How one customer stays one customer across five channels

A customer asks a question on WhatsApp in the morning, fills in a form on your website at lunch, and calls in the afternoon. To most tools, that’s three different people. To your team, it’s three half-conversations they have to mentally stitch together — usually by asking the customer to repeat themselves.

Keeping one person as one person, across every channel, is the job of the contacts primitive. Here’s how we think about it.

The problem isn’t matching — it’s matching safely

It’s easy to merge two records when they share a phone number. The hard part is the grey area: a web-chat visitor with no identifier yet, a call from a number that isn’t saved, a name typed three slightly different ways. Merge too eagerly and you blend two real customers into one — a worse failure than not merging at all, because now the agent says the wrong thing to the wrong person.

So resolution runs on a spectrum of confidence, not a yes/no:

• Strong signals — a verified phone number, an email confirmed on a booking — merge automatically.
• Weak signals — a similar name, an overlapping device — associate the threads without collapsing them, and surface the maybe-match to your team rather than guessing.
• No signal yet — a fresh web-chat session stays its own provisional contact until the customer gives something that ties it to a known record (books, shares a number, logs in).

Why the agent cares

The agent reads the resolved contact before it replies. That’s what lets it say “welcome back” to a returning customer instead of re-asking what they already told you on another channel — and what lets a human stepping in see the whole story, not just the last message.

The honest trade-off: we’d rather under-merge and ask than over-merge and be confidently wrong. A split record is a tidy-up. A blended one is a trust problem. We optimise for the second never happening.