An in-house tester costs far more than their salary. Outsourced testing costs more than its invoice. Comparing the two honestly means counting both properly.
Here is the arithmetic, with the parts people leave out.
What an in-house tester actually costs
Salary is roughly 70–75% of the total. The rest is employer taxes, benefits, equipment, software licences, recruitment, and management time.
Recruitment alone is significant — agency fees typically run 15–25% of first-year salary, and even hiring directly costs weeks of someone's attention.
Then there is ramp-up. A new tester is not productive on day one. Learning a product properly takes one to three months. That is real cost, paid in full salary.
The rule of thumb: total first-year cost is roughly 1.4× the salary, plus a quarter of unproductive ramp-up.
What outsourced testing actually costs
The invoice, plus three things.
Briefing time. Every engagement needs someone to explain the product, provide accounts, and answer questions. Budget two to four hours for a first engagement, less afterwards with the same tester.
Coordination. Someone on your side owns the relationship. Small, but not zero.
Context loss. This is the real hidden cost, and it applies mainly to agencies with rotating staff. If a different person tests each time, you re-explain your product each time and they re-discover the same shallow issues. The deep bugs — the ones only found by someone who knows how your product is meant to work — never get found at all.
A freelancer or a retained tester avoids this. A rotating pool does not. When comparing outsourced options, ask whether you get the same person.
The comparison at realistic volumes
If you need testing on 4 releases a year:
In-house is indefensible. You are paying a full-time salary for occasional work. Buy it in — four fixed-price engagements will cost a fraction of a quarter's salary.
If you need testing on every release, releasing monthly:
Twelve engagements a year. Outsourcing is still usually cheaper, and a monthly retainer gets you continuity. Ours is $1,199 a month for 20 hours, which is well under the loaded cost of a part-time employee.
If you release weekly and have multiple user types:
The calculation flips. Testing is now continuous, coordination overhead is constant, and the product knowledge an in-house person accumulates becomes worth more than the rate difference. Hire.
The costs nobody counts on either side
In-house, forgotten: management time, holiday and sick cover, the risk that your only tester leaves and takes all the product knowledge with them, and the possibility that developers test less carefully because someone else does that now.
Outsourced, forgotten: briefing time, the risk of a bad supplier costing you a launch, and — the big one — context loss if the person changes.
The honest recommendation
Most companies under 30 people should outsource. Testing is not continuous at that size, and a fixed-price engagement per meaningful release covers the actual risk at a fraction of the cost.
Most companies over 50 people releasing weekly should hire. Coordination overhead exceeds the rate difference, and continuous product knowledge starts to compound.
Between those points it depends on release frequency more than headcount. Count releases, not people.
Want to run the numbers on your situation? Tell us your release frequency and team size, and we will tell you which way the maths points — including when it points away from us.