Quick answer
Micro interactions are small, single-task moments in an interface, such as a button's loading state or a toggle sliding on: a trigger, rules, feedback, and loops or modes. Judge one by behavior, not looks: missing feedback shows up as repeated clicks, pauses and abandoned steps.
Formula: Rage clicks per 100 sessions = rage clicks ÷ sessions × 100; duplicate-submission rate = duplicate submissions ÷ all submissions
What are micro interactions?
Micro interactions are the small, single-task moments in an interface: a button that changes as you press it, a switch that slides to "on", a field that turns red when an email address is missing its "@", a spinner that tells you an order is being placed. Each one handles exactly one job and tells the user what just happened.
The term was popularized by designer Dan Saffer in his book Microinteractions: Designing with Details (O'Reilly, 2013). Nielsen Norman Group describes them as trigger-feedback pairs: the trigger is a user action or a change in the system's state, and the feedback is a narrow, targeted response to that trigger, usually small and visual. Both spellings, "micro interactions" and "microinteractions", mean the same thing.
The useful way to judge one is not how it looks in a design file. A micro interaction is working when users do not have to guess. When the feedback is missing or late, people click again, pause, or give up, and those reactions show up in behavior data as rage clicks, duplicate submissions and abandoned steps. This article treats micro interactions as something you can measure, not decoration.
The four parts: trigger, rules, feedback, loops and modes
Saffer's model splits every micro interaction into four parts. Take a "Save" toggle in a settings page as the running example.
| Part | What it is | In the "Save" toggle | What goes wrong when it is weak |
|---|---|---|---|
| Trigger | What starts it: a user action (click, tap, swipe, key) or a system condition (new message, upload finished) | The user taps the switch | The control does not look tappable, so nobody triggers it |
| Rules | What happens after the trigger, and what can and cannot happen | The setting is sent to the server; while saving, a second tap is ignored | Hidden rules surprise users, e.g. a second tap cancels the first |
| Feedback | What the user sees, hears or feels so they learn the rules | The knob slides, the track turns green, a short "Saved" appears | Nothing changes for two seconds, so the user taps again |
| Loops and modes | Meta-rules: how it behaves when repeated or over time, and any temporary special state | If saving fails, the switch slides back and shows "Couldn't save. Retry" | A failed save looks identical to a successful one |
Rules are invisible, so feedback is the only way users learn them. That is why most measurable problems are feedback problems: the rule worked, but nothing told the user it did.
Common microinteraction examples
Most UI micro interactions fall into a handful of types. The last column is the signal that tells you, in data, that the feedback is not doing its job.
| Type | Trigger | Feedback that should appear | Data signal when feedback is missing |
|---|---|---|---|
| Button states (hover, pressed, loading, disabled) | Click or tap | Pressed state immediately; loading state if the action takes longer than about a second | Rage clicks on the button, duplicate submissions |
| Inline form validation | Leaving a field, or typing | Error message next to the field, in words, not only color | Repeated submit clicks, form abandonment at one field |
| Loading and progress indicators | A request starts | Spinner under ~10 seconds; progress bar with time left above that | Page exits during the wait, reloads |
| Toggles and switches | Tap | The new state shown at once; rollback with a message if it failed | Switch flipped back and forth in one session |
| Notifications, toasts and badges | System event or completed action | Short confirmation ("Copied", "Message sent") or a counter | Users repeat the action to check it worked |
| Undo snackbars | Destructive action (delete, archive) | "Deleted. Undo" for a few seconds | Support requests to restore deleted items |
The timing in the table follows Nielsen's response-time limits: about 0.1 second feels instantaneous, about 1 second keeps the user's flow of thought, and around 10 seconds is the limit of attention, beyond which a percent-done indicator is needed.
How to find micro interactions that lack feedback
You cannot see missing feedback in a screenshot; you see it in what users do next. Use these rules:
- If one element collects rage clicks, check its pressed and loading states first. A rage click is several rapid clicks in one spot, which usually means "I am not sure that worked". Start from the heatmap and filter to the element.
- If users click something that is not interactive, the affordance is wrong. Those are dead clicks: something looks clickable but produces no change. The fix is visual, either make it look static or make it work.
- If the same submission arrives twice within seconds, the button allowed a second press. Count duplicates in your order or form database; this catches slow double clicks that are too far apart to count as rage clicks.
- If many users stop at the same step, watch a few sessions. A pause before abandoning often means feedback was unclear, such as an error message that appeared off screen.
- If feedback appears late, measure responsiveness. Interaction to Next Paint measures the delay from a click, tap or key press to the next frame the browser paints. Google treats 200 milliseconds or less as good and above 500 milliseconds as poor, so the pressed or loading state should paint well inside that budget even when the server takes longer.
For a wider view of where these problems pile up across a journey, see user friction.
Worked example: adding a loading state to a submit button
Sample data: a checkout page's "Place order" button calls a payment service that takes about 2.4 seconds. In the old version, nothing changed on screen until the confirmation page loaded. The team added a micro interaction: within 100 milliseconds of the click, the button is disabled, its label changes to "Placing order…" and a spinner appears. Server time stayed the same. Both periods are 14 days.
| Before (no loading state) | After (loading state) | |
|---|---|---|
| Checkout page sessions | 1,600 | 1,650 |
| Rage clicks on the page | 240 | 33 |
| Rage clicks per 100 sessions | 240 ÷ 1,600 × 100 = 15.0 | 33 ÷ 1,650 × 100 = 2.0 |
| Orders submitted | 1,500 | 1,600 |
| Duplicate orders (same cart within 60 seconds) | 90 | 8 |
| Duplicate-order rate | 90 ÷ 1,500 = 6.0% | 8 ÷ 1,600 = 0.5% |
Rage clicks fell from 15.0 to 2.0 per 100 sessions, a relative drop of 1 − 2.0 ÷ 15.0 = 86.7%. The duplicate-order rate fell by 5.5 percentage points. At 10,000 orders, that is 600 duplicates before and 50 after: 550 fewer orders to cancel or refund.
Is the drop noise? A 95% confidence interval for the difference in duplicate rates is 1.96 × √(0.06 × 0.94 ÷ 1,500 + 0.005 × 0.995 ÷ 1,600), about ±1.25 points, so 5.5 points is roughly 4.2 to 6.8 points. The interval stays far from zero.
Two details make this a fair comparison. First, nothing got faster: the micro interaction changed only what users saw while waiting, which proves the point that its value is measured in fewer repeated clicks, not in looks. Second, both metrics were needed. Some duplicate orders came from two clicks a second apart, which do not meet a rage-click definition based on rapid clicks; the order database caught those.
Common mistakes with micro interactions
- Animating instead of responding. A 600-millisecond flourish before the pressed state appears is worse than no animation. Show the state change at once and keep any motion short; the response-time limits above still apply.
- Blocking input during animation. If the user cannot act until a transition finishes, the animation adds to interaction delay and can push INP past 200 milliseconds.
- Too many at once. When every element bounces, nothing signals importance. Reserve motion for state changes the user needs to notice.
- Feedback carried only by motion or color. Some users turn motion off. WCAG 2.2 success criterion 2.3.3 asks that motion triggered by interaction can be disabled unless it is essential, and CSS
prefers-reduced-motionlets a page respect that setting. Pair motion with a text label or state change. - No failure state. A switch that looks saved when the request failed teaches users not to trust it. Every rule needs a visible "it didn't work" path.
- Judging by demo, not by data. A micro interaction that looks good on a fast office network can still be missing on a slow phone. Compare rage clicks and duplicates before and after, split by device.
Finding weak micro interactions in TapCub
The checkout example above runs like this in TapCub:
- Find the element. In All-platform analytics › Behavior › Heatmaps, pick the checkout page and a device. The page list shows clicks, rage clicks and sessions, so rage clicks per 100 sessions is one division. The Rage clicks tab shows where they land; TapCub records a rage click as the third click within 300 milliseconds and 50 pixels. Heatmaps need autocapture (Pro plan and above); setup is in create a heatmap.
- Mark the release. Add the date the loading state shipped in Configuration › Annotations, then set the heatmap's date range to the period before and to an equal period after, and compare (up to 92 days each).
- Watch what users did. All-platform analytics › Behavior › Session replay lists sessions with a rage-click count and can filter to sessions with rage clicks. Replay is off by default and must be turned on by an admin or owner, with input masking kept on.
- Check responsiveness. Monitoring › Performance reports Interaction to Next Paint from real visits, with the 200 and 500 millisecond thresholds, broken down by page and device.
What TapCub does not do: it does not detect dead clicks yet, and it does not flag duplicate orders. The duplicate-order rate comes from your own order system.
See also: What is a modal window in UI? Rules and examples.
Sources
- Nielsen Norman Group, Microinteractions in User Experience (definition as trigger-feedback pairs; cites Saffer).
- Nielsen Norman Group, Response Times: The 3 Important Limits (0.1, 1 and 10 seconds).
- web.dev, Interaction to Next Paint (INP) (200 and 500 millisecond thresholds).
- W3C, Understanding Success Criterion 2.3.3: Animation from Interactions.
- MDN Web Docs, prefers-reduced-motion.
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.

Product Experience Specialist
Writes about product experience, practical workflows, and growth, turning complex processes into clear steps for everyday decisions.