Why you shouldn't automate tests for new features
Delivering a new feature with 100% automated tests seems like a good idea — but reality is different.
The problem
New features are a sea of uncertainties.
→ Even with good Discovery, validation with beta users and interviews, → When the feature goes to production... often it's not what the user really needed
What happens?
→ You need to adjust the logic and interface based on real usage → And now you have two problems: refactor the feature and rewrite all automated tests
Result: Lead time blown for a delivery that hasn't even generated real value yet.
Not every company is at the same stage
→ In startups, the product changes all the time — hypotheses, not truths → In established mission-critical software, the focus is ensuring nothing breaks
Automating a still unstable feature is a waste of time — and money.
The product's maturity level defines how much you should invest in E2E tests from the start.
Automated tests have a purpose
They serve to ensure the application's current behavior remains reliable over time.
That is, it only makes sense to automate behaviors that actually exist and have been validated in production.
Automating something that hasn't been really used is crystallizing a hypothesis.
What's the best strategy then?
→ Manual and exploratory tests — yes, that's right → Done by people outside the development team, to validate UX and robustness → Use rituals like bug bash or internal bug bounty: – They integrate the team – They spread knowledge about the new delivery
How we do it at Voidr
At Voidr, we continuously monitor changes in our partners' environments.
→ When the feature is already validated and stable, we proceed with E2E test automation. → When there's still uncertainty or technical risk, we opt for an exploratory approach, focusing on breaking the delivery before the end user finds the problem.
Are you automating hypotheses — or validated behaviors?
The difference completely changes the ROI of your quality strategy.
Victor is CTO & Co-founder at Voidr, where he leads quality and test automation initiatives for mission-critical systems.
Follow on LinkedIn