Anonymized. Client details available under NDA.
The engagement started where most do: the platform was in use, but nobody trusted it. Marketing leadership wanted three things: resolve the usability failures that were slowing the team down, reduce audit and compliance risk, and get to a point where modern marketing (preference management, personalization, chat) was possible.
What we found
| Area | Before |
|---|---|
| Data model | No data dictionary. 356 duplicate fields. |
| Database | 139,054+ records created by an integration with no documentation of what they were or why. |
| Compliance | No deceased-member suppression. No written procedures and no proof of opt-in. |
| Attribution | 151,966 leads with no acquisition program, so no way to tie marketing to outcomes. |
| Deliverability | No DMARC policy in place. |
| Access | Manual user management; no single sign-on. |
| Reporting | Inconsistent success definitions across programs; reporting anomalies nobody could explain. |
What phase one delivered
Four months, foundation first.
- Data dictionary published. For the first time the team could see what each field meant and where it came from.
- Database cleansed. The 47,711 records broke down into 60 deceased, 12,422 unmarketable, and 35,229 inactive since January 2023. Deceased suppression is now automatic.
- One principle made explicit: member preferences are never overwritten by the core banking system. That single rule removed a whole class of complaints.
- Personalization rebuilt. Tokens were rebuilt and tested in Outlook (the client's dominant inbox); reusable snippets and fragments put in place.
- Send policy set. Communication limits of two emails a day and five a week, active.
- Standard templates, tags, and channels so programs report consistently, with segment analysis by product, persona, and behavior.
- Unsubscribe reason captured as a KPI, so the preference center launching next would have a baseline.
How the roadmap was sequenced
The order is the lesson. Nothing in phase four is possible until phases one and three are done.
| Phase | Focus | Contents |
|---|---|---|
| 1 · Build | Foundation | Audit, usability fixes, the 47K cleanse, segmentation |
| 2 · Enable & launch | Adoption | Training and documentation, preference center and compliant footers, a 90-day chat pilot |
| 3 · Govern & secure | Risk | DMARC, single sign-on and provisioning, written procedures and a RACI |
| 4 · Scale & attribute | Measurement | Multi-touch attribution, personalization at scale |
At the end of phase one, the risks still outstanding were listed plainly for leadership: the missing DMARC policy, the absence of procedures and opt-in proof, the 151,966 leads without an acquisition program, and manual access control. A roadmap that only lists wins is a sales deck.
What to take from this
- Publish the data dictionary before you clean anything. You cannot decide what to delete until you know what the fields mean.
- Cleanse in categories (deceased, unmarketable, inactive) with counts for each. It makes the decision defensible and the legal-hold conversation short.
- Write the one or two principles that settle recurring arguments, such as which system wins when preferences conflict.
- Sequence governance before attribution. Attribution built on an ungoverned instance will be rebuilt.
Want this done in your instance? Most engagements start with an audit.
Book an intro call →