Server response time isn't user experience.
The number that matters is what the browser paints. Everything else is a proxy for that, and most proxies are wrong at peak.
About
Built because the existing tools couldn't tell us what we needed.
Evaluat is a real-browser performance testing platform. We started building it because we got tired of "all green" load test reports that didn't predict the production incidents that actually mattered.
I spent a decade running performance work on large Adobe Commerce / Magento sites. Black Friday peaks. Multi-region launches. The kind of e-commerce where 200ms of LCP is a million-pound conversion problem.
The pattern was always the same. We'd run k6 or JMeter against the API for weeks before peak. The reports came back green. Then the day arrived and customers were watching the product page render in 6 seconds, with a layout shift that pushed the buy button below the fold, while Hotjar and Cookiebot fought each other for the main thread.
The HTTP layer was fine. The user-visible experience was a disaster. We had no way to measure the gap.
Evaluat started as a private tool. One real browser per virtual user. Web Vitals captured at load. Session video so you could watch the moment a step broke. After a couple of years using it on production e-commerce and getting hounded by friends who wanted access, we turned it into a product.
It's still in private access. We're working directly with each customer because the product is sharper when we do. The marketing site is dry on purpose. The product is what we're proud of, not the brand.
— Ahmad Farzan, founder
What we believe
Five things we keep coming back to.
The numbers should match what the customer sees.
If your test report says the page loads in 800ms and the customer sees 6 seconds, the test report is the broken thing. Not the customer.
Forensic detail beats aggregate percentiles.
When something breaks for 14 users out of 42,000, p99 won't show it. You need to open the session for one of those 14 and watch what happened.
Pre-launch software shouldn't pretend to be a brand.
This site is intentionally restrained. The product talks. We're not interested in graphics that promise revolution.
Demo-led beats demo-canned.
Every demo is on your real site, not ours. The report is yours to keep whether you sign up or not.
Build once. Use everywhere.
The scenario you write for a performance test should also be your CI smoke test and your continuous monitor. Three products, one foundation.
Team
Small, technical, and talking to customers directly.
Evaluat is founder-led right now. As we grow, this section will grow with us.
Ahmad Farzan
Founder · Engineering
A decade running performance and DevOps on large Adobe Commerce / Magento e-commerce sites. Got tired of HTTP-layer load tests that missed everything customers actually noticed.
LinkedInHiring
UK · Remote · Soon
We're starting to think about hire #2. Browser internals, performance engineering, or product design. Drop us a line.
Where we are
UK-based. Customer-led. Working in the open.
Evaluat Ltd is incorporated in the UK. We work mostly with EU and US customers, with EU and UK data residency commitments on request. There's no investor deck, no Series A press, and no media training. We talk to customers, we write code, we ship.
We publish a public changelog. We're slowly building a blog of post-mortems and methodology pieces. If you want the same content earlier, just talk to us.