The standard way to launch a brokerage is to buy a trading platform from one vendor, a CRM from another, and a client portal from a third. Each is individually reasonable. Together they create a permanent job nobody costed.
What the job actually is
Three systems means three ideas of who a client is. The platform knows a trading account. The CRM knows a lead who became a customer. The portal knows a login with documents attached. Nothing links them except an integration somebody wrote, and the integration is only ever as fresh as the last time someone maintained it.
The daily consequences are mundane and expensive:
- A client asks why their withdrawal is pending. The desk can see the request in the portal and the balance in the platform, but not the verification hold that caused it.
- Sales calls a client who took a heavy loss that morning, because the CRM has no idea.
- Compliance asks for a client's full history and someone assembles it from three exports.
- A leverage change applies in one system and not another, and nobody notices until an order behaves unexpectedly.
None of these is a crisis. All of them are a tax.
What "one record" changes
In Raptor the CRM record is the client record. The portal writes to it, the terminal authenticates against it, the back office administers it, and reporting reads it. There is no synchronisation step because there is nothing to synchronise.
Concretely:
- Verification state is visible to the desk at the moment a client calls, including why a document was rejected.
- Funding state has one value. Pending means pending in every screen.
- A permission change applies immediately everywhere, because permissions are evaluated in one place and enforced server-side.
- Trading activity is part of the relationship, not a separate report — what they trade, in what size, in which session.
The part that is genuinely harder
A shared data model is less flexible than three loosely coupled systems. If you want a bespoke CRM field that the platform must respect, that is a change to one model rather than a mapping in an integration layer. We think that trade is correct — a mapping layer is where correctness goes to die — but it is a real trade and worth naming.
The other honest constraint: migrating onto a single model is more work up front than pointing an integration at an existing vendor. The payoff is that the work finishes.
What to ask any vendor
Not "do you integrate with X". Ask:
- Where does the client identity live, and how many copies of it exist?
- If I change a client's permission, which systems learn about it, and when?
- Who owns reconciliation when two systems disagree?
The third question is the one that reveals the answer.
Related: Raptor CRM · Client Portal · Back Office