Factors of Interface Quality
In short Vlad Golovach reduces interface quality to four criteria: speed of work, number of errors, speed of learning, and subjective satisfaction. He then catalogs the factors behind each. For analysts, the four criteria give a solid structure for usability requirements, and the factors make a checklist for prototype reviews.
Based on: V. Golovach, User Interface Design, v1.2 (in Russian), part I “Factors”.
On this page
Key ideas#
Speed of work#
- Most of the work is mental. Golovach uses Donald Norman’s seven stages of action, from forming a goal to evaluating the result. Six of the seven are mental, so simplifying the user’s thinking matters more than speeding up clicks.
- Direct manipulation. Acting on objects directly, such as dragging a file to the bin, beats issuing commands because it takes fewer actions. The object should look realistic; the action need not.
- Fitts’s law. The time to reach a target grows with distance and shrinks with target size. Two consequences: a target at the screen edge is effectively infinite in that direction, and menus or dialogs that open at the pointer reduce the distance to almost nothing.
- Interruptions. People get distracted. The screen should show where the user stopped, what has been entered and what to do next.
- Long operations. Collect everything you need from the user before a long process starts, and keep progress indicators honest, because users plan their time by them.
Human errors#
Golovach’s position is that there are no human errors, only design errors. He names four types: lack of domain knowledge; slips, when attention drifts; failing to read the system’s state; and motor errors, such as missing a small target. Errors noticed and fixed during an action cost far less than errors found afterwards, and the main tool for that is real-time visual feedback.
He ranks three ways to make a system less sensitive to errors, from weakest to strongest:
- Ask for confirmation before dangerous actions. This works mainly for beginners, since experienced users confirm automatically.
- Check actions before accepting them, for example by offering only valid commands.
- Let the system choose values (last used, most frequent) and ask the user only to verify them. This is the most effective option and the hardest to build.
Learning#
People learn only with an incentive (they must believe the effort will pay off) and only if they know a feature exists. Four things make a system understandable: a clear mental model; metaphor, which is powerful but risky because it constrains functionality; affordance, Norman’s term for an object whose look shows how to use it; and standards, which are strong only while popular. Golovach distinguishes several kinds of help. The most important is domain help, which explains how to do something well. He recommends “spiral” help: one to three sentences first, more detail on request.
Subjective satisfaction#
Golovach cites research showing that attractiveness weighs about as much as efficiency in how people rate an interface. Aim for an unobtrusive, elegant design built on consistent visual rules rather than decoration. Error messages should answer three questions: what is wrong, how to fix it now, and how to avoid it next time. But most error messages signal a design that let the error happen. People value customization; in business tools, offer presets rather than full configuration.
Memory, search and navigation#
Golovach suggests planning for about five items in short-term memory rather than the classic “seven plus or minus two”. Grouping raises the effective capacity. For search, query forms work well for finding specific records, but spotting patterns needs a visualization of all the data. A navigation system should show users where they are, where they can go, where they have been, where it makes sense to go next, and where they came from.
How to apply it#
- Structure usability requirements by the four criteria, with a measurable target wherever it matters.
- Prefer prevention to messages. Require constrained inputs, valid-only options and smart defaults before you require error messages.
- Specify error messages so that each one gives the three answers.
- Require resume points for long or interruptible tasks.
- Check navigation against the five questions.
Review checklist#
- Frequent targets are large and close; menus open near the pointer.
- The screen shows where the user left off.
- All inputs for a long operation are collected before it starts, and progress is shown honestly.
- Invalid choices are not offered.
- Defaults use the last or most frequent values.
- Each error message says what went wrong, how to fix it and how to avoid it.
- No screen asks users to hold more than about five items in mind.
- Navigation answers: where am I, where can I go, where have I been, where next, where from?
Example#
At The Workshop, a fictional chain of bike-repair workshops, the parts-ordering screen produces frequent wrong orders. The analyst classifies the errors. Managers type part numbers from memory (slips). They miss a supplier’s “discontinued” flag (state not read). And they don’t know which parts fit which frame size (domain knowledge). The requirements change accordingly:
- parts are chosen from a list filtered by bike model instead of typed;
- discontinued parts are not offered at all;
- the last ordered quantity is prefilled;
- a short “which size fits?” note appears next to frame-specific parts.
Sources and further reading#
- V. Golovach. User Interface Design, v1.2 (in Russian; freely distributed), part I “Factors”.