Library self-checkout kiosk: designing for the 5th and 95th percentile at once
A public library in Sha Tin is replacing its staffed counter with a self-checkout kiosk used by about 900 borrowers a day, including elderly residents, children and wheelchair users. The design team proposes: a fixed counter 1,050 mm high and 700 mm deep; a 15-inch screen set at 20° from horizontal, 400 mm back from the front edge; a recessed keypad with 30 mm keys; a barcode scanner window at 1,150 mm; and a bagging shelf at 1,000 mm. The room has 500 lux ceiling downlights behind the user, and the screen glass is glossy. Staff report that borrowers lean over the counter to read the screen, elderly users sometimes press keys twice, and wheelchair users cannot reach the scanner or see the screen because their knees hit the 1,050 mm apron. Available data: Hong Kong adult stature from 1,480 mm (5th percentile female) to 1,800 mm (95th percentile male); comfortable standing counter height about 900–1,050 mm; accessible counter height 750–850 mm with at least 650 mm knee clearance; comfortable forward reach about 500–600 mm.
1. Evaluate the proposal against the data and propose a resolution that works for all three user groups.
Read the worked answer
The proposal fails three groups at once, and the failures can be traced to specific dimensions:
- Wheelchair user. A 1,050 mm counter with a 1,050 mm apron gives zero knee clearance; at least 650 mm clear height and 750 mm width are needed. Maximum comfortable forward reach is 500–600 mm, but the scanner is 400 mm back and 1,150 mm high, so it is outside the seated reach envelope. A screen only 20° from horizontal cannot be seen over the counter edge from a seated position.
- Standing adults. 1,050 mm is at the top of the comfort band, so the counter height itself is defensible, but the screen is wrong: for a standing user the screen should be tilted about 15–30° back from vertical, with the top near eye height (about 1,500 mm for a 5th percentile female). A near-flat screen forces borrowers to lean over, which is exactly what staff report. The scanner at 1,150 mm is also low for many adults.
- Elderly users. Glossy glass with 500 lux downlights behind the user produces veiling glare. The 30 mm keys are adequate (19 mm minimum recommended), so double pressing is a feedback or debounce problem, not key size.
Resolution: a two-level counter.
- Accessible bay: 800 mm high, 900 mm wide, at least 700 mm knee clearance, with screen and scanner within 500 mm of the front edge.
- Standing bay: about 950 mm high, with the scanner at 1,000–1,100 mm and the screen tilted 20° back from vertical.
- Fit a matte anti-glare film or a hood over the screen.
The examinable reasoning: design to the tails - 5th percentile female for reach, 95th percentile male for clearance - and treat the wheelchair user as a primary user, not an exception to be accommodated afterwards.
2. Diagnose the double key presses and propose error-proofing measures, with a way to measure success.
Read the worked answer
Double keying usually has three different causes, and each needs a different fix. Treating them all as 'users pressing too hard' would be a poor answer.
- No debounce. A mechanical key bounces for 5–20 ms, and a fast scan loop can register several presses. Fix in firmware with 20–30 ms debounce and a 300 ms per-key lockout.
- No immediate confirmation. If the screen responds only after a network round trip, users press again. Fix with local feedback within 100 ms: a key highlight, a short beep, and a large on-screen echo of the entered digit.
- Layout and workflow. Adjacent keys and an enter key placed next to a digit invite mis-hits. Use at least 19 mm key pitch, 3 mm separation, and a distinct colour and shape for cancel. Where a long borrower number must be entered, make barcode scanning the primary path and the keypad the fallback.
The general error-proofing principle is to make the correct action the easiest action, and to make an error visible and reversible before the transaction commits.
Measurement: observe at least 30 borrowers across the user groups; record correction events per transaction; target ≤ 2% of transactions with a correction, against the current baseline, and compare before and after the firmware change so the cause is isolated.
3. Outline a test plan that would validate the design before installation.
Read the worked answer
A usability and anthropometric test plan should recruit to the tails, not the middle.
Participants. At least 12: three wheelchair users, three adults aged 65 or over, three children aged 8–12, and three adults spanning the 5th–95th stature range. Record each participant's stature and reach so results map back to the population.
Tasks and success criteria. Locate the card slot, scan a book, read the due date, and complete a checkout unaided in ≤ 90 s.
Measures.
- Reach: can the participant touch the scanner and card slot without standing or leaning past the counter edge?
- Viewing: screen luminance and contrast under 500 lux with the participant in place; note any glare.
- Performance: error rate and time on task.
- Subjective rating 1–5 for ease of use.
Method. Build a mock-up with adjustable height (700–1,100 mm) and adjustable screen tilt, so the dimensions are determined by the test rather than assumed. Then iterate once and retest with a different group of 12 to avoid learning effects.
Acceptance. All participants complete the task unaided, and at least 90% rate 4/5 or better. Any dimension that fails for one group is adjusted before the fixed joinery is ordered. The examinable judgement is that ergonomic decisions should be evidence-based and tested with real users at the extremes, not justified by average dimensions.