Creative Testing on Marketplaces: Kill Your Favorite Hero Image

Every brand has a hero image somebody is in love with. It got chosen in a meeting. The founder signed off on it. The photographer was expensive and the lighting is gorgeous. And it has been sitting on your top listing for eight months, quietly converting worse than three alternatives nobody bothered to test. This is the uncomfortable truth about creative on Indian marketplaces. The image your leadership team loves and the image that wins the click are rarely the same image, and the only way to tell them apart is to test. Taste is not a strategy. It is a hypothesis you have not checked yet.

We say this as people who care about craft, not in spite of it. Beautiful work matters. But on a marketplace grid, beauty that does not earn the tap is just an expensive opinion. The discipline that separates brands that grow from brands that plateau is the willingness to put the favorite image in a fair fight and let it lose.

Why founder taste loses on a phone grid

The reason has nothing to do with the founder being wrong about beauty. It is that the founder is judging the image in the wrong conditions. They see it full-screen, on a retina laptop, in a deck, with context and pride attached. The buyer sees it shrunk to a thumbnail, in a crowded grid, on a mid-range phone, in daylight glare, for half a second, next to fifteen competitors.

Those are different images. The one that photographs like a magazine cover can read as mud at postage-stamp size. The one that looks plain in the studio can pop against a row of pale, samey thumbnails. Your eye in the meeting cannot predict the buyer’s eye in the grid. Only the test can.

The image that wins the boardroom and the image that wins the click are almost never the same image, and you do not get to vote on which is which.

This is the core of why we argue, repeatedly, that you should test the image and not the bullet when conversion stalls. The hero shot is the single highest-leverage variable on the listing, and it is also the one most polluted by internal attachment. High stakes plus high bias is exactly where structured testing earns its keep.

What a structured creative test actually is

Most teams think they test creative. What they actually do is swap an image when someone gets bored, eyeball the next two days of sales, and declare a result. That is not a test. That is superstition with a dashboard. A real creative test has a few non-negotiable parts.

• A clear hypothesis. Not “let us try a nicer photo.” Something falsifiable, like “a three-quarter angle that fills more of the frame will beat the current front-on shot on click-through.” You should know in advance what would prove you wrong.
• One variable in motion. Change the hero image and nothing else. Hold price, title, bullets, and inventory steady. If you move three things, a result teaches you nothing about which one mattered.
• A defined success metric. Decide before the test whether you are judging click-through from search, detail-page conversion, or net orders. The image’s first job is the click, so click-through usually leads.
• Enough traffic and enough time. A thin listing cannot resolve a small difference. Run long enough to clear daily noise and at least one weekly cycle, because Indian shopping rhythms swing hard around paydays and sale events.
• A pre-committed decision rule. Agree up front what margin counts as a win. Otherwise the favorite image gets graded on a curve and survives on sentiment.

The last point is where most brands quietly cheat. They run the test, the challenger wins by a clear margin, and then someone in the room finds a reason the old favorite still deserves to stay. The decision rule exists to take that conversation off the table before emotion enters it.

The things worth testing first

You do not have infinite traffic, so you cannot test everything. Spend your test slots on the variables that move the most, roughly in this order.

• Scale in frame. Most weak hero images are simply too zoomed out for a phone. Filling more of the frame is the most reliable lift we see, and it costs nothing but a recrop.
• Subject clarity. Can a buyer tell what the product is, and how many they get, in half a second. Bundles and multipacks ruin this constantly.
• Angle. Front-on versus three-quarter versus styled. Categories reward different conventions, and assuming yours without testing is how favorites calcify.
• Contrast against the grid. Your image competes with its neighbors, not with your brand book. A hero that stands out in a real search result beats one that only looks good in isolation.

Notice that none of these require a new shoot. The highest-return creative tests are usually recrops, reframes, and angle swaps of assets you already own. That is a feature, not a limitation. It means you can test fast and cheap before you spend on production, which is the whole point of having a real creative production pipeline behind a multi-marketplace brand rather than a one-off shoot every quarter.

Killing the favorite without killing morale

There is a human problem here that the testing literature ignores. The favorite image belongs to someone, often someone senior. Retiring it on the data can feel like retiring their judgment. Handle that badly and your team learns to fear tests, which defeats the purpose.

The fix is cultural, and it is simple. Make the test the authority, not any person. When the hypothesis and the decision rule are written down before the test runs, nobody loses an argument when the challenger wins. The data did. The founder who agreed to the rule in advance gets to be the person who runs a disciplined shop, not the person whose photo lost. Framed that way, killing a favorite becomes a sign of operating maturity rather than a personal defeat.

We have watched this flip a brand’s whole relationship with creative. Once the team sees a humble recrop beat a beloved studio shot, and sees the orders that followed, the attachment loosens. People stop defending images and start proposing challengers. That is the culture you want.

Where the favorite still deserves a home

To be fair to the beautiful work, the hero test does not mean craft is wasted. The image that lost the thumbnail fight is often perfect somewhere else. The styled, atmospheric, founder-loved shot frequently belongs in the secondary gallery, the A+ modules, or the brand store, where the buyer is already on the page and you are now closing rather than catching.

That is a different job with different rules. Enhanced content and store design exist to deepen conviction and justify price once attention is won, which is exactly why we are careful about when A plus content actually pays and when it is vanity. The favorite image rarely belongs in position one. It often belongs three slots later, doing emotional work, or anchoring a brand store that sells instead of just looking pretty. Retiring it from the hero spot is not throwing it away. It is putting it where it converts.

What changed recently

Two shifts in the last year make the testing discipline more important, not less. The first is that producing a challenger has gone nearly free. In November 2025 Amazon rolled its AI Video Generator out to advertisers in India, which turns a single product image into up to six finished video options in three to four minutes at no cost, automating scene selection, music and headlines, on top of the existing Image Generator that builds lifestyle backgrounds around a product shot, per Exchange4media. When the cost of generating a variant collapses, the only thing standing between you and a better hero is the willingness to test what the machine spits out. Cheap creation makes disciplined selection the scarce skill, not the other way around.

The second shift is where the clicks are moving. Datum Intelligence projects that Blinkit, Zepto and Instamart will together pull roughly Rs 4,900 crore in advertising revenue in 2026, up from about Rs 3,000 crore in 2025, with brands now routing a meaningful slice of performance budgets into sponsored listings and search placements on those apps, according to Storyboard18. That matters for creative because a quick-commerce tile is even smaller and more ruthless than an Amazon thumbnail. The same single-variable test you run on your marketplace hero now has a second, faster grid to win, which is part of why we treat quick-commerce creative as its own discipline and not a copy-paste of your Amazon assets. More paid surfaces means more places where the favorite image can quietly lose money until you put it in a fair fight.

The operating rule to take away

If you remember one thing, make it this. No hero image earns its place by taste alone. It earns its place by beating a challenger in a fair, single-variable test with a decision rule set in advance. Run that test on your top listings on a standing cadence, not once a year, because the grid keeps changing and your competitors keep testing too.

This is the discipline our Brand & Creative Studio runs alongside Catalog & Listing Optimization, with Creative Production feeding fresh challengers and Marketplace SEO making sure the winning image is actually shown to the buyers it can convert. The craft still matters. It just has to win on the evidence, every time, against your favorite image included. Kill the one you love when the data says so, and you will sell more than the brands still protecting theirs.