Category: Uncategorized

Auctioneers ask the same question every time: “this AI thing — does it actually save us time, or is it just another tool to learn?” Here’s the maths from real UK rooms running Snaplot for the first three months.

The baseline: how long does cataloguing actually take?

Time-and-motion across six rooms (general antiques, industrial, estate clearance, vehicles, fine art, modern design) gave us:

Lot type	Average minutes per lot (manual)
General antiques (Victorian furniture, ceramics, silver)	6–8
Box of “bits” (mixed-content lot)	3–5
Industrial / commercial (machinery, plant)	5–7
Vehicles (cars, vans)	10–14
Fine art (paintings, sculpture)	12–18
Modern design (Ercol, mid-century)	4–6

Call it 7 minutes per lot as a rough average. A 200-lot weekly sale = ~23 hours of cataloguing. A full-time job for almost three days, every week.

What changed with Snaplot

Same six rooms, same lot types, three months in:

Lot type	Manual minutes	With Snaplot (photograph + edit)	Saving
General antiques	6–8	1.5–2.5	~70%
Box of bits	3–5	1–1.5	~70%
Industrial	5–7	1–2	~75%
Vehicles	10–14	2.5–4	~75%
Fine art	12–18	4–8	~55%
Modern design	4–6	1–1.5	~75%

What the time savings actually translate to

The 200-lot sale at 7 min/lot → 23 hours. With Snaplot, ~6 hours. That’s not “I have an extra afternoon”. It’s “I have an extra 2.5 days every week to do everything else the auction house needs”.

What does that look like in practice across the six rooms tested?

• One room added a sale per month. The cataloguing time freed up enabled an extra timed online sale, no headcount change.
• One stopped using a freelance cataloguer. £400/month saved.
• One re-deployed cataloguing time to consignment chasing. Bigger sales, same effort.
• One actually went home at 6pm. Their words: “I haven’t had a Wednesday evening in fifteen years.”

Where Snaplot doesn’t save time

Honest disclosure — the time savings are smaller in two situations:

1. High-value specialist lots where the cataloguer needs to research provenance, attribution, exhibition history. The AI gets the routine done; the specialist still spends the hours on the bits that need expertise. Saving here is more like 30–40%.
2. The first week. You spend time learning what the AI does well and what it doesn’t, building a quick mental model of when to trust the draft and when to rewrite. Most rooms hit steady-state by the end of week two.

The hidden saving: catalogue consistency

Beyond the time on the clock, the AI writes in one voice across the catalogue. No “this lot was done by Sarah on Tuesday and reads like Sarah, this one was done by Mark on Friday and reads like Mark”. Repeat bidders notice consistency. So do post-sale buyer queries — fewer of them, because the descriptions follow the same structure every time.

Try it on your next 100 lots

Free 100-lot trial, no card required. Run a parallel test on a sale — Snaplot for 50 lots, your existing process for 50 lots. Compare time spent, output quality, post-sale queries. The maths becomes obvious within one sale.

Start your free trial →

The difference between a lot description that gets ignored and one that draws bids isn’t florid prose — it’s discipline. Bidders skim. They want enough information to decide quickly, with the assurance that they’re not going to be stung at collection. Here’s a working framework, distilled from what consistently works in UK rooms.

1. Start with what it is

The first line should answer “what am I bidding on?” without ambiguity. Not “a lovely Victorian piece” — that’s marketing waffle. “A George III mahogany bow-fronted chest of drawers, c.1790” tells the bidder the era, the wood, the form, the date. They can decide in two seconds whether to keep reading.

2. Add the distinguishing details

One sentence, the things that aren’t obvious from the photo:

• Maker (if known and confidently attributed)
• Period or date range
• Materials, including secondary woods or platings
• Construction features that signal quality (cock-beaded drawers, dovetail joinery, original handles)
• Provenance, if you have it

3. Dimensions go next

Always. “H 89cm x W 110cm x D 53cm.” Bidders have walls and rooms to fit things into. Missing dimensions cost you bids from anyone who isn’t already in the room.

4. Condition — be specific

The condition report is where buyer disputes are won or lost. Be specific, not euphemistic:

• “Some wear” → useless. “Light surface scratches to the top consistent with age; one small chip to the left front foot.” → defensible.
• “Restored” → vague. “Polished surface; later replacement drop handles to the bottom drawer.” → bidders know what they’re buying.
• “As found” → red flag. “Sold as found — drawer linings showing signs of woodworm flight holes, no active infestation observed.” → discloses without alarming.

5. End with what’s missing

If there’s no original key, no maker’s label, no certificate, say so. Bidders assume things are present until told otherwise. Don’t let them be surprised at collection.

6. House voice — pick one and stick to it

Read your last three catalogues and ask: does this read like one person wrote it? Inconsistency is exhausting for repeat bidders. Some houses go formal-conservative, some descriptive-warm, some lean into specialist jargon. Pick a register, write a one-page style guide, brief whoever’s cataloguing.

7. What NOT to do

• Don’t oversell. “Stunning”, “exquisite”, “rare” lose meaning when used on every lot.
• Don’t speculate on attribution. “Possibly by Sheraton” is fine. “By Sheraton (unsigned)” is a buyer dispute waiting to happen.
• Don’t hide damage. The internet has ruined any room that tried this five years ago.
• Don’t translate marketing copy from the consignor. “Family heirloom passed through three generations” is not catalogue copy. “Property of a private collector” is.

The 30-second test

Read your description out loud. If a regular bidder couldn’t decide whether to bid in 30 seconds — too long, too vague, or too padded. Cut.

The role of AI here

An AI cataloguing tool like Snaplot writes to this framework by default — what it is, distinguishing details, dimensions where photographable, condition from photo evidence, what’s not visible. The output is a draft you edit; the framework is enforced from the first try, so what you publish is consistent across the catalogue without having to police every cataloguer’s prose individually.

Try it on 100 lots free →