Quick answer
A user interface (UI) is the layer where a person operates a product: the layout, controls, wording and feedback they see and use. It is not the whole experience (UX) and cannot be judged by looks. Judge it by behavior: click rate among people who saw it, follow-through, and repeated rapid clicks.
Formula: Element click rate = sessions with a click on the element ÷ sessions that reached it
What is a user interface?
A user interface (UI) is the point where a person and a product meet: everything the person can see, touch, say or type to operate it, and everything the product shows back. On a website it is the layout, menus, buttons, form fields, words and the small reactions (a button that changes state, an error message) you meet while using the page. Wikipedia describes it as the space where interactions between humans and machines occur, and the same word covers a phone screen, a car dashboard, a terminal window and a smart speaker.
Two points keep the idea practical. First, the interface is not the whole product: the same search box can sit on top of a great or an empty catalogue. Second, how good an interface is cannot be read off how it looks. It shows up in what people do with it: what they click, where they stop scrolling, where they give up. The sections below define the types and parts of an interface, then show three rules for judging an element from behavior data.
Types of user interface
| Type | You operate it by | Everyday example | Strength | Weakness |
|---|---|---|---|---|
| Graphical (GUI) | Pointing, clicking, tapping on windows, icons and menus | A website, a mobile app, a desktop program | Options are visible; low learning cost | Takes screen space; repetitive tasks need many clicks |
| Command line (CLI) | Typing commands | A terminal, PowerShell | Fast and scriptable once you know the commands | Nothing tells you what you can type |
| Voice (VUI) | Speaking | A smart speaker, a phone assistant | Hands-free | Hard to browse many options; recognition errors |
| Touch and gesture | Swiping, pinching, shaking, moving | A phone, a tablet, a headset | Direct and natural | Gestures are invisible until learned |
| Natural language | Writing a sentence in your own words | A chat assistant | No menu to learn | You may not know what it can do |
Most products mix types. A web app is a GUI, but its search box behaves like a small command line, and its support chat is a natural-language interface. For web analytics the GUI matters most, because clicks, taps and scrolls leave a record that the other types do not.
UI vs UX
| User interface (UI) | User experience (UX) | |
|---|---|---|
| Scope | The surface people operate | The whole relationship with the product and the company |
| Question | What does it look like, and how is it arranged? | Does the whole journey meet the person's need, without hassle? |
| Includes | Layout, controls, wording, feedback | The interface, plus content, speed, support, price, delivery |
| Example failure | A button too small to tap | A well-designed search over a catalogue missing what you want |
| Typical evidence | Clicks, scroll reach, rage clicks on one element | Task completion, return visits, satisfaction |
Nielsen Norman Group makes the same split: the user interface is one part of the user experience, and usability is a quality of the interface (ease of learning, efficiency, pleasantness). A polished UI cannot rescue a missing feature; a rough UI can still deliver a good experience when the product does exactly what the person came for. For the wider set of measures, see UX analytics; this page stays with the interface itself.
The five parts of an interface
Any screen can be read as five parts. The table pairs each with a question and with a behavior signal that shows whether it works.
| Part | What it is | Question to ask | Behavior signal |
|---|---|---|---|
| Layout | What sits where, in what order | Can people find the main action without hunting? | Scroll reach: share of visits that get as far as the element |
| Navigation | Menus, links, breadcrumbs, search | Do people reach the page they want in one step? | Which page people open next; clicks per menu item |
| Controls | Buttons, fields, toggles, checkboxes | Can they operate it the first time? | Click rate among visits that saw it; rage clicks |
| Text | Labels, headings, error messages | Does the wording match what people expect? | Clicks on a link before vs after a wording change |
| Feedback | The product's reaction to an action | Does it say something happened? | Repeated rapid clicks; errors; time to the next step |
The last row connects to the first of Nielsen's ten usability heuristics, visibility of system status: keep users informed about what is going on, with feedback in a reasonable time. A button that does nothing visible for two seconds invites a second and third click, which is exactly the pattern a rage click records.
How to judge an interface element with behavior data
A click count on its own cannot say whether an element is good. Three rules turn it into a judgement.
- Divide by who could see it. Click rate = sessions with a click on the element ÷ sessions that reached it. Clicks ÷ all sessions mixes a visibility problem with a design problem.
- Check what happens after the click. A control that is clicked but not followed by the next step promises something the next screen does not deliver: follow-through = sessions that reach the next step ÷ sessions that clicked.
- Look for repeated rapid clicks. Several clicks within a fraction of a second in the same spot usually mean people expected a response and did not get one. Count them per 100 clicks and see whether a few sessions or many produce them.
Always split by device: desktop and mobile layouts differ, so their numbers are not interchangeable. And settle the rule's thresholds (for example, "investigate a follow-through below 60%") before you look, so that the data does not choose the threshold for you.
Worked example: one button on a pricing page
Sample data: 1,000 sessions on a pricing page with a "Start free" button. The button was clicked 80 times in 64 different sessions. The button sits low on the page, and 500 of the 1,000 sessions scrolled far enough to see it. Of the 64 sessions that clicked, 48 reached the sign-up page. Six of the 80 clicks were rapid repeats, coming from 4 sessions.
| Measure | Calculation | Result |
|---|---|---|
| Clicks per session, naive | 80 ÷ 1,000 | 8.0% |
| Click rate, all sessions | 64 ÷ 1,000 | 6.4% |
| Click rate, sessions that reached the button | 64 ÷ 500 | 12.8% |
| Follow-through to sign-up | 48 ÷ 64 | 75.0% |
| Rapid repeat clicks per 100 clicks | 6 ÷ 80 × 100 | 7.5 |
| Clicking sessions with repeats | 4 ÷ 64 | 6.3% |
Rule 1 changes the picture: the button converts 12.8% of the people who saw it, double the 6.4% you get by dividing by everyone. Half the visits never reach it, so the first fix is its position, not its colour. Rule 2 says three in four clickers carry on, which is healthy; the 16 who do not are worth a few session replays. Rule 3 finds 6 repeat clicks in 4 sessions: small, but check whether the button changes state when pressed.
The same 80 clicks can mean opposite things. If all 1,000 sessions had seen the button, 64 ÷ 1,000 = 6.4% would be the true rate and the button would be half as effective as it is. The click count alone never tells these two cases apart.
Common mistakes with user interfaces
- Treating UI as UX. A redesign of buttons and spacing is a UI change. Whether people finish their task is an experience question and needs outcome measures, not an opinion of the screen.
- Using click counts as the verdict. Counts rise with traffic and with visibility. Use rates over the sessions that saw the element.
- Reading no clicks as no interest. An element below where most people stop scrolling or hidden on phones gets no clicks regardless of its design.
- Mixing devices. Desktop and mobile are different interfaces; merge them and neither is readable.
- Comparing across a redesign without a date. Clicks from the old layout land on positions that now hold other elements. Start the range on the release day.
- Judging by taste. Whether a design looks modern is not evidence. Set the rule first, then read the data.
How to check an interface in TapCub
Here is the pricing-page example worked through in TapCub.
- Switch on the inputs. In Web analytics › Configuration › SDK code, turn on autocapture (Pro plan and above) and switch on Scroll depth, which is off by default. Data builds up from that moment.
- Open the heatmap and pick a device. In All-platform analytics › Behavior › Heatmaps, pick the pricing page and desktop or mobile. Scroll reach gives the share of visits reaching each depth (the 500 of 1,000), and Elements lists the most clicked controls with their share and visitors, which gives the numerator of rule 1.
- Read frustration. The Rage clicks tab and the page list's rage-click count show repeated rapid clicks (the third click within 300 milliseconds and 50 pixels). TapCub does not detect dead clicks yet, so a zero is not proof that nothing looks clickable and does nothing.
- Measure follow-through. In All-platform analytics › Behavior › Behavior analysis, build a funnel from the pricing page path to the sign-up page path. Each step shows its count and the largest drop is marked, which gives rule 2.
- Compare before and after. The heatmap has no compare period and TapCub does not run A/B tests: open the same page for the 30 days before and the 30 days after a change, and note the release in Configuration › Annotations. For single sessions, open session replay (off by default; an admin turns it on).
How to create a heatmap walks through the setup screen by screen.
See also: User Interface Design: 8 Principles and How to Test Each With Data.
See also: What is a modal window in UI? Rules and examples.
Sources
- User interface, Wikipedia (definition, types and history of user interfaces)
- 10 Usability Heuristics for User Interface Design, Nielsen Norman Group, Jakob Nielsen, 1994, reviewed 2024
- The Definition of User Experience (UX), Nielsen Norman Group
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.

Data Engineer
Writes about event design, data quality, and collection workflows, with verifiable implementation methods for reliable analytics.