Why Vintage Sellers Get Buried in Whatnot's Feed — and How Labubu Got Unburied

OK so I've gone and read Whatnot's engineering blog. All of it. Six posts deep at this point, which I realize is not a normal hobby. But I found the one that actually explains why some of us feel invisible no matter how good our shows are — and it isn't the one about the feed. It's a post about their taxonomy, which is a fancy word for the list of categories you pick from when you set up a show. Whatnot's own Discovery team published it in March. It lays out this exact problem, start to finish, and how they fixed it. They just happened to be talking about Labubu instead of about us.

Before anything else, the same promise I made the first time: none of these posts are “here's how to win.” Nobody at Whatnot wrote a guide for us. This is me reading their engineering write-ups and working out what they mean for people who sell. Where I'm inferring, I'll say so. Where they said it outright, I'll quote them.

1. The feed is a matchmaker with a memory, not a leaderboard.

It's not asking “who's the biggest seller today.” It's asking “which show should I put in front of THIS shopper right now,” based on what they've been watching and buying lately. So you are not stuck in one giant line behind every seller on the whole app — which honestly takes a ton of pressure off. You're not trying to beat everybody. And here's the part that really matters — that matchmaker remembers. Somebody who watches and bids and buys from you once is more likely to get your next show shoved right in front of them. So your repeat buyers aren't just nice sales — they're quietly telling Whatnot to show you to them again. It's why I'll take 20 people who keep coming back over 2,000 who followed me once and then vanished. A follow is a vanity number. Somebody actually showing up and bidding again — THAT'S the stuff that counts.

That's the good news, and it's real. But there's a flip side I didn't write about last time, and it's the whole reason this article got longer. A matchmaker with a memory is fantastic at giving you back the people you already have. It is much, much worse at handing you somebody new. Because to match a total stranger to your show, it needs something to go on — something about your show it can actually read. Hold that thought.

2. Your first few minutes still matter — just not in real time.

This is the one that changed how I run my shows, and it survives the correction — it just runs on a slower clock than I told you. Early watches and early bids are what tell Whatnot “this show is popping,” and that does get folded back in and pushed out to more people. But on an hourly cycle, not instantly. So front-load it anyway: put your best, most eye-catching item up first, start the bidding low so hands get moving, and call out your watcher count out loud. Watching turns into bidding turns into buying — early on, your only job is nudging lurkers into that first bid. Just don't sit there refreshing and expecting the feed to react by minute eleven. It won't. Think of it as an hour-scale loop, not a dial.

And one more thing from that May post that I think is genuinely kind to hear. The reason they wrote it is that a bad code deploy stopped one chunk of their data from being produced at all, and the ranking quietly got worse and nobody caught it, because it degraded slowly instead of breaking loudly. They've since added alarms for it. But here's my point: sometimes the machine is just broken, and even Whatnot doesn't know yet. Not every bad week is something you did. I really wish somebody had told me that two years ago.

3. Now the part I actually wrote this for: Labubu.

In March 2026, Whatnot's Discovery team published a post about their taxonomy. And they say straight out what that category tree controls. Their words: taxonomy “shapes onboarding (what categories new users see), browse (how they navigate), search (how results are ranked), and recommendations (what their feed looks like).”

Read that again. Search ranking. And your feed. That's not me reading between the lines — that's Whatnot saying the category tree decides who finds you. I'd been treating that as a theory. It's documented.

Then they tell this story. When Labubu blew up in 2025, it didn't have its own home — it was filed under Toys. And here's their sentence: “When Labubu surged in 2025, it was subcategorized under Toys, and new users struggled to find relevant shows.”

A thing got popular. It was buried inside a category too broad to describe it. And new shoppers couldn't find the shows. Not because those sellers were bad. Because the tree didn't have a branch for them.

So what did Whatnot do? “When we elevated Labubu to the first page of onboarding (a similar treatment given to other big categories like Pokémon Cards), it reduced this friction and helped the category explode.”

They gave it its own spot. And it exploded. Their word, not mine.

I want to be really clear that I don't think anybody did anything wrong here. That's a good decision, made fast, and it's exactly what a marketplace should do when a community shows up. What gets me is that this is our story too — and it already has a documented happy ending, for somebody else.

4. So what does our branch of the tree actually look like?

Here's where I have to be straight with you about something I didn't fully clock until I sat down and counted. Whatnot's category for us is Antiques, Vintage & Ephemera. It has four subcategories. Four. Total.

Antiques. Ephemera & Postcards. Postage Stamps. Vintage Decor.

That's the whole tree for roughly two centuries of objects. Every piece of milk glass, every hen on nest, every quilt, every crock, every set of barware, every McCoy planter, every Christmas box lot — all of it lands in one of four boxes, and realistically most of it lands in the same one.

Now go look at what a coin seller gets. When you list a coin, Whatnot's own form asks you for composition, precious metal weight, fineness, year, mint, mint mark, strike type, and whether it's graded — and if it is, the grade, the certification, and the certification number. That's a dozen specific fields that describe the actual object. A sports card seller gets team, athlete, whether it's autographed, country of origin.

We get none of that. There is no maker field. No era field. No pattern field. No material field. Not because our stuff is simpler — a Pyrex pattern is at least as specific as a mint mark — but because nobody built those fields for our category. And you can't fill in a box that doesn't exist. That's not a Shelf-Snap problem I can solve for you either, and I'm not going to pretend otherwise. No app can type into a field Whatnot didn't make.

So count up what we actually control. Your title. Your description. One of four subcategories. And your three tags. That's it. That's the entire set of levers.

And here's the thing about those three tags that I cannot stop thinking about. Go scroll the Vintage Decor category right now and look at the little gray line under each show. I did, and on a single screen I counted the same three words — Vintage, Antique, Home Decor — over and over, on show after show after show. We've all got three slots, we're all picking from the same list, and we've collectively decided to spend them on the three most generic words available.

Which means that of the four levers we have, we're all voluntarily wasting one. If everyone's tags are identical, tags tell the feed nothing about who's who. My read, not Whatnot's: if you're going to have three slots, spending even one of them on something specific and a little bit weird — something nobody else in the category is using — is at minimum a way of being distinguishable. I don't have proof it moves placement. Nobody does. But right now most of us are handing the machine three words it has already seen four hundred times today.

So here's what I'm doing about it, and I want to be really upfront: this is an experiment I started this week, not advice. I put lobster on my shows. As a tag. I do not sell lobsters. I went looking and as far as I can tell nobody else in this category is using it, and that was the whole appeal — because why not, honestly.

My reasoning, such as it is: if three tag slots are the only show-level thing I control, and two of them are words four hundred other people used today, the third may as well be a word that's actually mine. Best case, the people who've bought from me before have one more thread tying them back to my shows. Worst case it does nothing and there's a lobster on my shows, which — fine. I can live with that.

I have no idea yet whether it does anything. It's been days, not months, and one seller's shows is not data, it's an anecdote with a lobster on it. But I'll report back either way, including if the answer turns out to be “absolutely nothing” — that's worth knowing too, and I'd rather tell you that than quietly never mention it again. If you want to run your own version, pick something specific to you rather than copying mine, because the entire point is being the only one using it. (Also I called it first. 🦞)

5. The ask — because this is fixable, and they've told us how

Here's the part that changed my whole attitude about this. I was ready to write a piece about how the algorithm is stacked against us. Then I read the rest of that taxonomy post and realized that's not the story.

Whatnot say they support “over 200 categories across seven countries, and that number increases weekly.” Weekly. And they rebuilt their whole system specifically to make these changes fast — they mention that updates used to take “days of engineering effort to create and roll out,” and their reasoning for fixing that is “the hottest trends don't happen over months, but start suddenly.”

So this isn't a technical wall. The tree changes constantly, there's a process for changing it, and they've publicly committed to changing it fast when a community needs it. Vintage just hasn't been the community they did it for yet.

That makes this an ask instead of a complaint. And I think it should be a specific one, because “support vintage sellers more” is not something anybody can action. So here's mine. This is my wish list, not anything Whatnot has agreed to or commented on:

One — subcategories that match how vintage is actually collected. Not four. Nobody hunts “Vintage Decor.” They hunt glassware, pottery and ceramics, kitchenware, linens and textiles, lighting, barware, holiday, advertising. Those are real collecting lanes with real communities behind them, exactly like Labubu had.

Two — four attribute fields. Maker. Era or decade. Pattern. Material. That's it, that's the whole request. Those four would do for us what mint and year and strike type do for coins. And they'd make our listings machine-readable for the first time.

Three — onboarding placement for the big vintage lanes. This is the actual Labubu fix. Whatnot's own September 2025 post says the feed for brand-new shoppers leans on the categories they picked when they signed up. So what's on that first onboarding page decides which sellers new buyers ever meet. If “vintage glassware” isn't on it, a new collector has to already know to go looking for us.

None of that is radical. Every one of those things already exists for other categories. We're not asking to be treated specially — we're asking to be treated like Labubu.

So what do you actually do on Monday?

Until the tree changes, work the four levers you've got, and work them hard. Set your category before you go live and make it match what you're actually selling in that block — Whatnot's April 2025 post names show category as the single most important thing the ranker reads at the moment somebody opens the app. Write your title like a headline. Use your description, all of it, because for us it's one of only two places a maker name or a pattern name can even live. And stop spending all three tags on the same three words as everybody else.

Then go be findable in the places the tree can't gatekeep — your repeat buyers, your regulars, the people who already know your name. The matchmaker is genuinely good at giving those people back to you. That part isn't broken at all.

(And yes — obsessing over exactly this stuff is why I built Shelf-Snap. When your inventory's listed in a few minutes instead of an hour on a laptop, you've got energy left to write real titles and run a sharp show. Free to start, it's on the App Store. That's all I'll say about it.)

Anyway. If you sell vintage and you've felt like you were shouting into a void — I don't think you imagined it, and I don't think it's your show. Go read the Labubu bit yourself. And if you've got a fifth thing that belongs on that ask list, tell me, because I'd like to make it a proper list with a lot of names behind it 🤝

Whatnot feed ranking FAQ

How do I rank higher on Whatnot's feed?

Set your show category correctly before you go live — Whatnot names show category as the most important thing its ranker reads at request time — then drive engagement early. Whatnot's behavior features (watches, bids, orders) refresh on an hourly pipeline targeting under 60 minutes, not instantly, so front-loading your best item still helps but on an hour-scale loop rather than in real time. Lead with your strongest item, start bidding low, and call out your watcher count. Running shows regularly keeps you visible between streams.

Do early bids really matter more than my overall sales history?

According to signals in Whatnot's own engineering writing, yes. The feed behaves like a matchmaker with a short memory rather than an all-time leaderboard — it reacts to what's happening in your show right now. Strong early watching and bidding tell the feed your show is worth pushing to more people while you're still live.

Does listing more items help my Whatnot show?

More inventory gives viewers more reasons to keep watching and bidding, which feeds the early-engagement signals the algorithm rewards. The catch is prep time — listing a big haul one item at a time eats the energy you need to run a sharp show. A bulk listing app like Shelf-Snap lets you list a full batch in minutes so you can go live with more, faster.

What's the fastest way to list a lot of items before going live?

Use a bulk listing tool instead of adding items one by one. Shelf-Snap lets Whatnot sellers snap photos, apply reusable templates for repeat item types, auto-number the batch, and export a Whatnot-ready CSV straight from an iPhone — no laptop required. It's free to download on the App Store.

Do better photos affect how Whatnot ranks me?

Indirectly, yes. Clean, eye-catching photos earn more clicks and watches, and those early engagement signals are exactly what the feed pays attention to. Shelf-Snap includes an on-device background remover (powered by Apple Vision) so your items look sharp without any extra editing apps or per-photo cost.

How often should I go live to stay visible in Whatnot's feed?

Consistency beats intensity. Because the feed favors recent, active shows, a regular cadence keeps you surfacing to viewers more reliably than occasional marathon streams. Reducing the time it takes to prep each show makes a steady schedule far easier to keep.

Why do vintage sellers get less reach on Whatnot?

Because Whatnot's category tree gives them very little to be ranked on. Whatnot's Discovery team states that taxonomy “shapes onboarding (what categories new users see), browse (how they navigate), search (how results are ranked), and recommendations (what their feed looks like).” The Antiques, Vintage & Ephemera category has only four subcategories — Antiques, Ephemera & Postcards, Postage Stamps and Vintage Decor — and its listing form offers no attribute fields, where Coins & Bullion offers around twelve including composition, year, mint and strike type. With little structured information to separate one vintage show from another, ranking leans on behavior signals, which favor sellers who already have buyers. That last step is an inference from Whatnot's published architecture, not something Whatnot states.

What is Whatnot's taxonomy and why does it matter for sellers?

Taxonomy is Whatnot's tree of categories and subcategories. It matters because Whatnot says it drives onboarding, browsing, search ranking and the For You feed. Whatnot supports “over 200 categories across seven countries, and that number increases weekly,” and rebuilt its tooling for speed after updates previously “took days of engineering effort to create and roll out.” When Labubu surged in 2025 it sat under Toys and, in Whatnot's words, “new users struggled to find relevant shows” — so Whatnot elevated it to the first page of onboarding, which “helped the category explode.” Where your show sits in that tree is therefore a genuine reach factor, not a filing detail.

Does Whatnot's feed update in real time?

Partly. Show category and a new buyer's onboarding category preferences are read at request time, when someone opens the app. But the behavior features — watches, bids and orders aggregated across users, sellers and categories — are recalculated on an hourly pipeline with a stated target of under 60 minutes end to end, per Whatnot's May 2026 engineering post. That post exists because the schedule had silently slipped to every two hours without triggering an alert. So early engagement in a show does feed back, but on an hour-scale loop rather than instantly. An earlier version of this article said five minutes; that was wrong.