Rethinking Meesho’s product card for 250 million people who browse by picture.
Most of our shoppers live in tier 3 and tier 4 towns, and some are new to e-commerce and shop only on Meesho. All 250 million of them just open Meesho whenever they have time and start browsing. They scan by picture only, and they operate in a mode of rejection. They are image-first: they stop when the image is nice, and everything else comes later.
Over time, more than ten teams had added their own features to the card. Each addition made it taller, so shoppers saw fewer products on every screen, and the pictures they rely on got harder to compare.
The pod’s mission: help 250 million shoppers find the right product faster, whatever the category and whatever brings them to browse.
We worked on the card in two directions, one experiment at a time, each measured on its own.
The new card is live for 250 million shoppers. Results are confidential for a listed company; I walk through them in interviews.
The full case study expands here. Live, it sits behind a password that comes with my application.
In the full case study:
Why
More than ten teams had placed features on one card with no shared rule for what earned a place. Each line pushed another product off the screen for a shopper who was only looking at pictures.
How
We started by understanding every case the card has to carry, then built a framework from them: the right information architecture for browsing cards, and one that scales.
Live; next version in test
Why
Seller-written titles were long, repetitive and stuffed with search words, and most of them carried shop names, which added no value. Shoppers glanced and read nothing. The facts that decide a purchase (fabric for a kurti, pack size for a snack) were not on the card, so people opened products just to check.
How
Defined a minimum viable truth per category: the smallest set of facts that lets a shopper decide without opening the product. Replaced the title with those facts as chips. A second version tightened the chips, added consideration tags (price drop / quality marks; built, not launched — too few products qualified), and forced three rounds of catalogue cleanup so the chips were true.
What worked
Chips were noticed late but valued once seen; shoppers stopped opening products to check fabric. Pack size was the most useful fact in ambiguous categories. A chip the picture contradicts is not believed, so catalogue quality had to come first. Showing the prepaid price up front confused cash shoppers — fed the cash-price work.
Inconclusive after three runs
Why
The grid padded every row to its tallest card. Once chips and tags varied by product, that padding was whitespace, and by the third row the shopper had lost a product she could have seen.
How
Let each card take its natural height in the For You feed so columns stagger. Three runs, each bundled with other changes.
What worked
Shoppers never noticed misaligned text under the images (confirms picture-first). No clean read because of bundling; result stays open. Rule kept: never force staggering — it is what variable content does when the grid stops hiding it.
Live, reading
Why
Grid suits products judged on looks. Headphones and other considered categories are judged on facts a grid card cannot carry.
How
Full-width list rows for considered categories. The platform picks list or grid by category; no user toggle.
What worked
Live, read in progress. Direction supports rows where details decide, grid where looks decide.
Shipped
Why
Shoppers judge from pictures and the card showed one; every second view cost a tap in and a tap back.
How
Up to three images swipe inside the card, showing more of the product and the variations it comes in, one autoplay to teach the gesture. An early version that rationed scrolling to selected products was killed on day one — breadth was the point.
What worked
Taps into the product fell while orders per tap rose: one finding, not two. The decision moved onto the listing.
Why
For a garment on a body the look is the decision, and the standard image was too small to judge it.
How
Taller 4:5 image on fashion cards, tested across the feed and then by category.
What worked
Fashion converted on fewer, larger cards; non-fashion paid for it in the same feed. The rule it points at is intent, not a blanket change.
Live
Why
Most shoppers pay cash on delivery and the card showed the prepaid price. A higher number at the door cost the category, not one order — shoppers stopped buying that kind of product.
How
Cash price in its own row under the prepaid price; one rule for the card: anything price-related lives in that row.
What worked
Cash shoppers stopped being surprised at the door; the cohort that had been hidden from the real price responded.
Scaled
Why
Shoppers assumed four to five days in cities and a week or more outside them; “Free Delivery” on every card told them nothing.
How
A day count at the foot of the card, shown only on products where delivery is fast, not on all of them. Three tests, including one with a “FAST” mark in front of the number.
What worked
The date did the work; the word in front of it carried nothing. Third test scaled.
