Estimated reading time: 4 minutes
Engineering leaders know the challenge: you need to modernize legacy infrastructure, but how do you prove the investment is worthwhile? And once you’ve made changes, how do you know they’re actually working?
Matt Buckley faced this exact dilemma at tax compliance software company Avalara, where his team managed Value-Added Tax (VAT) reporting products built from decades of acquisitions and integrations. With billions in tax remittance flowing through their systems and potential liability in the hundreds of millions, every deployment carried significant risk.
“We had subject matter experts, manual regression testing, conversations about ‘well, maybe it’s this, maybe it’s that,’” Buckley says. “But we couldn’t quantify the ROI of proposed modernization efforts or identify actual development bottlenecks.”
Start with measurement
Buckley soon realized that he needed to establish some baseline metrics before changing any systems.
He turned to Uplevel to track velocity, quality, and deep work metrics before Avalara adopted GitLab, a single end-to-end DevSecOps platform.
The aim was to simplify the teams’ cumbersome toolchain, automate security standards, and increase visibility across the software development lifecycle.
The initial assessment “was pretty terrible. Merge requests were taking days to reach production because of the complexity of our current toolchain,” he says.
Your inbox, upgraded.
Receive weekly engineering insights to level up your leadership approach.
Expect new bottlenecks
Buckley’s team executed a comprehensive modernization of the core infrastructure supporting Avalara’s tax returns products. The transformation involved first containerizing all legacy applications, then implementing shared GitLab CI/CD pipelines that deployed to centrally managed Kubernetes clusters.
Previously, some individual teams had been using GitLab pipelines but deploying to their own isolated lower environments, only bringing code together in production – a process that frequently caused integration issues.
The new approach created consistent pipelines from development through production. The reliability engineering team managed the underlying Kubernetes infrastructure, while development teams focused on deploying their applications through standardized shared pipelines.
“Once we’d implemented shared pipelines, we had a different bottleneck – teams treading on each other’s toes,” Buckley explains. “When somebody breaks the pipeline, no one owns it. Once the lower environment is broken, no one can push changes to higher environments, blocking everybody.”
Without measurement, this could have derailed the entire initiative. Instead, throughput and quality metrics immediately surfaced the new constraint. Buckley’s team implemented a rotating “champion” system where team members alternated responsibility for unblocking shared infrastructure.
Use data to tell the modernization story
The results were dramatic:
- Merge request cycle time went from 20+ hours to 3 hours (85% reduction)
- Merge request throughput jumped from 0.7 to 2.2 PRs per developer per week (230% increase)
- Deployment frequency shifted from 30 annually to 30 monthly
But perhaps more importantly, the data told a story that resonated with business stakeholders.
“I used Uplevel data to communicate about the transformation to non-technical stakeholders,” Buckley says. “The graphs were clear – it was easy to point out where metrics were low, then got higher and higher.”
Product Director Rachael Marshman observed the customer impact: “We’re gaining confidence with our customers by making strong predictions and delivering features on time, planning roadmaps more effectively, and getting feedback more quickly.”
More like this
Quality vs quantity
Traditionally, one risk of focusing on velocity is sacrificing quality for speed. Buckley’s team tracked bug rates alongside productivity metrics, ensuring they maintained quality standards throughout the transformation.
“We maintained the same ratio of complex vs. simple merge requests even as volume increased,” he says. Bug rate remained below their benchmark for the duration of the project, proving the improvements were sustainable.
Build organizational momentum with proven results
The quantified success created organizational adoption beyond Buckley’s immediate team. Other teams developing Avalara’s tax returns products implemented similar approaches, and even mission-critical systems handling massive financial exposure, migrated to the new platform within six months.
“A lot of what I did was present the case for what we did and how that was more effective both for my teams directly, but it also justified it to all the other teams in our division,” Buckley explains.
The key was having independent validation. As Buckley puts it: “It’s another set of data, arrived at independently, which showed clear improvement at the exact moments leading up to when we delivered features.”
Final thoughts
1. Measure before you move. Establish baseline metrics before any infrastructure or technology changes. You can’t demonstrate improvement without knowing your starting point, and initial measurements often reveal unexpected problems.
2. Plan for new constraints. Successful optimizations often create new bottlenecks because there is no such thing as a “lift and shift” approach when you’re dealing with complex organizational processes. Continuous measurement helps you identify and address these quickly rather than letting them derail your initiative.
3. Let data drive adoption. Quantified results are your best tool for organizational change management. Clear metrics make the case for broader adoption far more effectively than technical arguments alone.
November 3 & 4, 2025
Book now and save up to €425 💸
The transformation enabled Avalara to do more with existing resources, while improving the customer experience through more reliable delivery. But the real lesson isn’t about any specific technology – it’s about using measurement to convert technology investments from leaps of faith into strategic business decisions.
“There’s been a ton of appetite to do more in the ways that Avalara works. There hasn’t always been a great way to tell the impact of that investment. This helps tell that story,” Buckley says.
Promoted Partner Content
