Chrono24 Has a Timing Problem
One dealer listed a watch on his own site at 03:56, then on Chrono24 at 06:42 - but priced 6% higher. A friend of mine built something that watches, uh, watches š
Does Chrono24 (C24) sell watches? I think thatās a pretty interesting question if you take more than a second to think about it. They have maybe 500k listings, thousands of professional dealers, even more private sellers. They enjoy upwards of 20-25 million site visits a month, offer an escrow service to make people feel safe, have a decent mobile app, and, by and large, offer a āquickā reference price which people now know to ādiscountā appropriatelyā¦ but hey, itās perhaps the most ubiquitous watch-world reference source I know.
But still, Iāll ask again: do they sell watches?
Well, aside from the few listings which are indeed āin houseā, I think theyād more accurately be described as facilitators than āsellersā. I also think they normalise our habit of window shopping for watches, and reduce friction for any potential purchase. So thatās not nothing, but ultimately, most listings arenāt their own, and the vast majority of the stock they display is scattered all over the world. As a marketplace, they provide an easy place for people to search for watches, and they offer a series of assurances and guarantees which help buyers feel a little safer when theyāre wiring considerable sums of money to random strangers all over the world. And clearly, this is a decent business so farā¦ but itās interesting to think about C24ās major asset being āthe place people search firstā because itās not actually an asset you can guarantee ownership on.
This is a large and successful business, but itās based on peopleās behaviour more than any reliable āmoatā (such as proprietary IP or defensible products). The scary thing about behaviour, for a business like C24, is that behaviours can be changed when people have a good enough reason to do so.
We will get to that shortly, but first, letās run through how a typical dealer-sourced watch gets onto C24. The dealer will buy it, via a trade-in, or auction, or just a direct buy-in. Then they will take photos, write about it, and usually list it on their own site. I have yet to find a dealer who does not have their own site and relies on C24 exclusively; Iām sure they exist, but there canāt be too many of them because it costs so little to have your own site, it makes no sense not to. Anyway, once itās listed on their own site, they tend to syndicate it to C24 where it is seen by many, many more people.
That order of operations is quite logical, because it makes the most sense for a dealer. Listing on their own site is free, so they do this first, and then C24 costs them more, so they do this next. C24 calculates a transaction fee dynamically on a per-sale basis as far as Iām aware, and I donāt know if thereās a published formula - but I read somewhere that itās around 6.9% for watches between $5-10k, and professional sellers have reported paying between 9-12% on top of a monthly listing āpackageā. This is why the listing on C24 will often be at a higher price, since the dealer has to eat a bunch of fees which C24 is charging for the eyeballs it shows the listing to.
The point here, is that C24 will almost always be the second listing of the same watch on the internet, and it will often be at a higher price than the first listing of that exact watch. Sure, there may be some exceptions, but as a general rule, if you look at the incentives, the outcome should not be surprising1. I donāt think Iāve ever seen this highlighted quite so starkly before, but this is a wild situation, and now I will show you why.
On 17 July, a German dealer called Zeitauktion listed a Grand Seiko SBGR307 on its own website at 03:56 in the morning, at a price of ā¬2,950 - this was about $3,383 on that particular day. The exact same watch from the same dealer then appeared on C24 at 06:42, but there, it was priced at $3,585.
So in under 3 hours, the same watch from the same seller turned up on the worldās biggest watch marketplace for ~6% more. Now you might be wondering how the fvck I know what a German dealerās website looked like at four in the morningā¦ well, a friend of mine built something that was watching the entire time, and he calls it the monster.2
The monster
So let me tell you about the monster; I suppose if you build something you get to call it whatever you like (I will keep him anonymous at his own request), so that explains the name. My anonymous friend has a background working with large corporate systems over decades. He has been deep into the vintage watch market for just as long, and he canāt write a single line of code himself.
Back in March, he bought a Mac Studio for a few grand and decided to start experimenting with an agent. I looked back through our chats, and it seems like he spent a fortnight fighting a losing battle. About halfway through the month he gave in and paid for a Claude subscription, and things went nuts from there. By June, he had a system which was getting close to 85k lines of code, none of which he had written himself! As of today this is now over 300k lines.
The monster has been created to ingest the internet of watches. It currently has a couple of million records, and covers 185 watch and clock dealers from Japan, the US, Europe, Singapore and the Gulf; it also has about 200 more in an automated queue and these are being ingested daily (up to 5 per day). It also contains auction archives dating back to around 1989 for some, and these records are even linked (automatically) across various auction sales based on case serials so you can trace a specific watch through its historical appearances at auction.
It also tags Rolex Certified Pre-Owned stock wherever it can find any, which allows for CPO listings to also come up in the same searchable pool as everything else - that CPO project went from āI wonder if anyone does thisā to more than 60 CPO retailers worldwide in under 4 hours.
Just for fun, he added a handbag feed too, which now covers HermĆØs and Chanel and LVā¦ this came about one day while we were chatting, and he wondered aloud why this couldnāt work for other luxury goods which have resale markets. Within a day the system had ingested over 30k live listings, and itās now at 250k records with about 50k live. He doesnāt plan to add any more bag dealers for now, but his wife did ask to be sent the LV and Chanel alertsĀ š¤£.
It gets betterā¦ because all the sources are checked every minute! Last month his dashboard showed there had been over 110k polls over 24 hours, and as of this week itās about 187k; letās call it 180k for simplicity. To put this in perspective, 180k checks per day is about 2 checks every second. If you wanted to do this manually, maybe it would take 10 seconds per page? So that would be 1.8 million seconds of work a day - or about 500 hours - which would need about 21 people working literally all day and night, or approx 60 people working 8 hour shifts.
Based on his own interests, new listings are sent to targeted Telegram channels; he has various setups which include dealer-specific channels, brand channels, and even model-specific channels.
Every time there is a new listing, the system automatically shows comparable recent sales to give you an idea of how good a deal this one might be.
Per the image above, the system also keeps an eye out for price drops. For example, in the past week it logged just over 4,000 price changes, which doesnāt include currency fluctuations and weird no-op updates which pop up when a dealer re-saves a page but doesnāt actually edit anything.
This all costs about $200 per month to run, and everything is executed locally using solar power. Now make no mistake, AI wrote the code for this monster, which is of course deterministicā¦ but I donāt think AI can āagenticallyā do any of this on its ownā¦ and if you did manage to get this to work agentically, it would cost you a fortune in tokens.
Okā¦ that doesnāt sound like a big deal
The Zeitauktion story I shared earlier is a neat example, because of the clean mathsā¦ 202 / 3,383 is about 6%, which is, give or take, what it costs a dealer to sell through C24. Now if you had been monitoring that dealerās website youād have got the watch 6% cheaper, and maybe even cheaper than that if youād bothered to haggle - and if you did that, the C24 listing would never have existed as far as you were concerned.
You might think that one watch on one morning is just a convenient anecdote, so I asked him whether this was typical... turns out, itās actually conservative.
As an example, for one dealer he monitors, the average price is around $30,000, and they have roughly 400 watches live at any time - average turnover is 23 days in stock but median is only five days. So that dealer, on average, lists their stuff on C24 only 14 hours after the same watch has gone live on their own websiteā¦ and half the time their stock is gone within five days. As my friend explained, cross-posting to C24ās marketplace is basically a batch job which someone does later in the day or the next morning.
And in case you want to check out an example for yourself, hereās one that was still live when I wrote this. Gallery of Dreams put a Breguet Classique Chronograph 5237 on their own website on 5 August at 22:00, priced at ā¬24,990 (around $28,826). The same watch only appeared on their Chrono24 storefront on 7 August at 06:56 - and of course, it was listed higher, at $30,729. Again, we have a huge head start (nearly 33 hours), and a saving of $1,903 (~ 6.2%). Hereās the C24 listing FYI - both listings were live at the time of writing, so you can go and look.
But here are two further examples with different dynamics.
On 13 July S.Song Watches published a Daniel Roth watch for sale. The monster saw it within 30 seconds after it went live - so fast, that all the dealerās photos were not even done being uploaded. On this original listing, the price was $13,500. Then, about half an hour later, the dealer corrected the price to $25,000. So for about half an hour the dealer had an 11.5k mistake on the open internet, and the monster had caught this within 30 seconds.
Here are couple of examplesā¦ these two Journes appeared and disappeared from the market very quickly. Now, other commercial platforms which offer market tracking services sometimes describe themselves as āā¦ comprehensive aggregators of secondary and auction-market transactionsā, who draw on āxxx dealers and xxx auction housesā with āthe broadest coverage available in the market.ā
As of writing, neither of the above Journes he watched appear and disappear made it into any other platformās record (per images above). I guess you will never be able to capture what you were not able to see in the first place. Honestly, the āalways watchingā aspect of this monster is really difficult to overstate; I have seen examples of how a watch is listed, captured, and gone so fast, that it might even seem like an error.
Incidentally, the impetus for all this was a particular Lange watch. A dealer in Japan listed one at an amazing price, and my friend found out about it the same day (via word of mouth) - point being, he found out after somebody else had already bought it. The very next day he started building the dealer monitoring system... I guess grudges are pretty good motivation!
Should C24 be worried?
In the beginning my friend assumed that there was no point trying to compete with C24 because, you know, why bother? They are huge and ubiquitous, and heās a guy with a solar-powered Mac. But then, a few days ago, it dawned on him that he doesnāt need any C24 listings, and I already explained why at the top of this essay. For every dealer who lists on C24, the monsterās alerts tend to pop up sooner, and often at a lower price.
Thatās all it takes, right? The monster does not need to compete with the sheer volume of C24 āinventoryā because it can front-run ALL the inventory on C24 and simply use C24 as a directory of āwho to watchā until all dealers are ingested over time. In fact, heās already automated that, tooā¦ so basically, if there is ever a listing in any of his existing priority channels from a dealer whose own website he isnāt yet āactively monitoringā, the monster will go and find that dealerās own website, and add it to his overall āmonitored setā.
My friend actually described how the monster goes about this as if it was thinking aloudā¦ and Iām going to quote this because itās easier to follow:
āOh, you listed a Lange on C24? And I donāt have you in the system yet? Letās find your website; then Iāll score it based on what you have that is of interest; then Iāll promote you into the pipeline accordingly; then when itās your turn Iāll first grab your entire history, then monitor you for all new listings every minute going forward; with automated alerts firing for watches of interest that you list. Oh, and Iāll regularly sweep your site to capture sales and price changes too.ā
Over time, if he keeps this going, the monster will get closer and closer to being all-seeing, across all dealers. It just needs time to find them allā¦ but this is all automated anyway so itās just a matter of time.
More importantly, I also think C24 canāt really copy this. The reason is that it will cannibalise their own business model. I mean, think about it; this āupgradeā, if they chose to copy it, would mean a little line on every listing saying āoh, by the way, this exact watch is on the dealerās own site right now, for 6% lessā. This of course means they would lose the sale altogether.
Also, imagine what would happen if dealers themselves ask the monster to monitor them - i.e. voluntarily. Why wouldnāt they? If there is a better channel out there where buyers might see your stock first, and it costs you nothing, why would you not want to be added? If that happens, and if access to the monster is open to usersā searching it and subscribing to its alerts, then the monster could become the listings feed which dealers prefer. And if that happens, what would C24 be good for?
I donāt know what my friend plans to do with this monster, but I would argue that if he opened it up to users to experiment and test it out, C24 would likely need to have an emergency board meeting and approach him for a buyout. I say this because if it ingests all dealersā listings on C24 as well as everywhere else, then this monster will essentially be a live view of the entire global online watch market, along with the tools and plugins to offer you comparative data which includes recent auctions around the world.
But maybe not?
Ok, so this might just sound like a āfaster data feedā right now, because this isnāt really a marketplace; so being the first to see a watch available for sale, one might say, is not the same as having the systems in place to complete a sale. Thatās an accurate and fair depiction; C24 does have a lot of bells and whistles like escrow, buyer protection, authentication, dispute resolution, payment rails, an app, and of course the superb reputation of being the go-to place when people want to search for a watch. So you might look at this and think ah, well this is just a man in a basement with a notification firehoseā¦ the only thing he appears to have solved is discovery but none of the rest.
So I raised this with him, and he explained that he already has most of the required pieces of that jigsaw. As an example, if he gets an alert for a watch in an auction via Telegram, he can double-tap it to ālikeā it, and the system will then ping him an hour before the auction ends, and then once more when thereās 15 minutes to go. He has the ability to extend that functionality to respond to that Telegram message with a maximum bid, and the system will actually bid on his behalf. It works the same way with dealer stock, where he could reply to an alert with a specific offer, and the system will email the dealer directly or, if the dealer already has a storefront, it can also complete the purchase on its own. The way he uses this right now doesnāt actually need this, but he assured me this was ātrivialā to add.
So it turns out, heās actually built a lot more than discovery here; there is no need to have a checkout system in his back end because dealers themselves already have this. In essence, if the dealer does have an online store, you want to be the first to buy from them, and if they donāt have a storefront, you want to be the first to see the watch, so you can contact them and secure it. In both cases, all youād ever need to do is reply to a Telegram messageā¦ which is pretty easy.
And so... the next question in my mind is to ask where the value is. What is escrow protecting anyway? Ultimately, if you found a watch on a dealer website, at a cheaper price, youāre just as likely to buy directly from the dealer are you not?
I know many large sites prohibit this sort of automated collection, and anti-bot measures get better all the timeā¦ I am well aware that āeveryone does itā probably wonāt work as a defence, so I donāt know how that will play out.
Another thing about the monster is that its value comes from attaching a monetary value to inefficiency - but like any arbitrage opportunities, the market works hard to reduce the arbitrage value to zero. If dealers know that fat-fingering a price will be punished in 30 seconds, they will find a way to put in checks and balances to ensure this never happens again.
But all this fat-finger stuff isnāt really the edge at all. I reckon the edge in this equation comes from the order of operations because thatās what creates an incentive. A dealer will usually list a watch on their own site first because itās free to do so, and theyāll list on C24 afterwards because C24 takes a cut.
So while you may be able to teach a dealer not to make typos, theyāre always going to understand basic maths; and as long as C24 takes a cut at the end, they will always come second. This, I think, leaves C24 with a pretty shitty option as far as I can seeā¦ to stop charging for transactions, and start charging for placement only. If they become the place which shows every dealerās stock the moment it goes live, and only bill dealers for visibility and leave their sales alone, maybe this all goes away. I also realise thatās a different business to the one they are running today, and, you know, turkeys donāt generally vote for Christmas.
Archive knowledge
The live alerts and site monitoring are the flashy parts of the monster, and you could argue that someone else could copy that if they were dedicated enough. But what they canāt build as easily is an archive of the past which powers the monster. There are a couple of million records of what prices watches were listed for, and what they sold for, and this includes a huge chunk of dealer data which was captured at the time of sale. If you were not monitoring constantly, the number is gone forever. And so, every day that this monster exists, it becomes even more powerful, and less reproducible.
Itās fair to say that the secondary watch market has never been efficient, and it may never be efficient - but what was never as clear to me was how much of that inefficiency was due to the spread of information and how much was related to inventory, condition, provenance, taste, and all the stuff we tell ourselves a price is made out of. Based on this, Iād say a fair piece of it was just latencyā¦ the delay between when something is live on the general internet, versus when it goes live on C24.
Hang on, what does this mean for me?
Think about what most of us do when weāre wondering about a watch; I know I almost always pull up C24 to check where the price is, ballpark. C24 is the reference point; but it isnāt neutral because itās always the second listing with their cut baked into it i.e., it is always skewed against you.
In a recent EveryWatch report they had a median secondary market sale at $5,900 across 795k transactions; 6% of $5,900 is $354, so whenever someone buys on the marketplace instead of the dealerās site (which had the watch a few hours earlier) thatās roughly $354 which didnāt need to leave their account. I have no idea how many of those 795k deals went through a marketplace, but Iād be amazed if it was a small number.
So, coming back to where I started, C24ās biggest asset is their ubiquity and their presence as the default place to look for watches. Being at the forefront of everyoneās mind is extremely valuable, and that only lasts as long as there is nothing else for people to think about.
As a thought experimentā¦3 suppose there was some platform which showed you a dealerās new stock within 60 seconds of it going live, before it reached C24, and usually a few percent below what itāll cost on C24ā¦ and say this was covering a growing chunk of all dealers worldwide, with recent comparable sales and auction results also attached to the alert so you know whether itās a good price or not. And letās say, this āreachā was growing every day, and that it could plausibly cover any online-listed watch on earth eventually. Where would you look first: on this platform, or on C24?
Finally, and specifically for todayās SDC post, he produced a sample report for Grand Seiko over the last 24-hours (see pdf below) - now imagine this, but for every brand, all the time.
ā
I know my friend reads SDC comments, so feel free to chime in with your thoughtsā¦ and if youāre from C24, and want to speak with him, let me know and Iāll connect you. I asked him if I could get a commission if he sells this, and he said he will think about it. If that happens, I will drop SDC prices š4
Just for funsies
This add-on is mostly for fun, but it also has real-world relevance if you use his tool a lot. Itās called Tickr which I think is a pun on Tinder & Grindr? Haha. Anyway, how it works if you select some slice of the data - in the demo shown below it was the Phillips Toki Auction. Then you just swipe left or right, or double tap, just like you would on any dating app. The system which has a ton of metadata (including images) will take note of what you like and dislike, and then build up an understanding of your taste (or whatever youāre training it forā¦ like, say youāre a dealer, and you have a ātasteā you want it to monitor for you because thatās what you want to source).
Right, so once it āknowsā what you like, whenever there are new auctions that go live anywhere in the world, apart from finding these auctions and ingesting the data, it will also score every lot against your preset taste and send you a pdf of the top 20 lots you might like. Right now itās just a proof of concept which needs some tuning to be reliable, but I thought that was pretty fun. Hereās how it looks in action:
Afterthoughts
This draft took about a week to put together, and the numbers moved as I progressedā¦ so I asked for a quick update and as of today, there are 2.2m watches in the archive, 196 dealers now integrated, 230k live listings being monitored (with 4,800 new listings in the last 24 hours alone), 5,500 sales and 7,800 price changes recorded in the last week, and 197k searches in the last 24 hours. Over the same 24 hours, it also pushed 1,566 watches worth about $22m into various Telegram channels, and my friend pointed out that this was just a subset of things the monster can see. He stopped creating individual channels for every dealer because heās only keen to highlight what he cares about. Point being, that $22m is only what heās chosen to filter out, and the firehose is much, much larger.
Footnotes
As it happens, my friend tells me there are quite a few dealers who list at exactly the same price on C24 as their own siteā¦ this is odd given what the marketplace costs them - but I donāt have a good explanation for this. I donāt think that changes anything because even if they list on C24 at the same price, the listing still goes up on their own site first, and that time delay is the key thing to keep in mind. Timing is the product as far as I can tell, and price variance is a nice bonus on top.
He has since come up with a better name, which came to him in a Miami hotel room, after which he reverse-engineered it into an acronym.
THOMAS: The Horologically Omniscient Monumental Alerting Systemā¦ named, he tells me, after the little engine that could. š
Ok I guess it isnāt really hypothetical, because the thing exists, it is actually running as I type this, and real watches are being bought off the back of its alerts; it just isnāt open to you and me lol!
Jokes aside, this isnāt a paid advertorialā¦ I just thought this system was fascinating, and worth sharing with watch nerds who might get excited about what the future holds.