Project 06Data & CRM

CRM audit

Twelve thousand contacts, audited record by record, because a database nobody trusts is a database nobody uses.

Role
Group Marketing Executive
For
Archetype Group
When
2025 to 2026
  • 12,000contacts audited
  • 20country databases
  • 1shared schema

01

At a glance

The whole project in three steps.

  1. The problem

    Twenty country lists full of duplicates and dead contacts, and a sales team that had stopped trusting the numbers.

  2. What I did

    Defined one schema and reconciled 12,000 contacts, with human review wherever a match was ambiguous.

  3. The result

    One database the team can defend, reporting figures that reconcile, and rules that keep it clean.

Skills it shows

  • Data governance
  • CRM management
  • Data quality
  • Process design
  • Marketing operations

Tools

  • HubSpot CRM
  • Salesforce
  • MS Excel
  • Zapier
  • Looker Studio

02

The challenge

Why the project had to happen.

Every office kept its own list. The same client appeared three times under three spellings, people who had left their companies years ago were still receiving mail, and nobody could answer "how many clients do we actually have" without a caveat.

Read the full context

The duplicates were the smaller cost. The sales and marketing teams had stopped believing the numbers, and a team that does not trust its own data starts making decisions from memory instead.

03

How I did it

4 steps. Open any one for the detail.

  1. Establish what a record should be

    Defined one schema for twenty country databases before cleaning anything.

    How

    Before cleaning, define clean. One schema across twenty country databases: what fields are mandatory, what a company record means versus a contact record, how markets are tagged.

  2. Deduplicate, then verify

    Matched duplicates automatically, then reviewed the ambiguous ones by hand.

    How

    Automated matching finds most duplicates and is confidently wrong about the rest. Twelve thousand records were reconciled with human review where the match was ambiguous, because a merged record cannot be un-merged from a guess.

  3. Retire the dead weight

    Removed contacts with no valid route, so the numbers became true.

    How

    Contacts with no valid route to them are not assets, they are a deliverability liability. Removing them improves the numbers by making them true, which is uncomfortable and correct.

  4. Make the clean state maintainable

    Set entry rules and ownership per office so the clean state holds.

    How

    A one-off cleanup decays within a year. Entry rules and ownership per office are what stop it silting up again. The deliverable was the process as much as the database.

04

See it

12,000 contacts — one dot per twenty

As found

Duplicates
merged on review
Unreachable
retired
Remaining
one schema, 20 markets

05

What it achieved

  • Twelve thousand contacts audited and reconciled onto one schema across twenty country-level databases.

  • Marketing sends against a list it can defend, and the reporting figures reconcile with each other for the first time.

  • The maintenance rules mean the state holds without another full audit.

06

What I took from it

  1. A merged record cannot be un-merged from a guess. Automated matching finds most duplicates and is confidently wrong about the rest.

  2. Removing dead contacts improves the numbers by making them true. That is uncomfortable and correct.

  3. A one-off cleanup decays within a year. Entry rules and ownership per office are what keep a database clean.