How Caribou saves 200+ engineering hours a year with Rootly.

To open an incident at Caribou, a single business analyst sat watching a Google spreadsheet manually, waiting for an alert to land so she could create the incident from it. Behind that spreadsheet ran a chain of tools that included Opsgenie for paging, Zapier for automation, plus Jira and Slack, with the Zapier glue snapping under load, which is exactly when an incident needs it to hold. This system was manageable at a smaller scale, but for a fast growing fintech responsible for a high volume of people's auto loans, it became a standing risk to the business. The team ripped the whole thing out, stood Rootly up inside a week, and got back roughly 200+ engineering hours a year.

"Rootly replaced five tools and changed our culture. We went from being afraid to call an incident to confident to call one." -Devree Lewis, Manager of Quality Engineering and Incident Management

A Franken system held together with duct tape

Devree Lewis, who manages quality engineering and incident management, called the old setup a “Franken system”. People hopped into a dedicated Slack channel, but there was no coordination of docs or artifacts. If an engineer wasn't online, Opsgenie paged them, and a true emergency meant calling all the alarms and alerting everyone. Retrospectives were a written-down process with no triggers, so they happened only when, and if, an engineer remembered to create one.

The pain concentrated in a few places:

• Manual admin work, the number-one problem. Paperwork, and stitching everything back together after the fact, with engineers going back and forth.
• Tool sprawl. Across Opsgenie, Zapier, Jira, Slack, and Sheets, things got lost in the seams.
• A compliance gap. For a fintech, there was no formal incident history beyond Slack and a retrospective doc that was not always written.

The trigger wasn't a single outage. It was constant growth meeting brittle glue. The Zapier automation kept breaking under duress, which is precisely when it was needed. Caribou needed to get rid of the manual work and the duct tape, which required  three things in particular: speed, governance, and real data, especially a true, captured mean time to resolution.

Up in less than a week, and a culture that changed

Caribou consolidated onto Rootly across on-call, incident response, and retrospectives. Rootly was up and running inside a  week, and the first incident ran through it perfectly, with no manual cleanup.

That first clean incident did something bigger than save steps. It changed the culture. Where the team had been hesitant to call an incident, knowing the overhead that followed, the new default became, "when in doubt, call one”. With the connective tissue, easy communication channels, and reliable documentation, declaring an incident stopped being a burden and became a quick, low-friction decision.

The operational gains followed immediately. The messy alert-and-response pseudo-system collapsed into one clean layer. Response SLAs are locked at five minutes, and nobody has trouble meeting them. And the coordination tax the team used to pay on every incident, gathering artifacts, chasing documentation, remembering the retro, is, in Devree's words, just gone.

"Rootly was up and running inside a week and the first incident ran through perfectly, no manual cleanup. Our response SLA is locked at five minutes and nobody struggles to hit it." -Devree Lewis, Manager of Quality Engineering and Incident Management

What changed

Retrospectives that always happen

The retrospective document is now created automatically and put on the calendar without anyone having to schedule it. Follow-ups flow into Jira through the integration, with SLAs attached, so the learnings convert into tracked, owned work rather than good intentions.

Calm, organized incidents

The workflow today is unrecognizable from the spreadsheet era. An alert comes in, or a human flags a problem, and Rootly already has the artifacts in place. The incident management team hops on, assigns roles, and the right teams are already present because of how alert sources route into the incident. Devree's description of it is simply “calm and organized”.

200 hours back, and real data for a fintech that's scaling fast

The headline business impact is hard hours. With Rootly, including its AI, Caribou gets roughly 200 engineering hours back per year that used to go to admin and manual paperwork, and that figure doesn't even count the business analyst's time. The second impact is data. For the first time, Caribou has real incident histories and real timing it can measure and send to leadership. That matters because Caribou is targeting aggressive growth and shipping that fast safely requires leadership to see what reliability actually costs. The data lets them make the call on what risk is worth taking, and when to push or pull back.

Proof points (Caribou lightning round, 1-5)

• Pager fatigue is lower: 5
• Time to first status update improved: 5
• On-call handoffs are clearer: 5
• Retrospective completion rate improved: 5
• Stakeholder communications are more consistent: 5
• Customers are happier: 5
• Engineering morale is in a better place: 5

There's a version of incident management that runs on duct tape; a paging tool here, an automation service there, a spreadsheet someone watches manually, and a retro that happens if anyone remembers. It holds together until growth puts it under load, which is exactly when the glue snaps. Caribou replaced all of it with one clean layer, and got back 200 engineering hours a year, automatic retros, a five-minute response SLA, real data for its leadership team, and a culture that now says, “when in doubt, call an incident”. For a fintech scaling fast, that's not just tidier tooling. It's the governance and speed the business runs on.