Ashish Sharma
About
A guided tool allowing users to measure their frames by placing them on their phone screen & measuring half the frame, users can accurately determine their frame size (XS, S, M, L, XL) in Lenskart sizing terminology.
Company:
Full time with LENSKART
Year:
2024
My role
I owned this project end to end.
> Research - understanding existing sizing methods & identify the problems
> Ideation and iterations - create new process for measuring the frame size & translating it in digital experience
> Working with engineering to spec and ship it
Three weeks. One designer.
How it started
Our CEO pointed to a ring sizing tool and asked one question:
❓ Can we do something like this for frames?
That was the entire brief. No scope. No spec. No guarantee it was even physically possible.
My job was to find out — and make it real.
Understanding the
problem
Frames at Lenskart come in sizes — just like a t-shirt - XS, S, M, L, XL. For best comfort user needs to know their size.
We had 3 sizing methods to help users get this right
> Manual selection: No guidance, no confidence.
> Past purchase: Only useful for repeat buyers. Useless for first-timers
> AR tool: Our flagship. But it doesn’t work properly on low-end Android.. Even on supported devices, accuracy was only ~80%
The result?
Sizing related returns:
~6% of all returns are sizing-related
Initial research
Through 34 user interviews across repeat and recent buyers (both returners and non-returners), we understood two main themes
> Frame Returns - 33% for sizing
> Frame sizing method - 70% from AR sizing and other from past purchase
Also watched users attempt the AR tool. Face alignment errors, and Android limitations caused visible frustration and abandonment.
> In 30% cases, user are getting different sizes at different tries from AR
"If you could find your current frame size yourself, would it help?"
And 70% users said "Yes".
Exploring directions
Before starting we did a thought experiment to know if happen make a sizing tool for measuring users glasses would it help?
We understood:
> User who wears oversize frames, for them AR won't work at all
>Works best if you can see your screen clearly without frames (glasses)
>Or if you have a backup pair to wear while measuring
>Won't work for people who wants to buy their first frame
> User will have best fit as they are already wearing best fit
We ideated few possible directions we explored:
> Photo of their frame with reference scale
> Photo or video of user already wearing their frames with some reference object
But after discussion with Dev but it was rejected
Core challenge
We needed to find out if the solution was even physically possible.
Three constraints hit immediately:
> Frames are wider than most phone screens — full frame measurement wasn't possible
> Without a clear reference point, measurement is inconsistent
> Frames need a stable position
Through hands-on testing we found the answer:
👉 Measure only half the frame — half gives you the full size
👉 Placing the frames upside down gave the best stability and a clear reference point
The solution
How users discover the Magic Scale
We didn't hide the Magic scale in settings. We brought it to where users already were.
> Home - Proactive nudge for new users or anyone wanting to redo their size
> Profile - Where all saved sizes live, for users returning intentionally
> PLP - We trigger sizing when a new user lands on the product listing page. Users who complete sizing convert 1.5x higher — so surfacing it here was a no-brainer
> PDP - We show the sizing nudge to do size near the size selection widget
Key decision
Onboarding screens
Initially, we didn't want onboarding at all — every extra step is a drop-off risk. But the tool was too novel to skip education entirely.
> For 1st iteration - No onboarding Everything explained on the Magic Scale itself. Users had to read instructions while placing frame simultaneously, too much cognitive load. It failed.
> For 2nd, Explainer video + direct to Magic Scale We added a real-person explainer video before the Magic Scale. But users forgot the instructions by the time they reached the measurement screen. It failed.
> For 3rd, Video + 3 onboarding steps + manual nudges Users finally understood the tool. This shipped as V1. But data showed drop-off in 3 steps and in the manual nudges.
> For 4th, Video + 2 steps + auto-dismissible nudges Reduced steps removed friction. Auto nudges appeared contextually - users couldn't ignore them. Drop-off reduced, completion improved.
👉We made skip available but not dominant — a wrong measurement hurts the user more than a drop-off hurts us.
Screen layout hierarchy
Before designing any interaction, we needed to figure out how to divide the ruler (Scale) screen. We started by testing existing ruler (Scale) apps.
One thing was immediately clear - rulers (Scale) placed at the edge of the screen were hard to use. The Frame would overlap the scale, making it impossible to read the measurement accurately.
So we made two foundational decisions that shaped every screen that followed:
> Move the ruler (Scale) to the middle - Always visible, never overlapped by the frame above or the interaction below
> Divide the screen into two clear sections - Top for frame placement, bottom for interaction
We applied this same logic from the very first onboarding screen — top shows the frame placement visually, bottom has the Next button. By the time users reached the Magic Scale, the layout already felt familiar.
👉 Consistency from onboarding → Magic Scale meant zero relearning at the most critical moment
Slider mechanics
While figuring out the measurement process, we tested with existing ruler (scale) apps — general purpose rulers (scale) where both ends of the slider are movable.
In workshops, we immediately saw confusion. Users didn't know which end to move first, or how to anchor the measurement. For a general ruler (scale) that makes sense — but for us the process was already unique. Users were learning something completely new.
We had three priorities:
👉 Understanding - the interaction had to be immediately obvious
👉 Accuracy - fewer decisions meant fewer chances for error
👉 Simplicity - Minimum action
We simplified to one decision:
Fix the left end at zero → user aligns glasses to zero, drags only the right slider to the centre of the frame
Two things we rejected:
>Both ends movable → confusion, two actions, more chances for error
>Dark mode → better focus but hurt visibility of glasses on screen
👉 One anchor. One action. Easier to understand, easier to get right.
Slider feedback
Once the slider interaction was defined, we needed users to feel confident about their measurement in real time. We built three feedback mechanisms that work simultaneously:
>Green gradient fills left to right as the user drags — showing exactly how much of the frame has been measured
>Frame silhouette at the bottom adjusts with the slider — giving visual confirmation that alignment is correct
>Dynamic pill shows the exact measurement updating in real time — always centred in the half silhouette, moves with the slider
Together these three removed the need for any instructional text during measurement. Users could see, feel and read the measurement happening simultaneously — building confidence at every moment of the interaction.
Keeping size context alive
Measuring is only half the job. The real question was - what happens after the user gets their size?
Once measured, we carried the full context forward into the shopping experience:
👉 132mm + size label shown together on every PLP card - not just the letter M, not just the number. Both together gave users complete confidence they were looking at the right fit
👉 "Perfect fit for you — As per Magic Scale" banner attributes the recommendation directly to the scale - maintaining the trust chain from measurement all the way to purchase
The measurement doesn't disappear after the tool closes. It follows the user through their shopping journey
Naming - Magic Scale, not Magic Ruler
We chose "Scale" over "Ruler" deliberately.
In India, nobody says ruler. We say scale — right from childhood. Using "scale" made the tool instantly familiar and memorable. The name tells you exactly what it does, in the word you've always used for it.
Building it
Shipping Magic Scale wasn't just a design handoff. Three technical challenges required us to get into the details.
Pixel density calibration
For the Magic Scale to be accurate, every phone screen needed to be mapped to real-world centimetres.
The logic: get the screen size in pixels → convert to cm → draw the ruler accordingly.
iOS was straightforward — only a handful of devices to account for.
Android was a different story. Thousands of device variants, no standard. We identified the 70 Android phone models that covered 95% of our users, manually built a database of their PPI, and created an API that detects the device and pulls the correct dimensions automatically. Every measurement on every phone is now accurate.
Drag slider equation
The slider wasn't a native Android component - it was a completely new system. Every element needed to scale correctly across different phone sizes. The scale had to stay centred, the silhouette had to adjust proportionally, the gradient had to map accurately.
We sat with the developer and we iterated together — writing and testing equations until the system was scalable across all phone sizes.
iOS loop video
The onboarding video needed to loop seamlessly on iOS. Native looping wasn't working reliably.
Our fix - loop a single video 5 times in sequence. Clean, stable, no interruption.
Impact
We expected the Magic Scale to perform around AR. It outperformed it.
Magic Scale conversion:
Magic scale
AR
Conversion
home to buy:
4.9%❤️
3.8%
Completion rate:
57% ❤️
70%
Return reduced:
From 6.2% to 5.4% in ~3 months
Revenue saved:
~ 18 Lakhs / month
From data for Magic Scale -
> 10% users were returning the frames
> 6% of returns are because of fit & finish, rest for sizing
The surprise insight
👉 We spoke to users who measured and purchased. The reason was consistent — "Because I measured it myself, I trusted the result."
Returns
Sizing-related returns dropped from 6.2% to 5.4% after launch.
63% user have sizing in system, out of these
> 52% uses AR sizing tool to get their frame size
> 16% uses Magic Scale for sizing to get their frame size
> 30% uses Past purchase for determining frame size
> 2% uses direct selection
Further scope of improvement
The Magic Scale solves measurement. But it doesn't solve memory.
132mm, M, and a date that's all we save. Come back after a month and it means nothing.
👉 The next unlock is personalization — attaching a name, a photo of the frame, or a photo of the person to the saved measurement
Thanks for reading 😊
