Fixing Size Mismatches and Discovery Gaps on Myntra

Phase 01 · Problem discovery

✳The problem they went after

Myntra shoppers cannot trust that a size will fit. Sizing shifts brand to brand, generic size charts do not match actual garment dimensions, and there are no peer fit reviews to fall back on, which makes buying high-effort and low-confidence. Across 29 surveys and 9 user interviews the problem validated universally, with respondents reporting return rates of 25 to 75 percent on size-related purchases.

✳What they actually did

Three hypotheses were tested. Beyond sizing, discovery came back generic and catalogue-driven rather than personal, causing browsing fatigue and abandoned sessions, while frequent shoppers said loyalty benefits felt inconsistent and did not justify sticking with the platform. PIF scoring on population, intensity and frequency put size mismatch at P0. The write-up then sizes business impact per problem: ₹50 crore from a 21 percent cut in size-related returns, 40 crore from better session-to-purchase conversion, and 15 crore from higher 60-day repeat purchase if the loyalty gap is closed.

✳What you can take from it

Worth stealing

PIF scoring on population, intensity and frequency to pick P0
29 surveys plus 9 interviews validating one problem
Revenue sized separately for each of three problems
Self-reported return rates of 25 to 75 percent

Read this if

Anyone turning mixed research into a ranked, defensible problem statement

The transferable bitThree hypotheses get tested and then actually chosen between, which is where most research decks stop. PIF gives the ranking a defensible shape, and putting a separate rupee figure against each problem turns findings into something a business reader can act on. Framing the call against AJIO, Nykaa and quick commerce gives the P0 decision its urgency.

✳The submission itself

Open the PDF →

Reading on a phone? The embedded viewer is cramped — open the PDF in a new tab instead.

Phase 02 · Solution design

AI Size Prediction and Virtual Try-On for Myntra Shoppers

✳The problem they went after

The write-up puts size uncertainty behind 40 percent of fashion e-commerce returns and 100 to 150 rupees per order in logistics costs. On Myntra the specific trigger is brand inconsistency: a high-intent shopper who wants the item still cannot confidently pick a size across dozens of labels. That hesitation turns into abandoned carts, returns, and revenue going to whoever guides sizing better.

✳What they actually did

SmartFit AI Predictor trains on purchase history, return reasons, brand-specific fit deviation matrices and fabric characteristics, then outputs a recommended size with a confidence score. The product detail page shows it next to social proof along the lines of 94 percent of users like you kept this size, plus a transparent explanation link. Virtual Try-On goes further, generating a 3D avatar from uploaded photos or the phone camera, simulating how fabric drapes, identifying fit pressure points, and giving a rotating 360-degree preview with a colour-coded fit heatmap.

✳What you can take from it

Worth stealing

Size recommendation shipped with a confidence score
Brand-specific fit deviation matrices as a model input
Social proof line: users like you kept this size
Colour-coded fit heatmap marking pressure points

Read this if

PMs designing ML recommendations users must trust before acting on

The transferable bitBoth solutions aim at the same instant, the moment of size selection, and both treat the deliverable as confidence rather than data. The transparency link is the detail to copy: a recommendation nobody trusts does not change behaviour, so the explanation belongs inside the feature rather than in documentation somewhere.