Rebuilding a revenue engine on one system of record.

Four systems, and a spreadsheet that decided the number.

Northbeam grew by acquisition. Each of the four regions it bought arrived with its own CRM, and none of them was ever replaced: the integration work was always scheduled for the quarter after next. By last year the group ran four separate pipelines: two instances of the same vendor's product on different major versions, one purpose-built tool for air freight, and one that had begun life as a shared mailbox.

Every Friday a commercial analyst exported all four to CSV, merged them by hand and produced the number the board saw on Monday. The merge took most of a day. It also required judgement, because the four systems disagreed about what a qualified opportunity was, and the rules that reconciled them existed only in her head and in the formulas of one workbook.

The cost was not the day. It was that nobody below board level trusted the output, so each region kept a second forecast of its own.

The hard part was not the data.

Moving 61,300 accounts and 14,200 open opportunities is a well-understood exercise. What took the time was agreeing what the records meant.

Three examples. The air freight team counted an opportunity from the moment a rate request arrived; the ocean team counted it once a quotation had been issued, typically eleven days later. Two regions used closed-won for a signed rate agreement, which commits nobody to any volume, and the other two used it for the first booking against that agreement. And because multinational customers had been sold to independently in each region, the same parent company existed four times over with nothing connecting the records.

• Nineteen deal stages across the four systems, mapping onto five the group could agree on.
• 8,940 duplicate accounts, of which 1,206 needed a person to decide rather than a match rule.
• A parent-child hierarchy that had never existed anywhere, built across 340 corporate groups.

One object model, then everything else.

Work started on the object model and stayed there for three weeks, which was longer than the client wanted. The rule we held to was that no field would be migrated unless somebody could say what a value in it meant and who was accountable for setting it. That removed 71 of the 214 fields present across the four systems.

Pulse then went in as the single pipeline. Gate provides the identity plane, so regional managers see their own region by default and group roles see everything, without a per-region permission table to maintain. Link carries bookings back from the freight operations system, so an opportunity closes when the first container moves rather than when somebody remembers to update it. Insight publishes the forecast the board reads, from the records the sales manager works in.

The Friday merge was retired in week nine. The analyst who used to run it now owns pipeline hygiene and territory design.