queue for a quick human check; block <40 from assignment and mark for enrichment.
- Email sends / campaigns: include trust_score >= 70 in your suppression logic for marketing sends (use explicit consent as a blocker, regardless of score). For transactional billing actions use a higher bar and an additional owner confirmation step when score <80.
Monitor four simple things daily: number of records processed by threshold, automation failure or bounce rate by trust band, volume in the review queue, and top provenance sources causing low scores. Alert owners by creating a short list or dashboard and a daily summary email to the steward (rotate the steward weekly). This keeps a small South Coast team informed without noise.
If you want a practical hand to set up the score, gate rules and monitoring for HubSpot, Salesforce or spreadsheet integrations, see our CRM and marketing data optimisation page for Fareham: [crm-marketing-data-optimisation-fareham.html].
If you'd rather not run this yourself, Optira can help set the pattern and an incremental rollout so your automations stop acting on risky records.