Every catalog has a few of them. A tool you can’t name. A piece of pottery with a mark you don’t know. Some industrial thing off a farm that three people at the preview will each call something different. You have to put a number on it anyway, because a lot with no estimate looks neglected and a lot with a bad estimate scares off the person who’d have paid fairly.
Here’s a method that holds up, and an honest account of where it breaks.
First, know what a number is worth
There are two completely different things that both look like a price estimate, and you have to be able to tell them apart.
The first is a guess. Somebody — a person or a language model — looks at the item, reaches into memory, and produces a figure. It feels authoritative. It is worth nothing. A model asked “what’s this vintage Rolleiflex worth?” will hand you a confident number pulled from a blurry average of everything it ever read, with no way to check it. That’s not pricing. That’s a slot machine with good manners.
The second is a comp-grounded estimate. You find recent completed sales of comparable items and anchor your number on what those actually brought. It won’t be perfect, but it’ll be defensible, and it tends to land close — not because the method is magic, but because you’re citing instead of guessing.
So the whole job of pricing a lot you don’t recognize comes down to one thing: turn a guess into a citation.
The method
1. Describe it before you name it
You don’t need to know what a thing is to find what it sold for. You need its observable features: material, size, marks, maker’s stamp, a number cast into the base, the shape, the wear pattern. A “5-gallon salt-glaze stoneware crock with a wing mark and two ear handles” is findable even if you’ve never heard the name Red Wing. Write down what you can see. That description is your search query.
2. Search completed sales, not asking prices
This is the step most people get wrong. A listing at $400 tells you what one hopeful seller wants. A completed sale at $180 tells you what a real buyer paid. Only the second kind counts. Look for sold/completed results — on the marketplaces, in past auction archives, anywhere the transaction actually closed. Ignore the aspirational asks entirely; they’re the reason so many reserves die on the block.
3. Pull several comps and take the middle
One comp is an anecdote. Five is a range. Gather the closest handful of recent sales, throw out the obvious outlier at each end (the one that went cheap because it was chipped, the one that went high because two collectors got into it), and take the middle of what’s left. That trimmed middle is your estimate. The lowest and highest real sales become your sanity check — if your number lands outside them, you did something wrong.
4. Publish the range, keep the receipts
Set your estimate as a low–high range, not a single figure, and hang onto the comps you used. When a consignor asks why the family heirloom is estimated at $200, “here are five identical ones that sold between $155 and $275 in the last year” ends the conversation. “That’s just what I think” starts a fight.
Watch the systematic bias
When we tested guess-based estimates against comp-grounded ones on real lots, the pattern was consistent: the guesses ran high. A 5-gallon salt-glaze crock a model estimated at $600–900 had actual recent comps of $155–275. An advertising jug guessed at $250–400 had comps as low as $31. Not off by a little — off by a multiple, and always in the optimistic direction.
That makes sense once you see it. A guess averages over the memorable examples, and the memorable examples are the rare good ones — the pristine, the signed, the auction-record piece that made the news. Nobody writes an article about the crock that sold for $22 at a Tuesday estate sale. So the guess drifts upward, away from the boring reality where most lots actually live. Comps drag it back down to earth. If your gut number and your comp number disagree, the comps are usually right and your gut is usually high.
When there’s nothing to anchor on
Sometimes you genuinely can’t find comps. Obscure regional pottery. A one-off piece of shop-built equipment. Local ephemera nobody outside three counties has ever cataloged. Thin verticals just don’t have a public sales record to lean on.
When that happens, do not manufacture a number. The temptation is real — a blank estimate feels like a failure, and any figure feels like progress. It isn’t. A fantasy estimate is worse than no estimate: it either scares off the buyer who’d have paid fairly, or it sets a consignor expectation you can’t meet, and now you’re explaining a shortfall on settlement day.
The honest move is to say so. “No estimate; opening bid $X.” Set a modest opening that protects against a giveaway, and let the room find the price. That’s what an open-outcry auction is for — price discovery on things nobody knows the value of. Bidders respect “we’re letting the market decide” far more than they respect a number they can tell you invented. You’re the expert precisely because you know the boundary of what’s knowable.
The rule that ties it together
Distrust a bare number. Whether it comes from your own memory, a price guide, or a piece of software with “AI” on the box — if it can’t show you what it’s based on, treat it as a rumor. A trustworthy estimate shows its work: here are the comparable items, here’s what they sold for, here’s the range that falls out of that. An untrustworthy one just asserts. Once you start asking “based on what?” of every number in your catalog, the fantasy figures get easy to spot.
This is the line Hammerwerks holds on pricing. When you hit a lot you can’t place, the “Research comps” button searches recent hammered lots for visually and textually similar items and hands you the actual comps behind a suggested range — which you accept or override. When it can’t find close matches, it tells you it couldn’t, instead of inventing a number to look smart. No black-box figures, no receipts hidden behind a confident guess. If you want to see how comp-grounded pricing looks in practice, watch the 90-second demo or book a call.