Capacity rehearsals
Push critical paths to 5x, 10x, 20x normal traffic. Find the third-party tag, the slow query, or the cache miss that breaks first.
Performance Testing
Performance testing that runs real browsers.
Push your site to a target concurrent user count. Measure what real browsers actually see. Get a per-session report on every one of those users, with Web Vitals, network, console, and video. Not a request script's guess at it.
Performance Testing is live now. Testing Suite and Monitoring are coming.
Black Friday checkout rehearsal
Run test Target
2,000
concurrent users
Duration
30min
10 ramp · 15 hold · 5 down
Region
London
eu-west · en-GB
Traffic shape one real browser per user
How it runs
Build a scenario. Pick a traffic shape. Hit go.
A performance test in Evaluat starts with a scenario: the journey you want virtual users to take. Homepage to checkout. Search and add to cart. Login and load the dashboard. You build it once in the visual editor.
Pick the traffic shape. Ramp 1,000 users over 10 minutes, hold them for 30, ramp down. Or fire them all at once. Or scale up to 10,000 if your peak is a sale launch.
Every virtual user runs in a real browser, in your chosen region, with your chosen locale and viewport. Web Vitals captured at load. Session video, network logs, and console output recorded for every one of them.
Want to see the per-session artifacts before booking anything? The free website speed test runs one page load the same way: same browser, same vitals, same video, on a shareable report.
What Evaluat calls performance testing is what other tools call load tests, stress tests, or spike tests. Different traffic shapes, same configuration. We use the umbrella term because the differences are knobs on one test, not separate features. Set the ramp-up, the steady-state, the ramp-down. Pick the scenarios. Hit go.
Configure the test
All the dials you'd expect. None you wouldn't.
Every performance test in Evaluat is built from the same six pieces. The dials are the same on every plan; plan limits scale with the tier.
- Load mode: Duration (fixed time, target concurrency) or Sessions (target completed runs).
- Users and duration: concurrent target and runtime. Minimum ramp rates kept realistic.
- Traffic shape: ramp-up, steady-state, ramp-down. Each phase configured separately.
- Scenarios with weights: multiple journeys in one test, distribute 100% across them.
- Region, timezone, locale, viewport: match the customer segment you're modelling.
- Popup handler toggles: pick which auto-dismiss rules apply for this run.
Test plan
6 dials Load mode Duration Sessions
Users & duration 2,000 · 30m
Traffic shape
Scenarios & weights 2 journeys
Checkout 60% Search 40%
Region · locale · viewport London · en-GB · 1280×800
Popup handlers
Cookie consent
Build a scenario once. Use it everywhere.
Scenarios are reusable.
A scenario you build for a 1,000-user performance test runs as a deployment smoke test in CI and as a 5-minute production monitor. Same definition, same configuration UI, three different lifecycles. The maintenance burden is one place, not three.
Parameterise the scenario with datasets so virtual users follow different paths. Wire a project-level popup handler once so the cookie banner doesn't break every script.
More on scenarios
What you get back
Five views over the same run.
Every test produces a five-view report. Aggregate enough for the executive summary. Detailed enough to find why something broke for 10 users out of 1000.
- Overview: active users, sessions completed, failed sessions, time-series Web Vitals.
- URL performance: every URL hit, per-URL Web Vitals, status distribution, timing breakdown.
- Sessions: every virtual user's session, individually addressable, with video.
- Console logs: every browser message, deduplicated and counted across the run.
- Network logs: every HTTP request, filterable across millions of rows.
Reports have stable URLs and can be shared read-only with people outside the team.
What teams use it for
From peak-day readiness to "what just broke?".
The patterns we hear most often on demo calls. Forensic detail beats aggregate percentiles, and every one of these is a test you can run today.
Release validation
Compare two test runs side-by-side. If LCP regressed by 400ms after this deploy, you'll see it before customers do.
Web Vitals budgets
Set thresholds on LCP, INP, CLS. The test fails when the budget breaks. Stop shipping regressions.
Third-party tag analysis
Analytics, A/B testing, consent banners. Measure exactly how much each one costs you in INP under load.
Regional performance
Run from multiple regions with the right timezone and locale. See what customers in Amsterdam actually experience.
Forensic debugging
Something failed for 12 sessions out of 40,000. Open one. Watch the video. One-hour root cause, not one-week.
Common questions
Performance testing, answered.
What browsers does Evaluat use?
Chromium-based browsers, the same engine your customers run as Chrome and Edge. Tell us on the demo call if you have a specific Firefox or WebKit requirement.
Which regions can I run tests from?
London (UK) and Frankfurt (Germany) are live today. Tell us what region you need on the demo call. We add regions when customers ask.
What's in a test report?
Five views: Overview (aggregate Web Vitals, time-series), URL performance (every URL with its own metrics), Sessions (every virtual user, individually addressable, with video), Console logs (deduplicated and counted), Network logs (every HTTP request, searchable across millions). More on reports.
What does "isolated browser" mean?
Each virtual user runs in its own browser instance. Its own memory, CPU, cache, cookies, and network stack. Nothing crosses between users. The contention you measure is real, not what a shared-browser model approximates.
How is that different from headless testing?
Headless browsers (puppeteer, playwright in headless mode) skip the rendering pipeline. Great for end-to-end test automation. Useless for accurate paint timings. Evaluat runs full browsers with the rendering pipeline so Web Vitals match what users actually see.
Does Evaluat work for SPAs?
Yes. Because each virtual user is a real browser, single-page apps render and execute exactly as they would for real users: client-side routing, hydration, lazy-loaded chunks, the lot. Web Vitals come out matching production.
Can I export the raw data?
Yes. Aggregated metrics export as CSV from every report view. Raw console logs are downloadable per session.
How many concurrent users can a test run?
Plan-dependent. Higher concurrency targets are available on Growth and Scale. Tell us your peak concurrency on the demo call and we will size the right plan.
More answers on the FAQ page.