How to evaluate the usability of an interface without A/B tests?

Question

Design

There is a web application. It is not possible to conduct A/B tests or field studies (observation, short usability tests). You can request any metrics for actions on pages. How to evaluate the ergonomics of the interface? Are there any universal metrics for evaluating ergonomics?
It's only about usability, without commercial metrics like DAU, Retention and the like.

Answer 1

As far as I understand, we are talking about closed production and operators.
The first conclusion: this is a professional tool and speeches about all sorts of funnels, conversions should not be carried out.
Is it possible to find the original goals with which the application was designed?
They had to make sure they did the right thing. According to the idea, it’s worth taking the metrics from there.
If this is not there at all, it sounds like a discovery phase - you need to re-understand a few things:
1. Who said that the interface is inconvenient? Who said there is a problem? Why should it be decided?
2. Then - for whom to decide? It's good if this operator is alone, but usually there is a supervisor, and above him there is also a manager who distributes the load. This is the creation and work with persons.
3. Screens - they are part of the process, you need to get the whole process. We made a whole AS-IS Workflow diagram where we looked for bottlenecks and incomprehensible white spots. And that's weeks of work.
4. Working conditions - operators can work in biological protection suits with oxygen cylinders and masks, or they can be at the plant in dust and a temperature of +40°C.
Damn, the question is not really average and the answer to it is quite like an article / conference report.
Further, it will look like an audit - you can’t take off without additional context.
And based on the usual metrics, you can always find the other side of the coin:
One time only gives rise to a bunch of questions: is it critical? Is it related to the number of errors? Better longer, but without errors, than faster and with more cognitive load? How complicated is the process and how much is it taught to the operator? What path is initially selected in the system? For example, for an operator of a nuclear power plant, time is critical only in case of accidents, otherwise it is system monitoring.
I think without a mat part and details it will be normal to advise something more difficult.

Answer 2

the main thing is the sales funnel, nothing better has been invented
specifically by you listed gauno, but they usually correlate with the funnel adjusted for the country
, alas, nothing has been invented except for a relative comparison, so you can not call it AB and use

Answer 3

When it is not possible to conduct A / B tests, you can use the same Yandex metrics service and test it on real users, see user paths, see a heat map, and based on this data, roughly understand the holes in the interface.
Another method is to assemble a focus group, now it is not difficult and not expensive to do this.

Answer 4

Read here what happens to people when they change A to B. You may find something suitable for you.