Outsource QA or automate it? What each costs a small SaaS team
If you outsource QA for one regression suite of 350 test cases, it costs about €5,100 to €9,300 at an offshore Playwright agency and about €14,350 to €19,250 for crowd testing. A hire costs an estimated €10,445, and agent skills your own team runs an estimated €2,155. Outsourcing still wins when nobody on your team has the hours to rule on defects.
QA (quality assurance, the testing of software before release) is easy to price badly, because quotes arrive in different shapes. One vendor sends an hourly rate, another a monthly fee. So we priced one unit of work: a regression suite of 350 test cases for a UI-heavy web app at a small software-as-a-service (SaaS) company. Our guide to automated regression testing for web apps explains what such a suite does once it exists.
We sell one of the six routes priced here, so read our numbers with that in mind. We ran the measurement on a real app and show its method and limits below.
How much does QA outsourcing cost?
QA outsourcing costs depend on the route, so we priced one suite of 350 test cases per route. An offshore Playwright agency costs about €5,100 to €9,300 and a German QA hire an estimated €10,445. Crowd testing runs about €14,350 to €19,250, as of October 2026. Managed QA as a service is billed per year, at about $30,000 to $83,100.
What are typical QA outsourcing rates per hour by region?
As of October 2026, hourly rates run from $20 to $50 in Eastern Europe and from $50 to $150 in the US, the UK and Australia. That is one vendor table, and it gives no basis for its rates. Another vendor's blog gives $30 to $55 an hour for Eastern European Playwright work.
Hours are the other half of the sum. Our model needs about 191 hours for 350 test cases; the basis sits under the hire below. At $20 to $50 an hour that is about $3,820 to $9,550. At $50 to $150 an hour it is about $9,550 to $28,650.
These are price lists as of October 2026. We saw no quote while writing this post.
How are QA as a service and managed services contracts priced, monthly or annually?
Managed contracts price the result and hide the hours. As of October 2026, one vendor publishes a flat $2,500 a month, which is $30,000 a year. Another quotes per customer, and its purchase median is about $83,100 a year.
The cost table below names both sources and the dates. When you compare QA outsourcing services, ask for the price per suite and the price per year. Then ask what happens to the tests when you stop paying.
How much do six ways to get regression tests written cost?
Six routes cost from an estimated €2,155 to about €19,250 per 350 test cases, plus $30,000 or more a year for managed QA. Each figure comes from a model or a vendor's public page, and we received no quote. Third-party prices are as of October 2026, with the source in the cell.
| Route | Cost per 350 test cases or per year | Time to first suite | Who owns the tests | What you still do yourself |
|---|---|---|---|---|
| Hire a QA engineer | An estimated €10,445 (range €8,090 to €13,491), from the Kimai report model | An estimated 191 hours of work, 24 working days | You own them | Manage the engineer and review results |
| Outsourcing agency | About €5,100 to €9,300: DeviQA blog figure, not a quote, as of October 2026. TestFort rates: $50 to $150 an hour (US, UK, Australia), $20 to $50 an hour (Eastern Europe), an uncited vendor table, checked 11 October 2026, as of October 2026 | About 191 hours of agency work; elapsed time is not published | Set by the contract | Prepare accounts and staging, rule on defects |
| Crowd testing | About €14,350 to €19,250: RapidUsertests pricing, €41 to €55 per tester for a 20-minute usability test, an assumption and not a quote for regression testing, as of October 2026 | Not published | No automated tests; you get test reports | Write the test cases and pay again every release |
| QA as a service or managed QA | About $30,000 to $83,100 a year. Bug0 Managed: $2,500 a month flat, Bug0 pricing page, read 11 October 2026, as of October 2026. QA Wolf managed tier: quoted per customer, Vendr purchase median about $83,100 a year, checked 11 October 2026, Vendr's page needs a browser, as of October 2026. QA Wolf Platform: 1¢ per AI credit and 15¢ per runner minute, as of October 2026 | Not published | Often the vendor's workspace; ask before you sign | Rule on defects, manage the vendor |
| Per-seat tools | $2,199 per seat per year and up. Katalon Professional, list price, Katalon pricing page, as of October 2026. mabl: no published price, Vendr purchase median about $36,000 a year, checked 11 October 2026, Vendr's page needs a browser, as of October 2026 | Not measured | You own the project in the tool | Write and maintain every test |
| Agent skills your team runs | An estimated €2,155, from a run with qa-signoff Pro features (parallel writing); the €1,500 Pro licence is not counted in it. 350 test cases, 43 hours | About 43 hours elapsed | You own plain Playwright code in your repository | Review results and unblock test cases |
Read how we got these numbers for the method, the sources of every rate and the limits.
What does it cost to hire a QA engineer?
A hire costs an estimated €10,445 for 350 test cases, at €54.71 an hour in our model. The rate comes from Ravio, Compensation explorer, Germany, P5 Specialist, median €99,800, checked 5 October 2026; the page needs a login. The percentiles are 10th €77,300, 50th €99,800 and 90th €128,900. We divide the median by a staff-year of 1,824 hours.
Hours come from Bons et al. (Information and Software Technology 158, 2023). They measured 6 hours to create 11 replayable scripts in an industrial case, which is 0.55 hours per script and about 191 hours for 350 test cases. The study used Selenium IDE, not Playwright, and did not measure review or maintenance.
So the suite costs an estimated €10,445, with a range of €8,090 to €13,491. Treat it as a floor: the salary excludes employer social contributions.
What does an outsourcing agency (offshore or nearshore) cost?
An agency costs about €5,100 to €9,300 for 350 test cases, as of October 2026. It sells hours at a lower rate than a hire. At $30 to $55 an hour (as of October 2026) for about 191 hours, an Eastern European Playwright agency lands at about €5,100 to €9,300. That is a vendor blog figure and not a quote.
Outsourced QA testing services differ by contract. A typical engagement covers:
- Test planning: the agency decides what to test and in which order.
- Writing tests: the agency turns test cases into automated checks.
- Regression runs: the agency runs the suite before each release.
- Bug reports: the agency reports what fails and why.
Your team keeps the access, the staging site, the review of results and the decision on every defect. You also pay for the agency's learning curve on your app, and you pay again each time the app changes.
What does crowd testing cost?
Crowd testing costs about €14,350 to €19,250 for 350 test cases, as of October 2026. It sells human testers by the slot. One slot per test case gives about €14,350 to €19,250 for 350 test cases, at €41 to €55 for a 20-minute test, as of October 2026. The provider prices usability tests, so the figure is an assumption for regression work.
The bigger problem is what you get: test reports and no automated tests, so the next release costs the same again. Crowd testing earns its place for first impressions of a new flow on real devices. For a suite that runs every release, it is the wrong tool.
What does QA as a service or managed QA cost?
Managed QA bundles building and running tests into a fee, about $30,000 to $83,100 a year in the table. The spread is wide because one price is a flat public tier and the other is a purchase median for a quoted tier.
Ask who owns the tests at the end of the contract.
What do per-seat test automation tools cost?
Per-seat tools cost $2,199 per seat per year and up, so five seats run $10,995 a year (list price, as of October 2026). They charge each user a licence and leave the test writing to your team.
These tools sell the software and leave the tests to you. Your people still write and maintain every test, so the hours of the hiring row come back on top of the licence. Seat prices also grow with the team, and the bill returns every year.
What do agent skills your own team runs cost?
Agent skills your team runs cost an estimated €2,155 and about 43 hours elapsed for the Kimai backlog in the measured run. The figure comes from a run with Pro features (parallel writing) and excludes the licence. AI coding agents write the Playwright tests and your team reviews them.
The licence is the second cost. qa-signoff Pro, the skills package behind our measurement, costs €1,500 one-time for 5 seats, as of October 2026, with 12 months of updates and a 15-day money-back guarantee. See the qa-signoff Pro pricing. Our guide to QA automation with AI agents shows how the pipeline works.
The tests are plain Playwright files in your repository, so they keep running if you stop using our skills.
What is QA as a service?
QA as a service is a subscription in which a vendor plans, builds, runs and maintains your tests for a monthly or annual fee. You pay for the outcome, a tested release, and the vendor decides how many hours it takes.
It differs from an agency, which sells hours for a project and hands the work back at the end. It differs from a tool, which sells software and leaves the work to you.
The cost of handing QA to a vendor is dependence. The vendor's staff learn your app, and their knowledge leaves with the contract. A short exit clause, with the tests handed over as code, keeps your tests yours.
What did the agent route cost on a real app?
In the Kimai model, built on a measured run of 66 test cases, the agent route costs about €2,155 and about 43 hours elapsed against an estimated €10,445 for a hire. That is about 4.4 times faster and about 4.8 times cheaper than the human QA engineer estimate.
We ran the agent route on Kimai 2.67.0, an open-source time-tracking app, with Playwright 1.49.0 and Chromium 131.0.6778.33.
The backlog held about 350 test cases. A fleet of 6 Claude Code instances running Sonnet 5.5 processed about 66 of them in the measured run, in about 1 hour 48 minutes. A tool we built, Kaizero, orchestrated the fleet. For the stages, read how AI agents wrote the 350 test cases.
In the measured run about 55 test cases landed as passing Playwright tests, and 11 blocked, about 17%. Blocked means the agents stopped and a person had to decide. For every figure, read the measured run on Kimai, with method and limits.
The estimate for all 350 comes from resampling those 66. The measured values are the 66 test cases, the 55 that landed, the 1 hour 48 minutes and the 20 minutes of prompting and review. The rest is a model built on them.
The cost split surprised us. In the model the agent route costs about €2,155: about €396 of AI usage fees, about €18 of prompting and about €1,741 of unblocking. Human time is the bigger line even when agents write the tests. The saving over a hire is about 148 hours and about €8,290.
A blocked case looks like this. TC-WEB-001 and TC-WEB-002 check that after logout the browser's Back button does not show the dashboard again. In Kimai it does, because the dashboard response carries no no-store header, which tells browsers not to keep a copy of the page. The agents stopped.
They marked both cases blocked, because the fault sat in Kimai and the tests were correct. A person decides whether to fix the app or change the test cases.
We didn't record how long that took. So we use the model: about 0.55 hours per blocked case, and about 31.8 hours for about 58 blocked cases, which is about 17% of 350.
Our limits, in plain words:
- The measurement covers one app.
- The 66 measured test cases leaned toward top priority.
- The human column is a model estimate: no human wrote tests for this backlog.
- 55 of 350 test cases landed, so the pass count covers 55 of 350.
The saving is therefore an estimate. We still use it, because the report states how it was built.
What costs does an outsource QA quote not show?
Four costs sit outside the price a vendor states. We saw no agency quote while writing this post, so we name these costs and leave them unpriced.
- Onboarding and knowledge transfer. Someone on your side explains the app and the test data. That someone may be the person you hoped to free up.
- Vendor management. A contract needs an owner who reads the weekly report and says yes or no.
- Test ownership on exit. Portable Playwright code in your repository leaves with you.
- Maintenance after the app changes. Every redesign breaks tests. Ask who fixes them and at what price.
A fifth cost applies to every route, and small teams forget it most: the person on your side who rules on defects. The Kimai model puts it at about €1,741 of the total of about €2,155. The €1,741 is a lower bound, because the salary rate excludes employer social contributions.
The AI usage fees were about €396, about 18% of the total. A team that worries about the AI bill and ignores that person is watching the smaller line.
How do you pick a partner when you outsource QA?
Pick the partner by what the quote leaves out. Four questions separate a price you can plan on from a price that grows, whichever of the QA outsourcing companies you shortlist. Ask them in writing before you sign. Each answer names a cost the quote may hide, and a vendor who will not answer has told you something too.
- Ask what the quote covers and how it is priced: hourly, monthly flat or annual contract. A rate without a stated basis hides the scope it leaves out.
- Ask who owns the tests when the contract ends: portable Playwright code in your repository or tests inside the vendor's workspace. Tests inside the vendor's workspace stay behind when you leave.
- Ask for the onboarding and knowledge-transfer plan and the management time it takes from your side. A quote that omits both leaves that time unpriced.
- Name the person on your team who rules on defects and unmet criteria, whichever partner writes the tests. Without that person, blocked test cases wait.
When should you outsource QA instead of automating?
Outsource QA when one of these describes your team:
- Nobody can review tests. A developer or product owner can own the suite, but only with the hours, and we estimate 2 to 4 hours per monthly release. Without them, agent skills and per-seat tools stall, because someone must rule on defects.
- You need independent eyes: a one-off audit before a launch, or a security obligation that requires an independent tester.
- You face compliance testing that an outside party must sign.
- You need a device lab with many real phones and browsers.
- You need exploratory testing, where a skilled person looks for what nobody wrote down.
An agency wins in these cases, and you pay for it in its rate.
Mixed plans work. A team can hire an agency for the launch audit and run agent skills for the suite that follows. The audit is a one-off, and the suite stays in your repository, where automated regression testing keeps it useful.
How do you decide between outsourcing, hiring and automating, as a small SaaS team?
Decide with your own numbers, in six steps. The costs above are a starting point, because they come from one app and from public price pages. A decision needs your count of test cases and a named person who reviews results.
- Write down the unit: how many test cases your app needs, for which roles and devices. A quote without a unit can't be compared with another quote.
- Ask three agencies for a price per suite and a price per year, in writing. An hourly rate alone hides the scope.
- Ask each vendor who owns the tests after the contract ends. Tests inside a vendor workspace stay behind when you leave.
- Name the person on your team who reviews results, and give them hours. Without that person every blocked test case waits, and you'll pay for the wait.
- Price the agent route with your own numbers, using the Kimai model as a start. A model from one app can be off for yours.
- Pick the cheapest route that leaves a named owner for every blocked test case. The cheapest route without an owner stalls.
If step 4 has no name, outsource. If it has one, read the measured Kimai run and build a regression suite that runs on every release.
What else do people ask about outsourcing QA?
Can AI agents replace an outsourcing agency for regression tests?
Agents replace the test writing of an agency and leave the defect rulings to a person. In the measured run, agents landed 55 of the 66 test cases they processed, and for the other 11, about 17%, a person had to decide what the app should do. The review needs a developer or product owner with time. Our guide to QA automation with AI agents covers the general question whether AI replaces QA testers.
Who owns the tests when an outsourcing contract ends?
Playwright code in your repository stays yours; tests in a vendor workspace may not. Tests built in a vendor's workspace can stay behind when you leave, and then you write them again. Put the handover of the test code, in a format your team can run, into the contract, and ask before you sign.
How long does an outsourcing agency take to deliver a first regression suite?
Expect about 191 hours of agency work for 350 test cases, from the Bons rate, because agencies don't publish the elapsed time. At one full-time person that is about 24 working days, the same as a hire. The agent route took about 43 hours elapsed in the measured run. Ask the agency for a delivery date per batch of test cases, in writing, and ask what it needs from you before the first test runs.
Run the agents on your own app: qa-signoff Pro is €1,500 one-time for 5 seats, with a 15-day money-back guarantee.
See Pro prices