Skip to content

Channel service

A channel is a distribution partner: the ride-hailing app, the bank, the retailer or the portal through which a policy reaches a customer. Every coverage belongs to exactly one.

The service is control plane, so it is deployed once in Singapore and every market reads from it. Channel records carry no personal data, which is what allows them to sit centrally at all. Why that boundary exists is in Regions and data residency.

A channel record carries two identifiers. One is a machine identifier and the other is a human-readable code, and both are stable. An integration should key on the machine identifier: a code is chosen by a person, and people rename things.

It also carries the country it operates in, and that is what ties a channel to a market. It is also why a channel is reference data rather than transactional. A distribution partner operating in three countries is three channel records, because the coverage each one sells is issued in a different market by a different data plane.

You can read a channel you have an identifier for. Nothing lets you enumerate the channels you do not.

That is deliberate and it holds across the control plane. A list endpoint on shared reference data is a directory of who else is on the platform, and the platform carries competing distributors and competing carriers at the same time. Separation by identifier rather than by a filter applied on the way out is the same principle described in Tenancy and scopes.

The practical consequence for an integration is that identifiers are given to you during onboarding rather than discovered at runtime.