Performance testing training: k6, JMeter, Locust, and what US data says

An app that works for one user is not necessarily an app that survives Black Friday, a tax deadline or a product launch. Performance testing is how you find out before your customers do. This article explains what a performance tester does, how k6, JMeter and Locust differ, what US sources say about speed and the job market, and how our "Perf QA" specialisation is set up.

Many testers learn functional automation first: click, fill a form, check a message. That is useful, but it does not answer another very concrete question: what happens when two thousand people do the same thing at the same time? That is the ground of performance testing, and it is a specialty of its own. For the wider picture of what employers ask for, see our QA Skills Barometer and our guide to test automation in CI/CD.

What a performance tester actually does

The job is not just "send a lot of requests". It revolves around a few plain questions:

• How many users can the application handle before response times become unacceptable?
• What happens beyond that point? Does it slow down gently or collapse? Does it recover on its own?
• Where is the bottleneck: the database, an external service, a missing cache, the server configuration?
• Does it get worse with each release? A performance test run on every delivery catches a regression before users do.

People usually separate load tests (expected traffic), stress tests (beyond the limits), soak tests (hours of steady load) and spike tests (a sudden surge). Each one answers a different fear on the team.

Why speed is a business topic in the US

Slowness costs money, and it has been measured. Two Google-published sources are worth knowing, with their limits:

• In its 2017 benchmarks article, Think with Google reports that 53% of mobile visits are abandoned if a page takes longer than 3 seconds to load. The underlying data dates from 2016 and comes from sites that opted in to sharing their analytics, so it describes an association, not a controlled experiment.
• In "Milliseconds Make Millions" (page last updated 24 June 2020), Google, 55 and Deloitte studied 37 brand sites in Europe and the US and more than 30 million sessions, measured in late 2019. For retail, an improvement of 0.1 second in mobile load times went with +8.4% conversion rate and +9.2% spend.

Neither figure promises the same result for your company. They explain why product teams care, and why a bank, a ticketing site or a retailer wants someone who can prove the system holds up before the peak.

What the US job data says, and what it does not

Let us be precise. The US Bureau of Labor Statistics groups software developers, quality assurance analysts and testers in one occupation. Its Occupational Outlook Handbook (last modified 27 August 2026) projects 10% employment growth from 2025 to 2035, "much faster than the average", with about 106,100 openings a year on average, and a median pay of $134,040 in May 2025. These numbers cover the whole group, not performance testers specifically, and most openings come from people changing occupations or leaving the workforce.

We do not have a reliable US figure for performance testing alone, so we will not invent one. Our own snapshot of 405 French "QA" job ads (2 October 2026) is a useful hint about rarity: JMeter appears in 1.7% of them, k6 and Gatling in 1% each, Locust was not measured. Performance skills are rarely listed in generalist QA ads, which makes them a way to stand out, though that finding comes from France and not the US.

k6, JMeter, Locust: three tools, three approaches

These are the three tools in our specialisation. They do not compete so much as complement each other.

• JMeter: the long-standing tool, with a graphical interface, widely used in large organisations. Handy for building a scenario without code, heavier to version and to run automatically.
• k6: scenarios are written in JavaScript, in a text file you version like the rest of your code. Easy to run from a CI pipeline.
• Locust: scenarios are written in Python, which suits teams that already use that language for tests or data.

Learning all three has a simple payoff: you join a client that has used JMeter for ten years, or a team that prefers k6, and you are not stuck. What matters most is the method (which scenario, which load, which acceptable threshold, how to read the results), and it is the same from one tool to the next.

The Perf QA specialisation at ADC

Here is what the specialisation page says:

• title: "Performance Tester" (Perf QA), to load-test, measure resilience and optimise throughput;
• tools: k6, JMeter, Locust;
• 399 hours over 3 months, 100% online;
• next session: October;
• mentoring included, 6-month platform access and 24-month replays.

Exact dates, prerequisites and any certifications should be confirmed with an ADC advisor before you enrol; the page says so, and we will not promise more here. One practical point: ADC is a French-language training organisation, and our courses are taught in French.

Launch price

For this specialisation, the launch price for individuals is €1,500 incl. VAT, payable in 1 to 4 interest-free instalments. For companies, the price is €7,500 incl. VAT. ADC is not eligible for French public funding such as the CPF at the moment, and we do not want you to think otherwise.

Who it is for

It suits someone who already knows the basics of testing and wants to add a rare skill: a functional tester, a developer who wants to understand load better, a DevOps profile. If you are starting from zero, our main QA automation and AI program is the more logical first step. Either way, talk through your situation with an advisor before choosing.

See the Perf QA specialisation and its price   Talk to an advisor

Frequently asked questions

Written with AI assistance. Course details come from the specialisation page; job-ad percentages come from the ADC barometer of 2 October 2026; market figures are quoted from the sources above.