Contents＋
Tap. Nothing. Tap again—and the page moves.
01 / Direct answer
Visitors do not simply lack patience; risk appears too quickly
The overlooked mobile problem is not merely a smaller screen but the loss of overview. Desktop can show proposition, proof, navigation and next step together. Mobile reveals one fragment at a time. If the first fragment lacks direction, the second is a large image and the third finally names the service, the visitor has already paid three comprehension costs.
02 / Magnification
Mobile magnifies every ambiguity
Illustrative scenario, not a client case: A Hong Kong director searches for IT support on the MTR. A cookie panel appears, then a full-screen photograph; the WhatsApp bubble covers pricing and the menu contains fourteen items. The visitor does not leave because a competitor is half a second faster. In a crowded, one-handed, variable-network context, the page offers no reason to continue.
Looking loaded and being usable are different
A background can appear quickly while the headline is missing, a button is unresponsive or a late image shifts the target. Visitors experience instability rather than one loading-time score. Observe main-content appearance, interaction response and layout movement, including whether cookie notices or floating chat controls cover fields.
03 / Hong Kong reality
Hong Kong visitors perform the first filter in fragmented moments
Baymard's large-scale mobile usability testing found that 63% of participants abandoned a product or site at least once because of preventable mobile usability issues, identifying loss of overview as a central challenge.
web.dev case studies connect performance improvements with revenue and conversion, but correlation does not mean one timing metric explains every exit.
Mobile abandonment combines three delays: delivery delay from server and assets, interaction delay from shifting layouts and unresponsive controls, and comprehension delay from copy and structure. Compressing images may improve the first while the other two keep the site feeling slow.
04 / Management error
A speed score can hide what actually feels slow
Teams declare the mobile problem solved when a PageSpeed score rises, or delete content after seeing a high bounce rate. The first confuses a lab score with buyer experience; the second can remove answers required for trust. Ask when the most important content becomes stable and whether the next step is obvious.
Smooth on office Wi-Fi doesn’t necessarily mean smooth on the journey home. Different conditions. Different story.
05 / Four frictions
Separate loading, interaction and comprehension
- Record a first load on an ordinary phone and mobile network, not only office Wi-Fi.
- Capture the page at five seconds, ten seconds and three swipes to see what the visitor knows at each point.
- Hide imagery and brand name and test whether content order still explains the service.
- Measure field LCP, INP and CLS, then compare them with actions on valuable pages.
- Remove or delay anything that obscures content without directly helping the task.
Lab tests compare changes under consistent conditions. Field data reflects actual devices and networks, but a low-traffic page may have insufficient data; absence is not a pass. Use a real device, a slower connection and a performance trace to reproduce the first failed tap and locate the script, image or server delay.
06 / Real-device execution
How to run a real-device diagnosis
A mobile audit must cover loading, visual stability, comprehension speed and action cost. Any layer that forces visitors to find their bearings again can turn a few seconds into a reason to leave.
- Open the five highest-traffic pages on a mid-range phone over a mobile connection.
- Record first useful content, shifts, obstructions, type size, forms and return paths.
- Retest first visits without cache and the journey back from WhatsApp.
- Repair one friction class at a time, then compare device- and page-level evidence.
07 / Measurement
Measure technical speed and decision speed separately
| Layer | What to record | What it means for management |
|---|---|---|
| Technical speed | LCP, INP, CLS and real-device observation | Locates loading, interaction and movement faults |
| Decision speed | Time to meaningful scroll or service selection | Shows whether the message needs decoding |
| Action cost | Steps and failure rate for forms or WhatsApp | Finds friction that actually blocks enquiries |
A quick exit can also be successful: finding a phone number, checking hours or recognising an unsuitable service. Trying to maximise time on page can introduce unnecessary steps. The useful question is whether an intending customer was prevented from finishing, not whether every short visit can be prolonged.
08 / FAQ
Three common mobile-exit questions
Is a speed score of 90 enough?
No. It is a laboratory signal; real devices, third-party scripts, pop-ups and message clarity can still fail users.
Should mobile copy be shortened?
Reorder before deleting. Essential answers should arrive earlier while deeper evidence can remain lower on the page.
Does a floating WhatsApp button always help?
Only when it does not obstruct content, expectations are clear and someone can handle the resulting conversation.
09 / Limits and conclusion
A very fast website can still trigger an immediate exit
Information-heavy finance, education or technical pages should not sacrifice necessary detail for brevity. Long pages can work when sequence is clear, important information appears early and users know where they are. The goal is less disorientation, not universally shorter pages.
Mobile visitors often leave quickly not because they lack patience, but because the site asks for too much attention before revealing a direction.
The score can stay in the report. The button still has to work when someone taps it.
10 / Sources