Marketplace search is not a mystery box. Meesho, Flipkart, Amazon, Myntra, AJIO and Nykaa run different code, but they optimise for the same outcome — the listing most likely to convert a search into a delivered, un-returned order — and they read the same five signals to find it.
Relevance. Whether your title, category and attributes match what the buyer typed. This is the cheapest thing on the list to fix and the most commonly left broken: a listing filed under the wrong category or missing its fabric, sleeve-length and occasion attributes cannot rank for the searches those attributes describe, no matter how good the product is.
Click-through rate. Of the buyers shown your listing in a grid, how many tap it. In fashion the grid is almost entirely image — a thumbnail, a price and a rating. Your first photo is doing nearly all of this work.
Conversion rate. Of the buyers who tap through, how many buy. This is price, reviews, size information and — heavily — whether the additional photos answer the questions the thumbnail raised: how it falls, how it looks from behind, what the fabric actually is.
Seller reliability. Dispatch inside the SLA, low cancellations, low returns. Marketplaces demote unreliable sellers regardless of product quality, because a cancelled order costs them a customer, not just a sale. Return rate is the one most sellers underestimate: a listing whose photos flatter the product converts well and then gets returned, and the engine reads that pattern faster than you do.
Paid amplification. Ads buy you the impressions to generate the signals above. They do not fix a listing that converts badly — running ads on a poor listing simply pays to prove it converts badly, faster.
The order matters. Relevance and reliability are hygiene: get them wrong and nothing else helps. Click-through and conversion are where fashion sellers actually win or lose, and both are decided by imagery long before they are decided by copy.
Amazon apparel listings typically allow up to seven images plus video. Most small sellers upload two. That is the single largest piece of unused conversion surface on the average Indian apparel account, and unlike ads it costs nothing recurring.
Slot one — the main image. Pure white background, RGB 255,255,255, product filling roughly 85 percent of the frame, no logos, watermarks or text. Amazon enforces this one, and a non-compliant main image can get the listing suppressed rather than merely demoted.
Slot two — full-length on model. The buyer's first real question after "what is it" is "what does it look like worn". Answer it immediately.
Slot three — the back. For anything with a back design, a drape or a fastening, this is the slot that prevents a return.
Slot four — fabric or print detail. A close-up that lets the buyer judge weave, embroidery or print quality. On ethnic wear this is often the slot that justifies the price.
Slot five — fit and scale reference. Something that establishes proportion, whether that is a measured overlay or a full-body shot with visible sizing context.
Slots six and seven — styling and context. How it pairs, how it looks in a real setting, the detail that makes a browsing buyer imagine owning it.
The reason to treat these as seven distinct jobs rather than seven photographs is that each slot's value comes from the objection it removes. Seven near-identical angles remove one objection seven times.
Sponsored Products is the fastest way to get impressions on a new Amazon listing and the fastest way to waste money on a bad one. The mechanism is simple: your advertising cost of sales is your spend divided by the revenue it produces, so it is governed by conversion rate at least as much as by bid price.
A listing with two images converts worse than one with seven. Running ads on the two-image listing raises ACoS, drains budget, and produces the conclusion that Amazon ads do not work — when what actually happened is that the seller paid retail price for clicks and then failed to convert them.
The sequence that works is unglamorous. Complete the listing. Establish a baseline organic conversion rate. Then advertise, and expect ACoS to be materially better than it would have been. The same budget goes considerably further against a complete listing, and the sales velocity it generates lifts organic rank, which keeps producing orders after the campaign pauses.
Most listing advice treats photography as a presentation detail. On a marketplace it is a ranking input, and it is the only input that touches three of the five signals simultaneously.
A better first image raises click-through, because the search grid is a wall of thumbnails and yours is competing on visual quality alone. A complete set of secondary images raises conversion, because every unanswered question — how does this drape, what does the back look like, is this the same shade in daylight — is a reason to close the tab. And an accurate set lowers returns, because the most expensive return in fashion is the one where the buyer says the product looked different online.
That last point is where flat-lay-only catalogues quietly bleed money. A flat-lay is a truthful photo of a garment lying down, and buyers cannot read fit or fall from it, so they guess — and a meaningful share of them guess wrong. On-model imagery does not flatter the product, it removes the guess.
This is the honest case for AI catalogue photography rather than the promotional one. It is not that AI photos are magic; it is that most small sellers ship two flat-lays because a studio shoot costs ₹15,000–₹50,000 a day and cannot be justified per SKU. Generating a full on-model set from one flat-lay at ₹10 a photo removes the reason the catalogue was thin in the first place.