Bundling Was Built for Things That Cost Nothing to Copy

You have three things open from the same seller and you are doing postage arithmetic. The second jumper is not really under consideration on its own merits. It is under consideration because it might make the first one cheaper to receive. Baymard Institute's 2026 aggregate of cart-abandonment research names the leading reason people walk away, and it is unexpected extra costs: postage, fees, tax. Roughly half of abandoned carts die there baymard. Every shopping app has read that number. Almost all of them read it as a checkout problem.

That is the part I think is wrong. The flat threshold at your cart behaves as if it were pricing theory written for goods that cost nothing to copy: cable channels, software licences, back catalogues. On a second-hand marketplace every item exists exactly once and has to be dug out, folded and posted, so none of those assumptions survive. On one-of-one physical goods, a flat cart threshold is a guess about a cost that theory never priced, and a person has to referee the guess.

Who Wrote the Theory, and What Were They Selling?

The modern case for bundling starts with a paper from 1976. In the Quarterly Journal of Economics, William Adams and Janet Yellen showed how bundling sorts buyers by what each would privately pay adams-yellen. Someone who values the first item at nine pounds and the second at two will still take the pair at ten. You capture a sale you would otherwise lose, at a price nearer the ceiling than the floor. It is a clever result. It is also, quite openly, a result about the seller's side of the table. Yellen was about thirty. She later chaired the Federal Reserve, where she became known for knitting through meetings, and then served as Treasury Secretary. The co-author who mapped how sellers capture more of what buyers would privately pay went on to set the price of money.

Twenty-three years later, Yannis Bakos and Erik Brynjolfsson named the conditions in Management Science bakos-brynjolfsson. Bundling pays when marginal cost is very low, meaning the next unit costs the seller almost nothing to supply. It also wants demand across the items to be uncorrelated. Cable channels. Software licences. Music catalogues. Now hold that against a wardrobe of second-hand clothes. Each listing exists once and has to be found and posted, and a buyer browsing one seller has taste that is anything but uncorrelated. Applying their conditions to resale is my reasoning rather than their claim, but the conditions are theirs, and resale meets neither.

Vinted Built the Honest Version. It Still Stops at the Cart.

Vinted's bundle is the best version of this I have used, and it is worth being precise about why. Three items from one seller travel in one parcel and carry one Buyer Protection fee instead of three. That fee has a fixed component as well as a percentage, so the saving is real money rather than a rounding difference vinted-buyer-protection. The buyer sends one offer across the group instead of haggling item by item. A seller can set a standing discount that applies the moment a buyer crosses a threshold. None of that is theatre. It answers a cost the buyer can actually feel.

The ceiling is that every part of it operates on things you have already found. You can only bundle inside a wardrobe you already opened, from a seller you already chose, so the feature never once helps you find anything. Amazon at least aims at the discovery question with its frequently-bought-together strip. The famous claim that recommendations drive around a third of its sales traces back to a McKinsey attribution. No method for it has ever been published. IKEA organises browsing by room, but those rooms were merchandised centrally, months before you arrived. In all three the bundle is a price, and the shopping happens somewhere else. That matters, because people commit to more when they can see the arithmetic and author the yes themselves.

The Discount Is Priced Off a Guess

A seller-set percentage is fixed before the bundle exists. It cannot see what the buyer has chosen. Two coats of similar value carry a genuine postage saving, while a fifty pence scrunchie paired with a ninety pound jacket is somebody tripping the threshold. So the platform hands the judgment back to a person, and the seller can decline vinted-buyer-protection. That is a sensible safeguard. It is also the exact moment the feature stops doing its job. A bundle promises fewer steps between wanting several things and having them, and this adds one at the point of commitment, performed unpaid by whoever is holding the phone.

Ask who absorbs that guess, and you find the postage again. In October 2024, Retail Technology Innovation Hub reported that Vinted had made its Royal Mail option permanently available to buyers, with sellers unable to switch it off rtih-vinted. It was the second such change in two months, and the first had been withdrawn after complaints. A seller posting as Lisa's £1 challenge put the consequence plainly. Some sellers cannot get to these places, she said, and when a buyer picks one of them the sale is usually lost. Cancelling is a faff, and nobody can be bothered to rebuy. The postage is the one number here that is genuinely computable. It is also the number the seller controls least.

The Weight Is in the Walk Home

A parcel is the object at the end of all this, and you can feel what a parcel costs long before you can calculate it. Think of the walk back from a market with the bags. The plastic handles thin out and bite into the fingers, the load begins to swing, and somewhere around halfway you stop and swap arms without deciding to. That stop is the actual price of the third thing you bought. Nobody at the stall charged you for it.

Which is why you buy three things from one stall rather than one thing from three. Your hands made that choice several minutes before your appetite was given a vote. Every bundle discount in every app is trying to put a number on that walk. Not one of them has ever measured it.

The Strongest Case for Leaving It at the Cart

Let me give the objection its fairest hearing. Checkout is where intent is highest and where the numbers finally exist, because until the buyer has chosen there is no parcel to weigh and no fee to divide. Suggesting bundles during browsing sounds obvious until you try it on unique stock. The item that would complete the pair may have sold four minutes ago, and a suggestion the buyer cannot act on is worse than silence. Second-hand supply is one of one, which is the condition under which recommendation engines are least useful and most irritating. A human approval step at these volumes is cheap and reversible, and fairer than most automated pricing. A seller declining a bundle is a person using judgment. A model repricing a wardrobe overnight is not. On that reading Vinted has not failed to solve discovery, it has correctly declined to pretend.

All of which holds, so long as the bundle's job is pricing. Baymard's 2026 aggregate found that unexpected extra costs are the leading stated reason people abandon a cart baymard. Total cost is deciding these purchases. The buyer begins deciding it well before the cart, at the moment they choose whose wardrobe to open. Baymard measures abandonment rather than wardrobe selection, so that last step is my inference and not their finding. The offer still arrives late. By the time the discount appears, the buyer has already paid the search cost of finding a second thing worth adding. The platform learns nothing from the thousands who tried and gave up.

The Parcel

On one-of-one physical goods, the flat discount at your cart is a guess about a cost the zero-copy theory never priced. The question worth asking is not how to automate the seller away, but what the carry costs and who is paying for the fact that nobody measured it. Lisa has already answered that, and the answer is that she loses the sale, because the platform picked the courier and the buyer could not be bothered to rebuy.