Brokerage risk platform

At a glance

Trading accounts in production

1,000+

Developers led

1 to 2

On site through go‑live

4+ months

Challenges and solutions

1. Duplicate and conflicting operations

   Challenge

   A retry, a double click or two people acting on the same account could trigger the same money-moving operation twice.

   Solution

   Operations were made idempotent, so repeating one has no further effect, and the number that can run at once for an account is capped.

   Trade-offThroughput is limited on purpose: where money moves, correctness comes before speed.

   class ProcessOperation implements ShouldQueue, ShouldBeUnique
   {
       // The same operation is never queued twice.
       public function uniqueId(): string
       {
           return (string) $this->operation->id;
       }
   
       // One operation per account at a time.
       public function middleware(): array
       {
           return [(new WithoutOverlapping($this->operation->account_id))->releaseAfter(10)];
       }
   }

2. Reports slowing down as data grows

   Challenge

   Report tables gain rows every day, and reports had to stay quick without throwing history away.

   Solution

   The largest tables are partitioned by week, and a scheduled job adds the next partition automatically, so a report reads only the periods it needs.

   Trade-offPartition maintenance becomes a scheduled job that must never be missed.

3. Spotting risk without manual review

   Challenge

   Unusual trading activity had to be caught early, across thousands of trades a day.

   Solution

   Automated rules cross-check activity against several independent signals and raise alerts for the operations team, instead of relying on someone noticing.

4. Self-service that stays secure

   Challenge

   Clients wanted balances and bills on their phones without installing another app.

   Solution

   A messaging bot verifies every request with signed payloads, asks for a PIN before sensitive actions, and delivers bills as PDFs.

   Trade-offOne extra step for clients, so an open chat alone is never enough to change anything.

How it fits together

A brokerage needed one back office for operations, risk and billing, fed by trading systems it does not control.

1. 01IngestScheduled jobs collect data from external trading systems
2. 02StoreSeveral MySQL databases, partitioned for scale
3. 03MonitorRules flag unusual activity for review
4. 04ActBilling, alerts and client self-service