Today’s art market is a global dynamo, comprising an ever-expanding population of buyers and sellers across generations and geographies who trade in, and often passionately compete for, the best art available in our time, hoovering up data and information to get the edge on their rivals along the way.
Bristling with energy, intellect, and cosmopolitan verve, the art market has seized the public imagination in an astonishing manner, but of course, it wasn’t always this way.
There was a time when the art market was a tiny, cloistered, boutique business plied by an intensely committed community of connoisseurs in a few capital cities who catered to the elite largely out of sight, with the actual prices of artworks guarded like state secrets to the degree that it was almost inaccessible to all but an initiated few. So, what changed?
In 1989, Artnet emerged on the scene and brought transparency to the art market through its signature Artnet Price Database, offering its users clear and precise information on the real prices of artworks in a way that changed history. Today’s art industry would be unimaginable without the tool, which covers more than 1,900 auction houses all across the world in every continent, from South America to Asia, to Africa, across art and decorative arts, jewelry, watches, and more.
This month, Artnet is making history again by releasing a major reboot of the Artnet Price Database, incorporating state-of-the-art technology, elegant design, and other bells and whistles to make it even more of an indispensable tool for collectors and art professionals.
To talk about the role of data in the art market, the game-changing power of the price database, and its brave new era, Artnet News’s editor in chief Andrew Goldstein spoke to Albert Neuendorf, Artnet’s chief strategy officer, and Fabian Bocart, Artnet’s chief data scientist.
Andrew Goldstein: Let’s go down memory lane briefly. What was the art market like before Artnet? And how did people know about the price history or comparables of art when they wanted to buy or sell?
Albert Neuendorf: If you go back 30 years, the art market was a much smaller, ultra-niche market that was highly inaccessible, and a lack of available information on pricing was a key reason for that. Most people didn’t have the tools to build the confidence to participate. It required a lot of expertise and connoisseurship.
If you think about Sotheby’s for example, they had to expend a lot of effort in maintaining prices and understanding how to set estimates and reserves for artworks they were offering. Each different location had a data room where team would cut out catalogues from auction houses around the world and index them into books so that they could keep a track of how each artist’s market was developing.
If you went to an art fair, you’d see dealers and collectors that would walk around with pen and paper and record the prices that other dealers were asking, just to get a comparison of what an artwork’s value might be. There was a huge amount of effort that was needed to get an understanding of what something might be worth, and that made the market highly inefficient.
How did somebody make sure that they were aware of the right price to buy an artwork back then?
Fabian Bocart: Historically, dealers will annotate auction catalogs with prices. So, they will go to the sales and write prices of lots in margin of their catalogues. And nowadays, academics collect these catalogues to create databases of past prices and so on. Those dealers would even annotate who was bidding on what, how many bidders, the bid’s increment and so on.
From the beginning, Artnet really has focused on the secondary market prices achieved at auction. Why was this the best way to provide transparency? Why auction prices versus primary market prices directly from galleries?
Albert Neuendorf: Auction prices are a public record, whereas gallery sales are not. The art market is unregulated and galleries and dealers don’t have to disclose their sales or open their books to anyone. We welcome a future where all sales prices are published, but there are a number of roadblocks to getting that done. Any seller’s best advertisement is their most recent result.
In recent years, primary growth driver for the leading auction houses has been private sales. Part of that success is that the auction houses publicly showcase that they’ve sold, for example a Modigliani painting at Sotheby’s for $157 million, and that builds confidence among collectors. That transparency into what auction houses have achieved in the market gets them a lot of clients. A gallery might have achieved a similar deal or often even better, but nobody knows about it. It’s a shame honestly, because dealers are such an important part of the market, they operate with an element of storytelling and expertise that informs what drives the art world. So full transparency in the market would be healthy overall. Today we have full transparency for auction results and that’s already a key driver for the market.
Fabian Bocart: Also, an important advantage of auction prices is that they come from a competitive bidding exercise. So the price that is achieved at auction is supposed to be built from several market participants and by contradiction, primary market actors negotiate over the counter often in one-on-one negotiation, which is a mechanism in which estimate of information plays a more important role in price construction than at auction.
I once asked a famous art dealer how he set his prices and he replied with three words, smoke and mirrors. We all know that the true value of art is in the eye of the beholder. So what makes auction pricing data so useful to actually finding the real price?
Albert Neuendorf: He’s not wrong. The value of an artist’s market or of a particular artwork is certainly subjective. And the art market is driven by passion and that storytelling capacity is key to what gets transactions done, gets people interested in an artwork and drives their desire to own something. Auction results though, allow the market to showcase a precedent. So to see what someone else has paid for a similar artwork allows a potential buyer to generate an idea of comparable value for another. And that transparency into historical transactions provides context and the confidence to bid to the same amount or even higher.
Fabian Bocart: Auction data is not only about price discovery, but also about reputation and transparency of the operations underlying the data. Auction houses tend to offer more transparency when it comes to commission and fees, and also use their reputation to impose standards in the market, for instance, attribution types and so on. And that reflects in the data. Dealers, on the other hand, may be more specialized. Frequent criticism of auction data by dealers is that it does not reflect the very high-end quality of certain artists because their best artworks change hands privately. But over time, even these exceptional works will at some point hit the block at auction.
Okay, so here’s the million-dollar question. How does Artnet obtain all of this auction data?
Albert Neuendorf: Initially, we manually entered all the data from various auction houses, printed catalogs going back to 1983. Those weren’t available digitally at the beginning when we started in 1989, and so we’ve manually entered those into a database. And that’s part of the reason why reproducing a database like Artnet is now impossible. You’d first of all have to find all of those historical printed catalogs, which in many cases don’t exist anymore.
Once auction houses started publishing digital copies of their catalogs, we then switch to scraping the information from their websites. However, we continue to manually check each and every lot before it’s entered in the database because we pride ourselves on the accuracy that is within the price database.
When did it become apparent that Artnet had become an indispensable tool for buyers and sellers?
Albert Neuendorf: The impact that the PDB has had cannot be overstated, but at the beginning it was really difficult. The art market often resists too much change. At the start, we faced some issues with large auction houses that didn’t want us to publish their results in a collated way, they felt like that was taking away their ownership of their results.
It was actually Sotheby’s that initially realized that it was beneficial to showcase their latest results. The best advertisement for them is to show the highest price records that they’ve achieved in recent auctions. And as soon as Sotheby’s adopted the price database, the rest of the auction houses had to follow because Sotheby’s had a competitive advantage.
What are other types of users who benefit from the Artnet price database?
Fabian Bocart: Artnet is primarily a catalyst for professionals in the art world, like dealers, galleries, museums, and so on but we also have a significant number of art collectors who want to benefit from the transparency brought by Artnet. Market participants want to find artworks that have similar characteristics so they can get an idea of the price range for a specific work or even a collection. They may also want to verify the provenance of the work or its past condition in a previous sale and so on. And dealers in particular, they use the database to reassure their clients about prices showing auction results of work comparable to the ones they sell.
How does Artnet maintain the data and keep it all up-to-date reflecting the latest auction results all over the world?
Fabian Bocart: Well, that’s the magic, we have eyes everywhere. So in practice we partner with auction houses to get their data about upcoming sales as soon as possible, and then we collect auction results immediately after the sale in most cases. We had a sophisticated data pipeline, including as we mentioned before, a manual check and other automated processes to fuel the price database. As soon as the results become available. We add on average 3000 auction results per day from these 1900 auction houses. We’ve been adding data to the database for the last 35 years, and nowadays, that’s about 16 million auction results for more than 400,000 artists.
How has Artnet’s approached the data gathering of data science generally evolved over the years?
Fabian Bocart: Well, the very first years were of course manual with the use of faxes and copying catalogs manually into databases. Through the years, the process got more and more automated, with data science and machine learning in particular. I think the years to come are going to see an exponential growth in A.I. application at Arnet, both in our capability to gather data and the services we will offer to our customers.
What kind of AI applications could you see potentially working for Artnet?
Fabian Bocart: We are developing stronger capabilities in image recognition so that in the future, our users will be able to upload photos of artworks and find the results immediately in the Price Database. We will also bring more process data and information to the public, ranking, market trends, statistics about artists, and even the artwork.
What are some things that are already embedded in the price database that even a veteran Artnet user might be surprised to find out when it comes to the capabilities of the data?
Fabian Bocart: Well, there is the online price database, but we also provide all sorts of information about the holistic market at Artnet. For instance, which auction houses offer the best liquidity for any given artist or where will be the best place to achieve the highest prices. Our data also allows us to build relatively accurate models of price prediction in the market.
Our clients are equally hungry for analytics reports. They need to know trends in prices for given artists or their volumes at auctions or the bought-in rate—that is the auction of artworks that do not sell at auction. This helps auction houses to gauge the liquidity of the market.
We internally also use an artist ranking mechanism where we basically score artists based on their popularity at auction, market performance, exhibition history, and many other factors, which allows us to know which artists best represent their movements in the market and so on. And of course, our clients have access to our price indices on demand. These track market performance for certain movements or for given pools of artists. And we generally track market trends across the board.
Okay. So we’ve been talking about the database in almost an abstract scientific way. What are some ways that it can be used to unlock opportunity in the real world for users?
Fabian Bocart: Basically Artnet offers information parity, which prevents people from being taken advantage of. In general, the main use is to search comparable artworks and estimate price points, but this data holds much more valuable information. For example, users can have a real grasp of the evolution of prices over time for specific artists, which can inform market participation.
For instance, at Artnet we have an auction arm, and used the price database with an artist named Richer Hambleton. We spotted a trend in the data and so could predict a fast progression of prices for this artist, so our auction house started to consign more and more of his work, and indeed, the market reacted positively.
And another strategy that dealers use is that they look for works that were recently bought-in at auction, in other words, mispriced artworks that did not sell in recent auctions globally, and then they reach out to these auction houses to try to negotiate a private sale with them at a lower price. Our users also look for low estimates of certain artists to find bargains ahead of upcoming auctions.
Fabian, as a data scientist you spend your life essentially inside the PDB. What are some of the most surprising or unexpected stories you’ve encountered in your time inside the database?
Fabian Bocart: One story I really found fascinating that revealed some unexpected aspects of our database is the saga of two brothers, heirs of a wealthy European dynasty who inherited their father’s collection in the 1990s. The trove was split between them, with artworks from Renaissance’s masters like Velazquez, Donatello, Caravaggio, and many pieces from studios or followers of these artists—extraordinary pieces. In the art world, a piece defined as “studio of the artist” is significantly less expensive than one by the hand of the master. However, in this case, 20 years after the collection was split, one of the brothers got a “studio of” the Renaissance master re-attributed as an original by the artist, and that brother sold the work privately (and discreetly) for nearly $20 million.
He was later divorced from his wife, who out of spite ended up spilling the beans about the sale to her former brother-in-law. Of course, a full-blown brotherly dispute followed with lawyers battling in court about the fairness of the original split of the collection. And so it happens that the only way to figure out whether or not the collection was fairly split 20 years before the private sale was to dig into the Artnet Price Database for a comprehensive review of comparable artworks sold at the time, the condition, the attributions, and so on.
Eventually using a combination of Artnet and other archival data, it was established that the artwork was indeed misattributed, the collection was re-split, and the brother who had sold the piece had to compensate his sibling.
Artnet is essentially an art market time machine itself. We offer snapshots of the market decades ago, and anyone can use it to navigate history both from the content it offers, but also history of the market environment in which it operates at the time of these auctions.
Now Artnet has relaunched the price database and this is a big moment for the company. That’s also a very meaningful moment for the industry at large. So what is new? What’s new under the hood? What’s new in terms of the interface for the price database?
Albert Neuendorf: We’ve built an entirely new user experience, which makes navigating the price database much easier. It’s built on a mobile friendly design, so it’s more accessible if you’re at an art fair or on the auction floor for example. There’s also vast new search capabilities allowing you to search by free text, which opens up opportunities for looking at artworks that have particular colors or that feature particular objects, or you can see the value still lifes overall.
It also has responsive and intelligent analytics. So as you search for particular criteria, you can see what those results mean in the broader context of an artist’s market. Another important aspect is that it’s built on an entirely new API first data architecture, and that allows us to not only build out intelligent analytics, which will benefit the user experience on Artnet, but also allows us to engage in much more meaningful data partnerships with clients from different industries like hedge funds or wealth management companies.
And so when it comes to Artnet’s users and the market at large, what kind of impact do you think this could potentially have? What are some beneficial outcomes that the new interface and the new data capabilities could actually bring to the market?
Albert Neuendorf: In the past, the Price Database has been a tool which is incredibly useful if you know what you’re looking for. The fact that we now have a new data architecture and can really build out intelligent A.I.-driven tools and analytics based off of that means that there’s much more opportunity to add value and allow users to get insight and contextual information that they might not have been able to before.
If you look at the art market overall, although it’s grown exponentially over the last 30 years, it remains an illiquid market. There’s around $1.7 trillion of fine art assets held privately around the world, and the market value overall is around $65 billion. So only around four percent of the total value of artists traded per year.
There’s a lot of opportunity to continue to make the art market more efficient, and that’s core to what art and it’s been trying to do since it was founded 30 years ago. We believe that an efficient market will help participants at every level.
We’ve seen in survey after survey that the greatest impediment for a new buyer to actually pull the trigger and buy a piece of art is a lack of trust and confidence in whether the price is fair or not. And that pricing opacity or that pricing sensitivity is actually a huge opportunity if you’re able to unlock that to create new pools of buyers. Where does Artnet’s data operation go from here?
Fabian Bocart: We plan to be much more assertive in our market predictions and find the future online tools to show what the data indicates when it comes to future prices and trends and price formation in the market in general. So we want to bring more transparency to the market, possibly in a predictive way and in general, we like to be at the forefront of research in data science and machine learning for everything related to the art market. So expect new tools coming soon and more transparency for the auction market.