Industries / Consumer goods / Eyewear
Fitting
Matching glasses to a face: choosing a frame that fits, measuring where each pupil sits behind the lens, cutting the lenses to the frame's shape, and checking the finished pair against the order.
6
Milestones achieved, from 1955 to 2019.
2
To watch for: announced, not yet confirmed outside.
3
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1955
Edging by machine
Edging is grinding a lens's outline and bevel so it fits the frame. By 1915 several machinery makers had automated it, the trade historian Joseph Bruneni wrote. The Lemay Electromatic of the early 1940s, which the trade paper Vision Monday calls the first fully automatic "all shape" bevel edger, made odd-shaped fashion frames practical. The Diamaline diamond wheel, introduced in 1955, quickly replaced ceramic wheels. In 1990 Nidek brought out an edger that worked from traced frame data instead of a metal pattern, by the account of its partner Santinelli International.
Credit
Arthur Lemay; Joe Santinelli, who helped design the Diamaline; Nidek
1964
Tolerances for finished glasses
The first US standard for prescription lenses, now ANSI Z80.1, appeared in 1964 and covered only glass lenses; it is meant to be revised every five years. ISO 21987, first published in 2009, set world tolerances for lenses mounted in frames. For example, a progressive lens's fitting point must be within ±1.0 mm of the height ordered. The boxing system of ISO 8624 (1991) gives the frame measurements these orders use.
Credit
the ANSI Z80 committee; ISO/TC 172/SC 7
1970s
Measuring pupils with light
Pupillometers rest on the nose like a frame and use a light reflected from each cornea to measure the pupillary distance (PD), the distance between the pupils, for each eye separately. Varilux, American Optical and Titmus each brought out their own in the 1970s, the optics magazine 20/20 reports. Testing nine instruments from three makers, Obstfeld and Chou found readings repeated with an average standard deviation of 0.26 mm, and advised checking accuracy regularly (1998).
Credit
Varilux (Essilor), American Optical and Titmus
1992
Cameras take the measurements
ZEISS says its Video Infral of 1992 was the first video centration system, a camera that finds PD and fitting heights in place of a ruler and marker pen. In 2018 its VISUFIT 1000 took the measurements in one shot with nine cameras and no calibration clip on the frame. Outside tests found that such systems do not agree with each other. Comparing a frame ruler and three devices on 21 volunteers, Garcia-Espinilla and colleagues at the University of Valladolid found fitting heights from Essilor's Visioffice averaged about 6 mm away from the ruler's, and concluded the measurements were "not interchangeable".
LessonA digital measurement still has to be checked against a fixed reference; instruments that each look precise can disagree.
Credit
ZEISS; Essilor
2011
Online glasses often failed
Karl Citek of Pacific University and colleagues had ten people order two pairs each from 10 of the most visited online sellers. Of the 154 pairs that arrived, 44.8 percent failed at least one optical or impact test, and 22.7 percent had a lens that failed impact testing. In the UK, Alderson and colleagues found 43 of 154 online pairs unacceptable, against 15 of 154 from optometry practices, often for poor centration or frame fit (2016). McMahon and colleagues found that people measuring their own PD in a mirror were neither accurate nor repeatable (2012).
LessonFitting without a person present has to be measured, not assumed. A 2021 study by Gordon and colleagues found failure rates of 8 to 11 percent across three online sellers, and no impact failures.
Credit
Karl Citek (Pacific University), in a study led by the American Optometric Association with the Optical Laboratories Association and The Vision Council; Alderson and colleagues; McMahon and colleagues
2019
Try-on in the phone
In 2019 Warby Parker's iPhone app used the TrueDepth camera behind Face ID, which maps the face in depth, to place frames on the shopper's own face, after using the same camera for frame recommendations. Outside studies then measured phone PD tools against pupillometers. Han and colleagues found the Warby Parker and Eye Measure apps within about 0.5 mm on average in 44 adults (2023), and Jung and Chu found 95 percent limits of agreement of about −2.9 to +1.7 mm for one app on 40 students (2024).
Credit
Warby Parker; Apple, maker of the TrueDepth camera
Watch for
Face scans in every store
Lenskart's chief executive Peyush Bansal told investors that its stores had launched a face scan to help customers choose frames for their faces, and that, with consent, it now gets more than 100,000 face scans a day.
Achieved whenan outside party reports outcomes such as remake or return rates for frames chosen with and without the scan, or audits the scan count.
Claimed by
Lenskart, February 2026
Frames sized for children's faces
Researchers at Aston University measured 480 white British children aged 4 months to 16 years from 3D photographs. They report that children's noses sit relatively low, wide and flat, that frames scaled down from adult designs give a bridge that "will typically fail to contact adequately", and that children's face form calls for a new approach to frame making. A 2025 review by Buthelezi and Hasrod reports that African facial measurements are underrepresented in frame design standards.
Achieved whenanother group repeats the measurements, or a frame maker or standards body adopts them.
Claimed by
Thompson, Logan and Cubbidge, November 2026
Open
01
Fitting height from a phone
A phone app that measures the fitting height for a progressive lens, the height of the pupil above the bottom of the lens opening, in the wearer's chosen frame. Validate it against an in-store centration device and careful manual marking, on at least 50 consenting adult volunteers wearing frames with plain demo lenses.
The briefCredit · unclaimed02
PD apps beyond adults
An open test of at least five PD apps and web tools against a pupillometer, covering distance and near PD, each eye separately, children (with a parent's consent and ethics approval), and a range of iris colours and face shapes.
The briefCredit · unclaimed03
A robot that mounts lenses
A low-cost robot arm with a camera that pushes edged demo lenses into full-rim frames, filmed over at least 100 attempts.
The briefCredit · unclaimed
Open 01 · the brief
Fitting height from a phone
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A phone app that measures the fitting height for a progressive lens, the height of the pupil above the bottom of the lens opening, in the wearer's chosen frame. Validate it against an in-store centration device and careful manual marking, on at least 50 consenting adult volunteers wearing frames with plain demo lenses.
Why it is still open
The tolerance is tight: ISO 21987 allows ±1.0 mm. Veero announced a fitting-height tool for online sellers in August 2023 and publishes no error margins or validation data. Searches of the research literature found no published validation of phone fitting-height measurement. The in-store devices themselves disagree (Garcia-Espinilla and colleagues, above), so the reference has to be fixed in advance.
Built with what exists
An iPhone's TrueDepth camera through Apple's ARKit; Google's MediaPipe Face Landmarker, which gives 478 3D face points on Android, iOS, the web and Python; a centration device such as ZEISS VISUFIT or Essilor Visioffice at a cooperating optician as the reference.
First step
Write the test plan and consent form, choose the reference method, and try it on five volunteers.
Done looks like
For at least 50 wearers and 3 frame styles, the average difference and 95 percent limits of agreement against the reference, the share of readings within ±1.0 mm, and the test plan and anonymised data published.
Open 02 · the brief
PD apps beyond adults
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open test of at least five PD apps and web tools against a pupillometer, covering distance and near PD, each eye separately, children (with a parent's consent and ethics approval), and a range of iris colours and face shapes.
Why it is still open
The independent studies so far tested adults only. Han and colleagues' 44 adults were 81.8 percent white and near PD was not measured, and Jung and Chu tested one app on 40 students. Fittingbox says its tool is within 1 mm for 7 of 10 measurements without stating a method. Buthelezi and Hasrod report that African facial measurements are underrepresented in design standards.
pmc.ncbi.nlm.nih.govpmc.ncbi.nlm.nih.govfittingbox.comebi.ac.uk
Built with what exists
Free PD apps on iOS and Android; a corneal-reflection pupillometer such as Topcon's PD-5 or Essilor's; a written test script.
First step
List the apps, write the test plan and apply for ethics approval for the child participants.
Done looks like
For at least 100 volunteers, each app's average difference and limits of agreement reported for each group, with the data published.
Open 03 · the brief
A robot that mounts lenses
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A low-cost robot arm with a camera that pushes edged demo lenses into full-rim frames, filmed over at least 100 attempts.
Why it is still open
A 2017 report on robots in prescription labs describes them moving job trays, loading the machines that cut and edge lenses, and picking lenses to pair with frames, with none mounting lenses. The closest design found, a Bausch & Lomb apparatus patented in 1999, is semi-automatic. Searches found no lens-mounting robot shown in 2025 or 2026.
Built with what exists
An SO-101 arm with Hugging Face's open LeRobot software (see the Note "No download"); a control policy such as ACT, which LeRobot includes, trained from about 50 teleoperated demonstrations; a 3D-printed fixture to hold the frame; a camera.
First step
Buy an SO-101 kit, print a fixture for one frame model, and record 50 demonstrations by hand.
Done looks like
100 filmed attempts on at least 3 frame models, with the success rate, time per pair and any chipped or scratched lenses tallied, and lens axis and position checked against ISO 21987.