Auto tests. But whether to outsource?

Hello, dear habrosociety.

I want to share some thoughts, as well as hear your opinion regarding outsourcing web application autotesting. A small vote is also proposed on the topic.

I must make a reservation right away that this is not an attempt at subtle PR of my company or something, but just a lively interest in the issue.

I think many of those who in the subject have noticed that recently there has been some excitement around automated testing. Numerous trainings, seminars, reports, tutorials on how to write, run tests correctly, reviews of tools, frameworks, design approaches, etc. Most companies involved in web development want to use auto-testing on their projects and get benefits from it in the form of a constant controlled level of quality and economy, due to the automation of repeatedly repeated checks. But often it turns out the other way around, when resources are spent on maintaining the tests themselves in working condition, and a stable result does not work out. I think everyone has come across the phenomenon that tests fall for various reasons (often due to curvature) just then, when you need to decide on the possibility of deploying the next release to production. There are many different reasons for this, but we will not dwell on this now. We just admit that there are problems and move on.

And here's the thought itself - Or it might be a better idea to outsource this matter to professionals who have this process debugged to the ideal state, integrate with their CI process and get the finished result in the form of reports without worrying about infrastructure and training people and other difficulties and leave the domain QA team, domain testing, manual testing and writing test cases, saving on the salary fund, since the salaries of testers-automation engineers have already been equal to the salaries of developers, which not logical. In addition, there is a shortage of professionals in the market.

Of course, with outsourcing auto-testing I also think there are problems. I have not encountered, I can only speculate. Share if anyone had experience.

I tried on the Internet to find companies that offer such services. My list was small and all the companies I found offer comprehensive testing services with opaque pricing a la call and we will calculate how much it can cost you. The first thought that comes to mind is that in financial terms this may turn out to be even less profitable than having your own team of auto-testers. But what if there was another model, say, pay per action, in which every action in the browser would be paid: click on an element, select an element from the list, fill in the field, etc. when writing code + monthly rent for infrastructure rental to run tests depending on the required capacities. It will be quite enough for someone to run tests in 1 thread, someone in 10, someone in 100.

With this approach, it would be accurate to the ruble to calculate what is more profitable to test manually, and what would be cheaper to automate.

I invite everyone to share their experiences and opinions in the comments, and only those who are directly involved in making such decisions are invited to the survey, in order to give it representativeness.