Harborline Cloud's engineers were spending more time filing infrastructure tickets and waiting on provisioning than actually shipping features. We built a self-service developer platform on Backstage, with golden-path templates and automated Terraform provisioning, turning a two-day wait into a twelve-minute self-serve action.
Harborline Cloud's engineers spent more time provisioning infrastructure than shipping features.
Every day an engineer spent waiting on infrastructure provisioning was a day not spent shipping product — and the bottleneck was entirely process, not technical capability.
Internal platform, not a revenue product — its "customers" are Harborline's own roughly 120 engineers, so success is measured in engineering time recovered, not direct revenue.
Internal developer platforms and developer-experience tooling, an increasingly standard investment area for engineering organizations past a certain size.
Manually managed infrastructure via a small platform team, with inconsistent, undocumented provisioning patterns accumulated across product teams over several years.
Not customer-facing competitive pressure — the pressure was internal: engineering time lost to provisioning delays was a direct, growing cost as the engineering org scaled past what manual provisioning could support.
Reviewed a year of infrastructure request tickets to quantify how much engineering time was actually lost to provisioning delays.
Interviewed engineers across six product teams about their actual infrastructure needs versus what they'd been provisioning out of habit.
Catalogued the inconsistent infrastructure patterns different teams had accumulated, to identify realistic golden paths rather than an idealized one.
Build on Backstage rather than a fully custom internal portal.
Reasoning
Backstage's plugin ecosystem covered service catalogs and templating out of the box, avoiding months of building portal infrastructure from scratch.
Trade-off
Some Harborline-specific workflows needed custom Backstage plugins rather than fitting an off-the-shelf pattern.
Define a small number of golden-path templates rather than fully flexible provisioning.
Reasoning
Unlimited flexibility was how the inconsistent infrastructure sprawl happened in the first place; opinionated defaults with an escape hatch struck a better balance.
Trade-off
A few teams with genuinely unusual requirements needed to work outside the golden paths, requiring manual platform team support.
Attach cost estimates to every template before provisioning.
Reasoning
Self-service needed to come with visibility, not just speed — teams needed to see the cost implications of their choices upfront.
Trade-off
Added engineering work to integrate cost estimation into the provisioning flow, delaying launch by roughly two weeks.
Developer
The platform's actual customer — every design decision was evaluated against their experience.
What happened
Terraform state conflicts emerged once provisioning became frequent and self-service rather than occasional and manual.
How we resolved it
Moved to per-service Terraform state files with remote locking, isolating conflicts to a single service rather than a shared state file.
What happened
A few teams' genuinely unusual infrastructure needs didn't fit any golden-path template, risking those teams feeling unsupported.
How we resolved it
Kept a documented manual escalation path to the platform team for non-standard requests, rather than forcing every case into a template.
Environment provisioning
Before
2 days
After
12 min
Approach
We built a self-service developer platform with golden-path templates and automated environment provisioning.
Outcome
New service provisioning time dropped from days to minutes, and deploy frequency tripled.
Environment provisioning
Deploy frequency
Platform team ticket volume
“New service provisioning went from two days to twelve minutes, and our platform team's ticket volume dropped 58%. The golden-path templates solved our common case so the team could focus on what actually needed judgment.”
Elena Vasquez
Head of Platform Engineering, Harborline Cloud
Two additional golden-path templates for less-common but recurring service shapes
Client planAutomated cost anomaly alerts for provisioned services, beyond the upfront estimate
Byld IQ recommendationExtending self-service provisioning to the remaining teams still on the manual ticket process
Client planBuildPath turns your version of this problem into a personalized roadmap in about three minutes — or talk to Byld first if you still have questions.