Test with AI

Performance and load testing services, measured against targets agreed up front

Before a launch, a sales peak or a growth step, our performance testing services find out how your system behaves under load before your users do. We run load, stress and soak tests against response times and traffic levels we agree with you before the first run. Performance test automation means the tests live in your repository and can be rerun before each release or in CI, so a slowdown between releases shows up as a failed check. Every script change is reviewed by an engineer. You learn, before the launch, whether the targets hold at the load you expect. We start with a one-week performance test plan.

Performance plan, targets set before the first run 1 passed 1 failed 2 review
  • Search results page at launch traffic Response time target agreed before the first run scripted
  • Checkout during a sales peak Needs a test environment close to production and realistic data environment
  • Nightly report export Slow degradation shows only in a long run soak
  • Partner payment provider Not loaded without its owner's written agreement out of scope
Illustrative example of a plan worksheet. Not client data.
  • 42 completed projects
  • 32 test automation engineers
  • Founders of CodeceptJS and Codeception
  • Founders of Testomat.io and Testomat.ai
  • Since 2015
  • 4.9 out of 5 on Clutch

At a glance

What it is
Performance and load testing
First step
Performance test plan About one week
You receive
Targets agreed, such as response times, throughput, error rates and the traffic level +3 more
After that
Test runs Two to six weeks

What we test

Services in this area

  • Load and stress testing before a launch or a peak

    The system is loaded to the traffic level agreed in the plan, then beyond it, to see where it slows down and where it breaks.

    • Response time, throughput and error rate against agreed targets
    • The slowest steps named with the evidence
  • Soak testing for slow degradation

    A long run at steady load shows what a short one cannot, such as a resource that fills up over hours.

    • Run length agreed in the plan
    • Results compared with the baseline
  • Capacity checks for growth

    When traffic is expected to grow, the load is raised step by step to find the level at which the targets stop holding.

    • The load level at which the targets stopped holding
    • What was not tested, written down
  • Performance checks in CI against a baseline

    Short checks against the baseline run on a schedule or before a release, so a slowdown between releases fails a check.

    • Scripts in your repository
    • Reruns after fixes

Performance testing services: what a performance test answers

A performance test answers four questions about your system at a given load: how long a request takes, how many requests it handles, how many fail, and where it breaks first. The answers mean something only against targets, so the targets are agreed before the first test. They come from your traffic, a launch or a sales peak, and not from our guess.

Load testing services, stress testing services and software performance testing services are the same work here: a plan, scripts, runs and a report against targets. If you are comparing a performance testing company with others, ask what its report would show; ours shows the targets, the result against each, and what was not tested. Why the tests are worth automating is in our post on the benefits of automated performance testing.

Test types: load, stress, spike, soak and capacity

  • Load: the system at the traffic you expect.
  • Stress: the system above the expected traffic, to see how it behaves at its limits.
  • Spike: a sudden, short and large increase in traffic.
  • Soak: a steady load held for a long time, to find slow degradation.
  • Capacity, also called breakpoint: the load raised step by step, to find where the targets stop holding.

The plan chooses which of these the engagement runs. Teams use these names differently, so the plan says what each run means.

What we put under load

The parts of the system agreed in the plan. Examples are the requests behind your busiest web pages, APIs and back-end services, the services behind a mobile app, or a database with production-sized data. Whether those parts work correctly is a separate question, covered by API testing, web test automation and mobile test automation. If you run an online shop, see e-commerce testing.

Environment, data and safety

Numbers are only as good as the environment they come from. The plan names a test environment as close to production as is practical, and test data at realistic volume. Load is never pointed at production, or at a third party, without written agreement. The report says how the environment differs from production, so you can judge how far the numbers carry.

What the report shows

  • Results against each target, with pass or fail.
  • The slowest steps, with the evidence from the runs.
  • A baseline that can be rerun.
  • What was not tested.

A debrief session walks your engineers through it.

Performance tests in your CI

Performance tests are code. Ours live in your repository next to your functional tests, are reviewed before they are merged, and can be rerun by your team after we leave. Short checks against the baseline can run on a schedule or before a release, to catch a slowdown between releases. Full runs happen before launches. Checks of the correctness of features belong to regression testing.

Where AI helps with load scripts, and where engineers decide

Agents help with the routine part of the scripts: drafting a first version from an agreed journey, and updating a script when an endpoint changes. Engineers review each change before it is merged. Targets, the choice of what to load and the reading of results stay with people. An agent does not know what a slow page costs you.

When a load test is the wrong first step

A load test is a poor way to find functional faults: under load they only show up as errors in a graph. If the application does not yet work correctly for one user, fix that first. Without an environment to test in, the numbers would describe nothing you run. And if nobody can say what traffic to expect, the first job is to estimate it, which the plan week can help with. If you are unsure where the problem lies, start with a QA audit. Looking for weaknesses rather than slowness is security testing.

From plan to a baseline you can rerun

  1. Performance test plan: about one week. Targets, what to load, the environment and a quote for the runs.
  2. Test runs: two to six weeks. Scripts, runs, the report and a debrief.
  3. Ongoing: baseline reruns before releases and after fixes, through QA outsourcing, a dedicated QA team or engineers who join your team.

Bought this way, performance testing as a service is a plan, runs and a baseline, not a tool subscription. Whether a next step follows is your call each time. Prices are not published; send what is launching and the quote comes within one business day. Functional test types are listed under test automation services.

How an engagement runs

Start small, with a fixed scope

Each step has a defined scope, duration and deliverables. You decide after each one whether to continue.

  1. Step 1 About one week

    Performance test plan

    We agree the targets with you, then write down the journeys and endpoints to load, the test environment and test data, what is out of scope, and which tool is used and why. Your existing tool comes first.

    You get

    • Targets agreed, such as response times, throughput, error rates and the traffic level
    • The journeys and endpoints to load, and what is out of scope
    • The environment and test data, and how they differ from production
    • A written plan and a quote for the test runs
  2. Step 2 Two to six weeks

    Test runs

    Scripts are written and reviewed, and the load, stress or soak runs in the plan are made. Routine script work is drafted with AI agents where that helps, and merged only after an engineer's review.

    You get

    • Scripts in your repository
    • A report with results against each target, the slowest steps with evidence, and a rerunnable baseline
    • A debrief session for your engineers
  3. Step 3 Monthly

    Ongoing

    The baseline is rerun before releases or in CI, and again after each fix, so you can see whether the fix moved the numbers. This part can be bought as QA outsourcing, a dedicated QA team or engineers placed in your team.

    You get

    • Baseline reruns before releases or in CI
    • A rerun after each fix, compared with the last one
    • A monthly report on the numbers against the targets

FAQ

Questions buyers ask

What do performance testing services include?

A plan, scripts, test runs, a report and a baseline. The plan sets the targets and chooses which test types the engagement runs, for example load, stress, spike, soak or capacity. The scripts live in your repository, and the report shows results against each target and what was not tested.

What is the difference between load, stress and soak testing?

A load test checks the system at the traffic you expect. A stress test goes above the expected load, to see how the system behaves at its limits. A soak test holds a steady load for a long time, to find slow degradation. A spike test applies a sudden, short and large increase. A capacity check, also called a breakpoint test, raises the load step by step to find where the targets stop holding. There is no agreed standard for these names, so the plan states what each run means.

When should we run performance tests?

Before a launch, before a sales peak and after an architecture change. Between those, short checks against the baseline can run before each release or in your CI, so a slowdown between releases, a performance regression, fails a check instead of reaching users.

Why is performance testing important?

A system that is quick with a few users can be slow at peak traffic, and the busiest day is when slowness costs most. A test shows the load at which response times, errors or throughput stop meeting your targets, while there is still time to act. The longer argument is in our post on the benefits of automated performance testing.

Can stress tests and load tests be automated?

Yes. They are scripts, so they can be rerun on demand or in CI. What stays with people is choosing the targets, deciding what a realistic load looks like, and reading the results. A script cannot tell you what a slow checkout costs your business.

Which numbers do you agree before testing?

Response times, throughput, the error rate, and the load level the test must reach. Response times are often stated as a percentile, for example the time within which 95 of every 100 requests finish, because an average can hide a slow minority. The numbers are written into the plan before the first run, so the report can say pass or fail against each one.

Do you test in production?

Not without your written agreement. The plan names the test environment, and third-party services are not loaded without their owners' agreement. Agents and scripts run against the environments in the plan.

Which tools do you use?

The tool is chosen in the plan, with the reason written down, and your existing tool comes first. Whatever is chosen, the scripts are in your repository so your team can rerun them.

Do you fix the bottlenecks you find?

The report names the slowest steps with the evidence from the runs, such as the requests whose response times grew with the load. Fixes usually sit in the code, the database or the infrastructure. Who makes them is agreed in the quote, and the baseline is rerun after each fix.

How much does performance testing cost?

Prices are not published. The quote depends on the number of journeys, the environment, the test types and the reruns. Describe what is launching and you get a scoped quote within one business day.

Start with a one-week performance test plan

Tell us what is launching or growing, which journeys and services carry the traffic, and what you measure today, if anything. We reply within one business day with scope, timeline and a quote.

  • 42 completed test automation and infrastructure projects
  • 32 test automation engineers and experts
  • Founded by the founders of CodeceptJS and Codeception, open-source testing frameworks
  • Founded by the founders of Testomat.io and Testomat.ai
  • On the market since 2015
  • Rated 4.9 out of 5 on Clutch, from 7 client reviews

Prefer to talk first? Schedule a call

    We reply within one business day.