Bonusly partners with RWX to replace flaky GitLab CI/CD with a platform its engineers trust

About

Bonusly is an employee engagement platform that helps organizations build high-performance cultures through frequent, visible, and meaningful peer-to-peer recognition. Trusted by HR and people teams at Workday, ZipRecruiter, and Salesforce, Bonusly has been named a G2 Best Software Award winner and consistently leads G2's Employee Engagement category.

Challenge

Pipeline instability directly threatened development velocity at Bonusly

Bonusly operates in the highly competitive B2B SaaS vertical, where development velocity stands as a core competitive advantage. As the company scaled, attracting attention from leading enterprises across the globe, its engineering organization faced mounting pressure to accelerate platform updates.

That pressure exposed the fragility of the company's existing CI/CD pipeline. At the time, Bonusly was using GitLab Runners across a hybrid of self-managed and Git-managed AWS infrastructure. While this configuration was operational, it was slow, flaky, and too complex for the team to trust for frequent production deployments. “Our process to verify and deploy code would take about an hour, and around 40% of the time, it flaked,” shares Jeremy Thomas, SVP of Engineering at Bonusly. “That level of instability makes engineers much more cautious.”

Jeremy knew that the pace of code development was accelerating in the industry, but Bonusly's delivery system still encouraged engineers to move cautiously. Facing an hour-long cycle that failed nearly half the time, Jeremy's engineers started batching changes, shipping less frequently, and making each release larger to justify the wait. Every failure cost time, eroded confidence, and made shipping feel riskier than it needed to be. And unlike test failures that point to real regressions, these were infrastructure-level breakdowns that engineers couldn't fix by writing better code.

“We do developer surveys every three months, and it was clear that CI/CD was a big detractor in terms of sentiment,” shares Jeremy. "It came to a point where I said, okay, that's it, we're actually going to solve this problem now." Jeremy was not looking for a faster version of the same pipeline. He wanted to change the development conditions that made every deployment feel risky.

RWX entered the picture through an Ohio tech community connection. While Jeremy was already evaluating two established CI/CD vendors, both came back with pricing that didn't hold up for a company of Bonusly's size. RWX took a different approach. Before any commercial conversation concluded, their engineering team built a working proof of concept and showed Jeremy, with real data, that his pipeline could run faster and more reliably on its platform.

That effort, unprompted and before a contract was signed, solidified his confidence in the partnership.

Solution

A modern CI/CD platform that creates a trusted development loop for smaller, safer changes

Partnering with RWX enabled Bonusly to replace a brittle, serialized pipeline with a reliable way to validate changes, move dependent pull requests through in sequence, and keep each release small enough to understand.

RWX provisioned Bonusly's account directly, working with the team's cloud ops engineer to run a parallel pipeline to the existing GitLab setup. Jeremy observed empirical evidence that RWX was faster and more dependable than the existing GitLab setup before any cutover decision was made. Any configuration questions that surfaced during or after cutover were resolved quickly. "We have a shared Slack channel,” shares Jeremy. “We'll ask a question, and within 10 minutes they'll have a response that's informed by our situation."

At the architectural level, RWX replaced Bonusly's sequential, VM-centric pipeline with a DAG-based CI/CD runtime. This execution model does not run one large job on a shared machine. Rather, RWX treats every pipeline as a set of independent, discrete tasks, runs them in parallel wherever dependencies allow, and allocates right-sized serverless compute to each task in isolation. By running independent tasks in parallel and reusing unchanged outputs, RWX helps Jeremy lower the cost of each iteration. That also means Jeremy's engineers no longer batch changes together, get faster feedback on every run, and ship with the confidence that a slow, serialized pipeline had taken away.

RWX also changed how Jeremy's engineers structure their work. With its end-to-end deployment cycle measured in minutes instead of hours, engineers can keep dependent changes small, move them through validation in sequence, and ship each one without waiting through another unreliable cycle. Each release affects less surface area, making it easier to review, validate, and diagnose. "Smaller code changes are less risky just because there's less footprint that you're affecting,” explains Jeremy. That means engineers can now isolate failures in minutes rather than sifting through the surface area of a large, batched release.

RWX's content-based caching directly addressed another dimension of the speed problem. Rather than rerunning every task on every push, RWX derives cache keys from actual filesystem contents and task definitions. So, when code hasn't changed, prior outputs are reused automatically. For Jeremy's team, that meant the pipeline stopped doing work the current change didn't require. Engineers didn't need to configure or maintain a cache strategy to get there; the platform handles it by default.

On the rare chance that failures do surface, RWX ensures they are instantly traceable. When a run fails today, Jeremy's team knows it points to a real regression in the codebase, not that the system let them down again. That shift showed up directly in Bonusly's quarterly engineering sentiment surveys: production deployment went from the single most important thing engineers wanted fixed to the last thing engineers were concerned about.

Results

Cleaner signals, lower release risk, and a team that ships without hesitation

With RWX, Bonusly's engineers can keep changes small, move dependent pull requests through in sequence, and trust pipeline failures. As a result, they have established a faster, more reliable CI/CD pipeline that supports rapid production deployments, high success rates, and smaller, stacked PR workflows, significantly reducing release risk. What had once been the single most pressing issue across the engineering organization quietly dropped off the list entirely.

Since partnering with RWX, Bonusly has seen a 75% reduction in CI/CD pipeline execution time, a 65% improvement in deployment reliability, and a 96% NPS score across the development team.

Looking ahead, that delivery confidence matters as Jeremy continues to expand into purpose-built AI agents designed for specific professional contexts.