At the idea stage
Once over. It is enough to correct the wording in the specification.
Five directions, each with its own question: does it do what it should, will it stand the load, can a person work it out and are there any holes.
We repeat the customer's actions by hand: placing an order, paying, returning, editing their own details. Edge cases, where things most often break, form a separate block.
We simulate a surge of visitors and find the point where the system stops coping. Learning your limit in advance costs less than discovering it on sale day.
The key journeys check themselves before every release. Both the repetitive manual work and the fear of shipping disappear.
We watch real people on real phones: where they get stuck, what they cannot find and why an order goes unplaced.
We find the holes before release: injection into database queries, mistakes in permission handling, excess information in server responses and logs, weaknesses in the sign-in mechanism.
The cost of the same defect depends on the stage at which it was found. That is the main argument for regular testing rather than a single run before handover.
Once over. It is enough to correct the wording in the specification.
Three times. The developer changes the lines before they reach a build.
Ten times. The work goes back to the developer and the whole cycle repeats.
Thirty times. An unplanned release with an emergency patch is required.
One hundred times. On top of the cost of the work come lost revenue and a dented reputation.
Investment in automation pays off not at once but after dozens of runs. That is why the machine is given the settled journeys that get walked at every release. Automating something rewritten weekly means spending the budget on maintaining the tests themselves.
The output is not a complaint about quality in general but reproducible findings with priorities, which a developer can start on immediately.
We study the user journeys, who the users are and what matters to the business. Giving every area equal attention makes no sense.
We fix the list of checks and the list of devices and browsers. The document is agreed with you and sets both the effort and the price.
Every finding comes with the steps to reproduce, a screenshot and a severity rating. Otherwise the report is worthless.
When a fix arrives we look not only at it but at the neighbouring areas: whether anything broke along the way.
They do test, but they confirm that what they built behaves as intended. A tester's job is the opposite: to find out what happens when you depart from the intention. The roles are fundamentally different and sit badly in one head - the eye stops noticing the obvious.
Yes, people come to us for this regularly just before accepting work from a contractor. We have no reason to hide findings, so the report is impartial. It makes more sense to run such a check before signing off rather than afterwards.
We start from your own analytics rather than a universal list. Usually two or three phone sizes and three browsers are enough - they account for the vast majority of visits. Testing everything is expensive and does not pay for itself.
Three numbers: how many simultaneous visitors it handles comfortably, at what point the slowdown begins and which component is the bottleneck - CPU, storage, the database or the code itself. With those you can choose between adding resources and rewriting.
Напишите, что подлежит проверке и к какой дате. Составим программу работ и рассчитаем стоимость по числу проверяемых путей.
Request received
It is already with a manager. You will get an answer within the working day, and urgent requests go to the duty engineer immediately.
There is no such city in the list. Check the spelling.