What the automated test package covers
The browser end-to-end test package includes 242 tests in 31 files and runs a real Chromium browser against a real backend. The tests replay user actions: logging in, the admin panel, work scheduling, work permits, reports, documentation, maps and analytics. This is the same scope a manual acceptance tester goes through after every release.
The test database is rebuilt from migrations once per test run, so the result does not depend on data from the previous run. Such acceptance is repeatable: two runs of the same version give the same result, something that cannot be guaranteed with manual acceptance.
Where did the scenarios come from?
The scenarios of manual testers, taken from session recordings (June to September 2026), were rewritten as automated tests. While porting them, 13 defects in the application were found, so the rewriting also worked as a form of audit.
How long a run takes and when it starts
The full E2E suite takes about 2 to 2.5 hours, so it runs manually on request and automatically at midnight on weekends, without blocking merges. The plant's team therefore does not wait for the result during the working day.
A parallel mode shortens the time almost linearly, but each test instance needs its own database and its own Redis database. This is an infrastructure cost on the vendor's side, which must be taken into account if the client wants a shorter run time.
Hours of acceptance per release
Calculation: how many hours automation replaces
The calculation below is based on assumptions that the plant can replace with its own figures. The time of a manual acceptance of one release depends on the number of paths the plant considers critical, not on the number of tests in the package.
Example: the plant assumed that a full manual acceptance of one release takes 40 hours of work, and after introducing automation people spend 10 hours on the paths not covered by tests and on analysing the result. With 12 releases a year, the saving is 360 hours.
What automation cannot replace
Automation does not replace a usability assessment or the checking of a new feature for which no test exists yet. Nor does it replace a visit to the plant floor and checking of devices such as transmitters or mobile devices. That is why the calculation keeps 10 hours of human acceptance.
What about a bug we already know about?
A known, not yet fixed error is marked in the test as an expected failure with a ticket number, never by skipping the test. Once it is fixed, the test itself reports that the annotation must be removed. The acceptor therefore sees a list of known problems instead of surprises.
Why unit tests are not enough
Running browser end-to-end tests after a fix catches user-visible errors that unit tests miss, because unit tests run against a mocked API. For this reason we count the savings on acceptance that a person previously performed at the screen, not on unit tests.
While porting the scenarios, seven of the 13 defects were fixed within two days, and their tests became ordinary regression tests. Every fixed bug therefore stays under protection, which over time reduces the hours needed for acceptance.
Defects from porting the scenarios
Annual saving in acceptance hours
| Parameter | Value |
|---|---|
| Releases per year | 12 releases |
| Manual acceptance of one release | 40 hours |
| Automated acceptance of one release | 10 hours |
Result: (40 - 10) x 12 = 360 hours per year. Multiplied by the plant's own hourly rate, this gives the annual saving in currency.
Note: Sample calculation based on the assumptions of the page author, not a measurement. The plant substitutes its own number of releases and acceptance hours.
Video
Frequently asked questions
Are 242 tests enough to accept the whole system?
They cover logging in, the admin panel, work scheduling, permits, reports, documentation, maps and analytics. New features and devices on the plant floor require additional checking.
Do the tests block merging of changes?
No. The full suite takes about 2 to 2.5 hours, so it runs on request and at midnight on weekends, without blocking merges.
Does the test run on mock data?
No. It runs a real browser against a real backend, and the database is rebuilt from migrations.
Calculate the savings on your own assumptions
Give us the number of releases and acceptance hours in your plant, and we will calculate how much work automation replaces. Book a conversation with InnerWeb.
Let's talk