Built by engineers who got tired of watching Lighthouse lie.

Grace Yoon spent the better part of a decade as a frontend performance engineer at consumer-scale products — the kind where a 100ms LCP regression translates directly to a measurable conversion drop. At each company, the same pattern played out: the Lighthouse CI check passed, the engineering blog celebrated the score, and the product team quietly watched mobile engagement metrics decline.

The problem wasn't that anyone was lying. It's that lab tools can only measure what a lab can control. They run a single pass on a configured connection from a fixed location. Real users arrive on mid-range Android devices, on variable LTE, having already loaded four other tabs — and they carry the full weight of every third-party script your growth team added in the previous sprint. The lab can't see the tag manager container loading a retargeting pixel synchronously. It can't see that your p75 on mobile 4G is 4.8 seconds while your p50 on desktop fiber is 1.2 seconds. The field data diverges, and the lab data gives everyone a clean conscience.

"I spent two years instrumenting PerformanceObserver events in production scripts just to get the numbers I needed. The tooling shouldn't require that. p75 on real devices should be the default view, not a bespoke analytics query."

— Grace Yoon, CEO & Co-Founder

Grace and co-founder Dmitri Voronov started building Pulsecadence in Seattle in 2021. The goal was narrow by design: a RUM tool that tracks p75 on real user devices, attributes render-blocking delays to specific third-party scripts, and fires on real-user data when a deploy breaks your percentiles — not a general observability platform with a RUM module bolted on. Pulsecadence does not replace your APM or your synthetic uptime checks. It measures the one thing those tools can't: what the page actually felt like to the user who just bounced.

Pulsecadence raised angel backing in December 2025 and is currently a team of eight people based in Seattle, WA.