Nielsen's Heuristics

The surest way to check usability — how easily and accurately somebody can accomplish what they came for — is to bring people in and give them a task. Recruiting five of them, aligning schedules and preparing incentives takes a week or two at the very least.

And half of what comes out of a session like that could have been found without anybody being recruited. When Pagedeck ran its first test, three of the eight findings said the same thing: pressing save produces no indication. Ten hours of participant time went to a problem anyone who knows the screen would find in one pass.

Nielsen's heuristics are the net you put in front of that. Ten principles sit in a list and a reviewer sweeps the screens against them. Does the screen say what's happening right now, is it written in the words users use, is there a way out when somebody lands somewhere by mistake.

It takes half a day to a day, in five steps.

1. Set the scope. Pick one flow rather than the whole product. Four screens, from creating a document to sharing it, is about one sitting's worth.
2. Gather three to five reviewers. One person finds roughly a third of the problems, so working alone misses half of them and four people surface most of what's visible. Don't fill the seats entirely with the people who built the screens.
3. Have everybody sweep separately. Sitting together tilts judgment toward whoever speaks first. Each person goes through twice: once to understand the flow, then again against the list of principles.
4. Record three things per violation — which screen, which principle it breaks, and how severe it is. Severity gets a number from 0 to 4. Without that number, twenty findings pile up at equal weight and nothing settles what to fix first.
5. Merge and collapse duplicates. Anything three or more people flagged independently is what gets fixed first.

What the method can't see is just as clear. It finds principle violations, but it says nothing about whether users want to do the task in the first place. So it runs first as a filter, and a usability test with real tasks comes after. The whole activity of experts reviewing against principles is called heuristic evaluation.

"Let's sweep it with the heuristics before we recruit. We can clear the obvious things first."

Putting the cheap inspection in front of the expensive one.

"That's a violation of the first principle. The screen never says whether it saved."

Attaching a principle number to a finding so it carries evidence.

"Twenty findings with no severity attached means nothing gets fixed."

The value needed to set an order is missing.

At Pagedeck, Noa and Theo sweep separately and Mia merges the results. Only the items all three flagged in the same place get fixed that week; the rest stay on the list.