The Problem

Northlane's freight coordination platform had been built in stages by three different development teams over four years. It worked, mostly, but the codebase had become genuinely difficult to extend. New features took two to three times longer to build than the business needed, integration requests from enterprise clients were regularly declined because the architecture could not support them, and the engineering team was spending more time on maintenance than on development.

There was also a performance problem that was becoming a customer problem. The platform handled real-time shipment tracking across thousands of concurrent loads, and latency had increased as volume grew. Operations teams at client companies had started keeping spreadsheet backups of critical shipment data because they could not trust the platform to be current.

"Our developers were good. The codebase they had inherited was not. We needed to separate those two things."

CTO, Northlane

The business case for a rebuild was clear. The risk of the rebuild was also clear: Northlane could not go offline. The platform operated 24 hours a day and any migration strategy had to account for continuity across the entire client base.

The Solution

We conducted a two-week technical assessment of the existing codebase before recommending a path. The recommendation was a strangler fig migration: rebuilding the platform module by module, running old and new in parallel, and switching client traffic over incrementally as each module was validated. Slower than a full rebuild, but eliminates the cutover risk entirely.

The new architecture was designed around an event-driven model that handled the real-time tracking requirements at significantly lower latency, with a clean API layer that made enterprise integrations straightforward rather than exceptional. The data model was normalized in a way the original had never been, which resolved a category of sync errors that had been generating manual intervention from the operations team every week.

"The migration was invisible to our clients. That was the whole point, and they pulled it off."

VP Operations, Northlane

The internal tooling built alongside the new platform gave Northlane's engineering team deployment and rollback capabilities they had not had before, substantially reducing the risk profile of future releases.

The Result

Migration completed over 14 weeks with no client-facing downtime. Platform latency for real-time tracking dropped by 67%. Feature development velocity increased, with the engineering team reporting that new capabilities that would have taken six weeks on the old codebase were shipping in two. Three enterprise integration requests that had previously been declined were completed within the first quarter after migration.

The spreadsheet backups that client operations teams had been maintaining disappeared within two months, which the account management team tracked as an informal but meaningful indicator of restored trust in the platform's reliability.