0.94%. 38%. 5.3 times. Those are the three numbers in this post. Here is the twist in the second one: the buttons people rage on work.
Picture the add-to-cart button on a product page. It is a real button, wired to a real handler, and it adds the item every time. Someone taps it. The page says nothing for half a second. They tap again, and again, and the cart now holds three. Who tests the half second after the tap? Nobody. It is not a bug. It is a silence.
Harvv is a pixel that measures that silence. Clarity and Fullstory record people so you can watch them. Harvv turns behaviour into data an AI can read: you ask why conversion dropped on Tuesday and it answers with the element, the page and the fix. The rage click is one of the signals it reads, and this post is what that signal says across the sample.
How often, measured
The median site in the sample has a rage click in 0.94% of sessions. The middle half of sites sit between 0.72% and 1.79%. The quietest is at 0.04%, the loudest at 4.7%. Desktop rages more than mobile, with more clicks per burst: a mouse makes three clicks in under a second easy, a thumb does not.
| Device | Sessions with a rage click | Sessions | Rate | Clicks per rage session |
|---|---|---|---|---|
| Mobile | 13,978 | 2,047,258 | 0.68% | 2.5 |
| Desktop | 5,751 | 380,132 | 1.51% | 3.3 |
| Tablet | 3,276 | 101,720 | 3.22% | 3.1 |
The surprise
Sessions with a rage click converted at 2.2% (500 of 22,861). Sessions without one converted at 0.4% (10,310 of 2,511,159). Rage sessions ran a median of 207 seconds against 22. Longer sessions do more of everything, including buying. So the person hammering your button is not leaving. They are trying to pay you.
Where it lands
We looked under every rage burst in the sample for 14 days to 6 October 2026: 9,998 bursts. The story is that rage clicks land on dead elements. Mostly they do not.
| Element under the burst | Bursts | Share |
|---|---|---|
| A real control (button, link, linked image) | 4,048 | 40% |
| Plain containers and text (div, span, li) | 2,274 | 23% |
| Icon glyphs (svg, path, i) | 2,087 | 21% |
| Images | 306 | 3% |
Four in ten rage clicks are on a button that exists and works. The problem is the half second after the tap, not the button. Three causes cover most of it.
A real button that answers too slowly
38% of rage bursts land on a real button. It works. The answer arrived after the second tap.
Fix: Show a pressed state or a spinner inside 100 milliseconds, then fix the long task behind the delay.
An icon that swallows the click
21% land on an svg, path or i element: the glyph inside the button took the tap and the button never heard it.
Fix: Set pointer-events: none on the icon. The button is the target, not the drawing on it.
A card that looks tappable and is not
23% land on a div, span or list item: the padding around a control, or a product card where only the title is the link.
Fix: Make the whole card the link. A real <a href> or <button> also gives you keyboard and screen-reader access for free.
What the tools call it
Four vendors publish a definition. They agree on the shape and none publishes the numbers that make it a rule, so two tools on one site count different rage clicks.
| Tool | Their words |
|---|---|
| Microsoft Clarity | "Rage clicks occur when a user repeatedly clicks in a clustered area within a short period."source |
| Fullstory | Rage Clicks "are logged when end users click multiple times rapidly in the same area."source |
| Pendo | "Rage clicks are repetitive clicks on a specific element in a software application, often caused by user frustration."source |
| Sentry | "Rage clicks are a series of consecutive clicks on the same unresponsive page element."source |
| Harvv | 3 clicks within 800 milliseconds, all inside a 50 pixel radius. A burst on a text field or quantity stepper is someone editing a value, and a burst that selects text is a triple-click; neither counts. The rule is read from the pixel's source each time this page is built. |
The pattern
Every site in the sample has a button like this. Not a broken one. A quiet one. And it is nobody's job to find it: the developer sees the handler fire, the designer sees the mock, the analytics dashboard sees a click. Only the visitor sees the silence, and they do not file a ticket. They tap three times, and either buy anyway or leave.
Which button on your site is the quiet one?
Questions people search
What is a rage click?
Several fast clicks or taps on one spot, the way a person presses a button again when nothing seems to happen. Harvv counts 3 clicks within 800 milliseconds inside a 50 pixel radius. In a sample of 2,534,020 real sessions, the median site saw one in 0.94% of sessions.
What are rage clicks in Clarity?
Microsoft Clarity’s FAQ: "Rage clicks occur when a user repeatedly clicks in a clustered area within a short period." Clarity shows them as a recording filter and a heatmap. It does not publish the click count, window or radius it uses, so two tools on one site count different rage clicks.
In Pendo, what is a rage click?
Pendo’s glossary: "Rage clicks are repetitive clicks on a specific element in a software application, often caused by user frustration." Pendo measures inside web apps, not public sites.
What does "dead clicks" mean?
A dead click is one click that gets no response at all. A rage click is several in a row. They rarely travel together: 6.7% of sessions with a rage click also had a dead click. Most rage clicks land on controls that did respond, late.
What is a good rage click rate?
Under 1%. In our sample the median site had a rage click in 0.94% of sessions, the middle half of sites sat between 0.72% and 1.79%, and the worst was at 4.7%. Above 2% is one control misbehaving. Find the element, not the average.
Are rage clicks bad for conversion?
Sessions with a rage click converted at 2.2% (500 of 22,861). Sessions without one converted at 0.4%. The person hammering your button is trying to buy. That is why the half second after the tap is worth fixing.
Method: sessions table, 90 days to 7 October 2026, a sample of sites with at least 1,000 sessions each after excluding sessions classified as bots or internal; per-site rates, then the median and quartiles across sites. Element shares from rage events in the 14 days to 6 October 2026. Vendor definitions quoted from each vendor's own page on 8 October 2026. The detection rule is read from the pixel source at build. Related: what is a dead click, why people add to cart and do not buy.