Field Sales Data Platform with GPS-Verified Activity

A commission-based field sales company ran its operation across a CRM, a call-recording tool, BI dashboards, and spreadsheets, none of which agreed. Pay depended on numbers nobody fully trusted. We built the platform that made them agree.

The problem

When commissions are computed from fragmented data, every payday is a negotiation. Field activity was self-reported, deal records lived in four places, and the schedule that decided what reps got paid was assembled by hand.

What we built

A production data platform on Postgres that ingests all four upstream systems continuously and derives one authoritative ledger:

• A single deal ledger. 11,769 deals across 168 active agents, 29 product providers, and 8 states, reconciled into one validated table.
• GPS-verified field activity. Every visit is checked against a 250-yard geofence. Of 7,291 submitted activities, 90.4 percent counted and 72.0 percent fully qualified; the boundary held to within a few yards in verification.
• A commission engine, written down. 56 rules encode the schedule that decides pay, so the computation is repeatable and auditable instead of tribal knowledge.
• Human-in-the-loop reconciliation. A 100-point, five-signal matcher pairs projected commissions against actually-paid statements and queues the ambiguous cases for review rather than guessing.
• An admin surface of thirteen screens for operations staff, backed by an ingestion daemon that polls upstream systems on a drift-corrected five-minute loop.

What it proves

Payment data is where trust in a company's numbers is won or lost. The reconciliation surfaced patterns the business could act on immediately, including that 25.0 percent of deals had missed a payment, a rate that varied roughly threefold across providers. That is the point of a source of truth: not prettier reports, but decisions grounded in numbers that survive scrutiny.