I built three landing pages for an avocado slicer. Each made a different pitch. The page I expected to do best lost, and the ugliest one sold the most. I started the pages with equal budgets, then moved spend toward the combinations that performed. My own preference would've sent the money somewhere else.
The client was an anonymous general ecommerce store selling kitchen gadgets in 2017. It had plenty of products and no clear winner. Facebook ads could send customers to the store, but the catalogue gave us a lot of possible things to sell and little discipline for choosing between them.
This was my first practical work in conversion rate optimisation, or CRO: improving how well a page turns visitors into customers. I binge-watched Ezra Firestone's ecommerce material and put the ideas into an actual campaign. The avocado slicer gave me one product on which to try different ways of selling.
The store needed a way to compare its offers
The store's revenue was scattered across products. Budget went into different bets without a structured way to compare the offers or the pages selling them. A sale could tell us somebody bought a gadget. It couldn't, by itself, tell us whether a different pitch for that gadget would've worked better.
I kept the avocado slicer at the centre of the page test. That gave the comparison a useful boundary. We could change the argument for buying it while still selling the same item.
A prettier version of one page would've left the underlying pitch largely intact. I wanted to compare different reasons to buy. That meant giving each page enough freedom to look and read differently, even when one version didn't match my idea of what a kitchen-product page should look like.
I tested three pitches with equal starting budgets
I tested three sales arguments for the same product. One used a hard sell and a simple package offer. Another appealed to people interested in cooking and trying things in the kitchen. The third copied the look of an existing page and tried to make the presentation visually appealing.
I expected the cooking-audience version to do best. It seemed like a sensible match between a kitchen tool and people who cared about cooking. The hard-sell version was the ugly one. It made a simple package easy to understand, and it was the version that performed best.
We used Facebook Ads and Shopify, with separate landing pages and supporting campaigns. Each campaign had its own ad-set and ad-level adjustments. I started with an even budget split, reassessed performance every three days, cut the losers and moved budget toward the winners. I also checked where the campaign links sent people so I could connect performance to the pages.
That cadence gave me repeated opportunities to change the allocation. The three-day interval describes how I worked. It doesn't establish statistical confidence or a rule that every store should copy. The surviving account gives no traffic counts, per-page conversion rates or confidence calculation.
The test also changed more than appearance. The pitch, package and presentation varied together, and the advertising changed as we went. We learned which selling combination to keep funding. A claim about the effect of a particular colour or headline would need a narrower comparison.
Same product, three arguments, three different-looking pages, drawn here as sketches of each pitch.
The winning concept produced roughly $600K in under 60 days
The ugly page outperformed the alternatives. My case account puts sales from the winning avocado-slicer concept at roughly $600K in under 60 days. I remember the period as possibly shorter, but under two months is the boundary I was sure of. I wouldn't tighten it further.
The broader store did over $3M in sales. I also recalled a little over $700K from my own advertising across the engagement. Those figures describe different scopes: the avocado-slicer concept, my advertising work and the whole store. The entire store's revenue can't be credited to my campaigns, and the recollection doesn't provide matching accounting periods for all three totals.
These are historical sales figures from my account of the work. They don't establish profit, return on ad spend or the extra revenue caused by the test. The surviving material doesn't include the advertising costs, margins or an untreated comparison.
We tested more products and gave winners dedicated pages
The work extended beyond one slicer page. Pancake mixers and cake-decorating tools also became part of the store's offers. We bundled products, looked through the general catalogue for other candidates and tested them. Products that performed earned dedicated landing pages and a place in the mix.
I recalled the store having those additional offers by day 90. My later account put the decorating kits and pancake mixers before the avocado slicer, reversing the earlier sequence. I can't give that expansion a confident product-by-product timeline. The consistent part is the process: we tried offers, kept the ones that sold and concentrated on them.
That gave the catalogue more focus. The decorating kits became a main area of attention, and the avocado slicer developed its own niche. The commercial change I can describe is a store organising sales around tested offers and pages. The account doesn't quantify an improvement in stock planning or cash management.
I think the package appealed to people curious about cooking
My preferred page assumed that people interested in cooking were the natural audience for a kitchen gadget. It lost. My retrospective explanation is that an experienced cook had less reason to buy this sort of dropship item, while someone curious about cooking could enjoy the idea of the gadget and the package around it.
That's an interpretation of the result. The surviving account doesn't establish those buyer identities through interviews or a customer-segment analysis. The simple package may have made the purchase easier to understand. I also thought the slicer's novelty and the interest in avocado toast at the time helped. I can't separate those explanations from the record we have.
The distinction changes what I'd take into the next job. Copying the ugly appearance would preserve the part I can see while potentially missing the reason somebody bought. The package, the pitch and the audience all deserve attention. A page can look suitable for cooking enthusiasts and still sell less than a blunt offer aimed at someone curious about cooking.
Ezra's material helped me question the assumption that a page had to look good to sell well. This campaign gave that question a practical answer: I had to fund a page I'd rated below another version. It doesn't make poor design a general strategy, and it doesn't make a polished page a commercial failure before anybody tests it.
My read: a shopper who already cooks seriously rarely buys a twenty-dollar gadget off a Facebook ad. A shopper who likes the idea of cooking will, and may add the rest of the bundle.
Compare different reasons to buy the same product
This case is useful for a merchant with a broad catalogue and no clear winner, or a brand deciding whether a redesign should change the sales argument as well as the appearance. It offers a way to frame the next experiment. The historical revenue is no forecast for another store.
I'd choose a product and write down the different reasons someone might buy it. Each page should express one of those pitches clearly enough that the comparison means something. I'd keep track of which ads lead to which page, set a decision rule before reallocating spend and judge performance using the economics available for that business. My old three-day reassessment schedule needs enough evidence behind it to be useful in a different account.
A team looking only for cosmetic approval won't get that answer from a sales test. A team willing to compare different offers can discover that its preferred page loses. Mine did. I let the result change where I put the budget.