Every CRM ambition we hear — hyper-personalisation, next-best-action, AI-driven journeys — depends on an unglamorous foundation: whether the data underneath is true. After years of auditing enterprise CRM estates, we can usually predict a programme's ceiling from one number: what percentage of its customer records it can actually trust.
Duplicates are a strategy problem wearing a data costume
A member with three profiles gets three inconsistent experiences and generates three inconsistent analytics rows. Deduplication is not a one-off cleanup; it is a standing capability, with survivorship rules the business has actually agreed to. Decide once, in writing, which source wins for name, for phone, for consent — then automate it.
Identity resolution before personalisation
You cannot personalise for a person you cannot recognise. Before investing in orchestration tools, invest in the match keys:
- Deterministic first — loyalty ID, verified email, verified phone. Probabilistic matching is a last resort, not a default.
- Instrument the join points — app login, till prompt, e-receipt — where anonymous activity becomes known.
- Measure match rate as a KPI. It is the ceiling on every downstream use case.
Consent is data quality too
A perfect profile you are not allowed to message is operationally identical to no profile. Consent flags deserve the same rigour as balances: a single source of truth, an audit trail, and propagation to every channel tool within minutes, not nightly.
The cheapest personalisation upgrade on the market is fixing the data you already own.
Where to start on Monday
Pick your top revenue journey, trace one real customer through every system it touches, and write down every place their record forks, stales, or loses consent. That one-page map will out-perform any tooling RFP. When you want a partner for the fix, we should talk.
