Research
Every number below is read off the shipped code, not off a note somebody wrote about it. Each one sits next to the standard or the study it came from, and the small raised numbers elsewhere on this site point here.
How to read this page
A design decision is easy to state and hard to check. So each entry gives three things: what The Ken does, the figure actually in the build, and the source. The sources are other people's. Every one of them opens in full.
The entries are numbered for life. If we ever change our mind about one, the number stays and the entry says what changed.
Hands
The Ken is used by people whose hands shake, whose fingers are stiff, and who press harder and longer than a touchscreen expects. Four decisions come from that, and all four are measurable.
Buttons are far larger than a phone's.
In the build: the main buttons on the Ken's screen are between 80 and 110 pixels tall. The smaller Settings buttons are 56 pixels, which is the accessibility floor plus about a quarter again.
W3C Web Accessibility Initiative, Understanding Success Criterion 2.5.5: Target Size (Enhanced). Sets 44 by 44 pixels at level AAA, and gives the reason as helping "users who may have trouble activating a small target because of hand tremors, limited dexterity, or other reasons".
And larger again than the accessibility floor, because that floor was written for everyone rather than for older hands.
In the build: each face on the home screen is a tile filling roughly a quarter of the screen, so the control used most often is the biggest thing on it. The Settings buttons in 1 are the small end at 56 pixels.
Z. X. Jin, T. Plocher and L. Kiff, "Touch Screen User Interfaces for Older Adults: Button Size and Spacing", in Universal Access in Human-Computer Interaction, Springer, 2007. Measured reaction time, accuracy and preference against button size, button spacing and manual dexterity, on the stated ground that existing recommendations were written for a general audience.
A tremor cannot press the same button twice.
In the build: a second tap that lands within a centimetre of the first and within 350 milliseconds of it is thrown away. On the on-screen keyboard the window is 250 milliseconds, because "77" and "nn" are things people mean to type. Separately, more than ten taps in a second is treated as a child playing with the screen, and the Ken stops accepting taps for half a second.
S. Agarwal and M. C. Biagioni, "Essential Tremor", StatPearls, StatPearls Publishing. Describes essential tremor as a 6 to 12 Hz postural tremor of the hands, and notes that its incidence rises with age. A shake in that range puts repeat contacts roughly 83 to 167 milliseconds apart, which is inside the window above; a deliberate second press is not.
A card that has just appeared will not act on a press that was already on its way.
In the build: when a card appears over the screen, its buttons are held for 1.5 seconds. The card says "Just a moment...", the button dims, and a dark bar sweeps along its bottom edge, so the wait is visible three separate ways rather than felt as a broken device.
Same tremor evidence as 3. The hold exists because a hand travelling towards a button it wanted a moment ago will land on whatever has replaced it.
Eyes
Sight changes with age in ways that are well documented, and the accessibility standards say out loud which age they were calibrated against. That made the arithmetic straightforward.
Text has to clear 4.5 to 1 against what is behind it.
In the build: two tests compute the ratio rather than trusting a number written in a comment. One reads the colour values out of the shipped files; the second reads the colour off the rendered page, which covers the cases the first cannot parse. That distinction is not theoretical. One page on this site carried a comment claiming 8.27 to 1 above a pairing that measured 3.96, and the comment is why nobody looked again for months.
W3C Web Accessibility Initiative, Understanding Success Criterion 1.4.3: Contrast (Minimum). States that 4.5 to 1 "was chosen for level AA because it compensated for the loss in contrast sensitivity usually experienced by users with vision loss equivalent to approximately 20/40 vision", and that 20/40 "is commonly reported as typical visual acuity of elders at roughly age 80". That is our reader, described by the standard itself.
Things that are not text have their own floor.
In the build: the bar that sweeps across a held button was once a pale wash that measured 1.05 to 1 on two buttons, so a person pressed, saw nothing, and had no way to tell a waiting Ken from a broken one. It is now dark ink at 80 per cent, measured on every button it lands on: 4.64 to 1 on gold, 7.96 on cream, 3.58 on green, 3.01 on red.
W3C Web Accessibility Initiative, Understanding Success Criterion 1.4.11: Non-text Contrast. Sets 3 to 1 for user interface components and meaningful graphics.
Colour never carries the meaning on its own.
In the build: the green is #2AA06A and the red is #D94A3D, a pair that stays distinguishable to someone who cannot tell ordinary red from ordinary green. Neither is the only signal: an answer button carries the word "Answer", a status dot carries a tick or a cross. In the admin portal a test holds those two colours and the amber to one definition each and refuses an off-palette green, red or amber anywhere in the markup.
W3C Web Accessibility Initiative, Understanding Success Criterion 1.4.1: Use of Color. Names people with partial sight, people with limited colour vision and "older users with reduced color perception" as the reason.
The text is set at large-print size and does not need to be enlarged first.
In the build: body text on the Ken is 22 pixels, which is about 16.5 point. Live captions during a call are 44 pixels, about 33 point.
Office for Disability Issues, Accessible communication formats, GOV.UK. Asks for "a minimum 12 point text size", says to "consider using 14 point text if the audience includes people with sight loss", and defines large print as "a point size of 16 and above".
Ears
A call is captioned on screen, and the volume is set by the family rather than by the person struggling to hear.
In the build: both call screens carry live subtitles, at the 44-pixel size in 8, and a test refuses a build where either screen loses them. Volume is set from the admin portal, from anywhere.
RNID, Prevalence of deafness and hearing loss. "Almost 80% of people over the age of 70 will have hearing loss", and "over half the population aged 55 or more have hearing loss".
Time
A message on screen waits until the person is finished with it.
In the build: notices stay for 20 seconds on the device and in the admin portal, and a tap dismisses one at any point. The portal's used to clear themselves after 3.5 seconds.
W3C Web Accessibility Initiative, Understanding Success Criterion 2.2.1: Timing Adjustable. Written because "people with disabilities such as blindness, low vision, dexterity impairments, and cognitive limitations may require more time to read content".
Why it exists at all
Being cut off carries a measurable health risk.
In the build: the whole device. One screen, the faces of the people who matter, and nothing between the person and a call.
J. Holt-Lunstad, T. B. Smith, M. Baker, T. Harris and D. Stephenson, "Loneliness and Social Isolation as Risk Factors for Mortality: A Meta-Analytic Review", Perspectives on Psychological Science, 10(2), 2015, pages 227 to 237. After controlling for confounds, social isolation carried an odds ratio of 1.29 and loneliness 1.26, which the authors give as a 29 per cent and 26 per cent increased likelihood of mortality.
The ways a scammer reaches an older person are the ways an ordinary device is useful, so The Ken does not have them.
In the build: no phone number, no email inbox, no web browser, no app store and no SIM card. The faces on the screen are the contacts the family approved in the portal, and a caller who is not one of them cannot get through.
Age UK, "Older person becomes fraud victim every 40 seconds", 30 July 2019, analysing the Crime Survey for England and Wales 2017-18. Around 8 per cent of respondents aged 65 and over, more than 800,000 people, reported being a victim of fraud in the previous year.
What this page is not
Citing the W3C's accessibility criteria is not the same as being audited against them, and we do not claim a certification we have not been given. Our claim is narrower and checkable: the figures above are the figures in the code, and the sources are the ones we used.
The Ken is a communication device. It is not a medical device, it does not diagnose or treat anything, and none of the research here is offered as clinical advice.
If you think one of these sources does not support what we have built on it, tell us and we will either fix the build or fix the page.