Walk into the operations room of a mid-size brokerage and count the browser tabs: the trading platform’s manager console, a CRM, two PSP dashboards, a KYC portal, a reporting tool, and the spreadsheet that reconciles the other six. Each tool was a rational purchase. The mesh they form was never designed by anyone — and it has properties nobody chose.
The visible costs are the small ones
Licence stacking, swivel-chair workflows and duplicate data entry are the costs that make it into procurement decks. They are real and they are minor. The expensive properties of a fragmented stack are structural:
- Reconciliation as a permanent job. When client state lives in five systems, agreement between them is not a fact — it is a nightly task that sometimes fails, and the failure mode is money moving on stale information.
- Incidents without an owner. A withdrawal goes wrong somewhere between platform, CRM and PSP. Three vendors each demonstrate, correctly, that their component behaved to spec. The client does not care, the regulator does not care, and the post-mortem dissolves into a diagram. Accountability diffused is accountability destroyed.
- Risk context that never arrives. The person approving a withdrawal sees the payment queue. The flow-quality picture lives in a different system, owned by a different team, on a different refresh cycle. The mesh doesn’t just slow work — it structurally separates decisions from the context they need.
What consolidation should mean
The answer is not “one giant monolith” — brokerages are right to be wary of betting the operation on any single vendor’s everything-box. Consolidation worth doing means three specific things. One data model: the client, their accounts, their money and their risk profile as a single record that every workflow reads and writes. One workflow layer: back-office actions — provisioning, adjustments, approvals — executed through governed, maker-checker workflows rather than scattered admin consoles. One accountable party: a partner whose contract covers the outcome, not a component, so “whose fault” has one answer before the incident, not after it.
The questions that reveal mesh cost
Three diagnostics for your own operation. When a client’s withdrawal is approved, does the approver see the client’s full risk context on the same screen — or does that live in another tool? If two systems disagree about a balance tonight, which one wins, and is that written down? And for your last operational incident: how many vendors were in the post-mortem, and did any of them own the outcome end to end?
Firms that can answer those questions cleanly have usually already consolidated — or built internally at real cost. Firms that cannot are paying mesh tax daily, invisibly, in the currency of reconciliation labour, incident duration and decisions made without context. The tax compounds with volume, which is why the mesh feels fine right up until the quarter it doesn’t.