ART: DATA, DATA, WHERE ARE YOU?
Big data is the buzz word of 2017 but the art industry has been slow to respond to it. We ask why, and its future in a market where players have thrived on its opacity.
Big data is the buzz word of 2017, but the art industry has been slow to respond to it. We ask why, and its future in an industry where players have thrived on its opacity
The art world is on a data scientist-buying spree. While data has always been precious, auction houses and dealers operating at the top-end of the market are leveraging quantitative market analyses, algorithms and high frequency price indices to gain a competitive edge in an increasingly crowded market place. The spate of acquisitions over the past year attests to this. Following Sotheby’s acquisition of The Mei Moses Art Indices (now Sotheby’s Mei Moses) last October, online giant artnet poached Fabian Bocart, co-founder of Tutela Capital, to head artnet Analytics a month later. Earlier this year, Artsy acquired Artadvisor, the data science start-up founded by AI expert Hugo Liu.
If the above is of any indication, the big data arms race is on, but the question remains: in an opaque industry where people guard their privacy like a polar bear guards its cubs, what could be big data’s biggest potential? And what are its limits?
To answer that, we first need to define what big data is. According to Daniel Doubrovkine, chief technical officer at Artsy, big data is "statistically significant information that can yield real insights with the help of algorithms and machines.” Thus, while a century-old auction house might have been meticulously noting down the names of their clients, their habits and details of their purchases since day one, this information – however rich – cannot automatically be classified as big data.
To achieve effective data mining, a sizable amount of data is required. In this regard, it does appear that the art industry is ripe for big data disruption. In 2005, the TEFAF Art Market Report put the value of the art market at US$35.9 billion; in 2015, that number has risen to US$63.76, translating to a growth of 77.6 per cent.
If we pan in onto the online art market – where data is argubly, more easily trackable – the numbers are even more encouraging. Online sales are at 3.75 billion in 2016, up 15 per cent from the previous year, according to the Hiscox Online Art Trade Report 2017. Online sales now take up a 8.4 per cent share of the overall art market, compared with 7.5 per cent in 2015.
Art has become an alternative asset class of its own but there are still those who believe that it should ‘rise above’ monetary concerns, according to Roman Kraussl, a finance professor at the University of Luxembourg.
“Pure finance considerations are still being treated as evil, as non-arty, as destroying the pure, the true value of art,” Kraussl, a research fellow at the Emory Center for Alternative Investments, notes. “It has nothing to do with the fact that we do not have enough data available. I am working with a database of over 10,000,000 auction records, going back to 1900.”
It’s also no secret that the art industry thrives on its very opacity. Even though public auction records are well, public, the dealer segment of the market remains private. Dealers are under no legal obligation to disclose the amount they profited from a sale, and it has long been a mantra of the shrewdest in the trade to ‘buy low, sell high’. Added to this the reality that more and more collectors are looking to the discretion that dealer sales offer. According to the 2017 TEFAF report, dealer sales had a 20 per cent growth, from 23.25 billion in 2015 to 27.9 billion in 2016. Meanwhile, auction sales are down, from 21.08 billion in 2015 to 16.9 billion in 2016.
“In such a market environment, some people do make big profits, akin to an insider IPO market. My research, my analyses, my indices, based on big data brings transparency to the market, and not everyone is happy about that,” notes Kraussl.
“Privacy - and the absence of regulation - is what makes the mechanisms of the art market very different from those for crude oil, stocks or property. There is no central registry where sales are recorded,” adds David Hopper, founder of Art Logic,
And it doesn’t look like there will be one soon. In a negotiation-based marketplace, where one dealer might be able to sell a work for US$10,000, and a less shrewd one could also sell it for US$ 9,000, no gallerist would want to disclose the final selling price for fear that the buyer would demand a refund for the difference.
“Changes in technology have made data analysis and prediction more commonplace in the art market, yet demand for large amounts of data upon which predictions can be based will outstrip supply,” adds an ambivalent Hopper.
As expected, Artsy, vocal about its goal to ‘democratise’ art through technology since its founding in 2009, remains the most optimistic about the disruptive potential of big data. “For the first time we have data from both the primary and the secondary markets that can power machine learning models and some early AI,” Doubrovbine notes.
In an industry where there are huge data gaps, Doubrovkine also argues that data breadth is more important than data depth. “markets typically behave consistently at scale and with broad enough data we can infer the overall system.”
POWER OF BIG DATA
Different companies have different ways of leveraging data, and at Artsy, that comes in the form of The Art Genome Project (TAGP). Devised by six art historians, one thing that TAGP does is to calculate the importance of an artwork within the artist’s body of work. Taking into account varied factors, including the size of the work, the artist’s popularity, and whether the work is in a museum or private collector’s
hand, a iconity score is generated. It’s interesting (and refreshing) to see that even in the age of social media, the art world still abide by certain traditions – while a work might score 2 points for being in a museum, it would only score 0.5 points if its artist has 10,000-plus followers on social media.
Prior to joining artnet, Bocart of Tutela Capital was perhaps most famous for putting forth the Fair Value Measure, or IFRS 13, an internationally recognised valuation for works of art that enables entities like banks and hedge funds to leverage art while conforming to regulatory demands for disclosure. In an interview with The Art Newspaper, the former trader stated his ambition to correct the art market’s major discrepany: while the 2015 TEFAF report values the art market at US$63.76 billion, he estimates the potential value to be somewhere nearer $1 trillion. “What we want to do is to provide the right information to stimulate that market, to help push margins lower and to give banks, hedge funds and investors access to a market [that is] in a similar position as the real estate market was in the 1980 and 90s."
While data collecting has been at the heart of Christie's business strategy since Day One, it is only in the last five to six years has the 250-year-old started doing large scale data-mining, according to Rebecca Wei, Christie's Asia President. She remains cautiously optimistic about the role of big data in the auction market, “big data matters when you have to decide whether to move online more aggressively, whether we need to close down part of the company or where we should enter a specific category, but when it comes to our day-to-day operations, small data is still much more important.” After all, the auction business is still considered a niche market, where one client’s purchases could represent 10 per cent of an auction’s annual sales. Data of individual clients are still more important. “We aren’t in the FMCG industry. We are selling a US$20 million diamond, so we need to know which client has the funds.”
Wei’s cautious optimism also hints at where big data will best effect change in the market. Compared to the top-end of the market, there is less public information about the middle to lower market, but it is much larger, volume-wise, and Doubrovbine points out that they’re seeing lot of repeatable and consistent patterns for art that sells under US$10,000. Quant analyses also appear particularly beneficial to the contemporary market, where the risk is largest, according to Kraussl.
Some have suggested that blockchain technologies might be a solution to the industry’s opacity – it could
provide security without compromising on the privacy of individual clients. “Blockchain technology has a lot to offer to any industry that deals with provenance, supply chain, chain of ownership, authenticities,” says Jehan Chu, art advisor turned investor of blockchain companies.
“If art is on the blockchain, it wouldn’t just be provenance we’re talking about, we can also add user experience, derivative products such as insurance,’ notes Chu. While a sophiscated system is required to accurately marry actual objects, be it a painting, a sculpture, an installation, with the digital identity on the blockchain. Chu says the biggest challenge is adoption. “The technology itself is very sound, but how do you get everyone from student artist who are doing their own art show, to major secondary market players to get onto it?” Eleesa Dadiani of the Uk-based Dadini Fine Art might be a trail blazer in this regard. On June 29, she announced that buyers will be able to use any of the six leading crytocurrencies on the market, incluidng Bitcoin and Ethereum, to buy art from Cork Street gallery.
DEATH OF THE ART ADVISOR?
The rise of AI is often linked to the death of the human advisor. Those we talked all emphasised that a big data-driven world would better, not eliminate the role of the advisor. “We can anticipate…a shift in the role of advisors. Collecting and consuming data will become a larger part of their routine, although their strengths will remain their knowledge of the work and the artist, connections, and relationships,” notes Hopper.
And indeed, while hard numbers might appear more objective next to their live-in the flesh human advisors, big data can easily be manipulated. “We all know that it is possible to generate any wishedfor outcome,” quips Kraussl. “TEFAF and Art Basel should set up a website, indicating the galleries, and people they approached. Otherwise these studies could be misleading.”
An overreliance on big data and algorithms might also lead to a ‘narrowing’ of the market. “[Purchases can be influenced by] notable press mentions, critical exposure and a certain volume of secondary sales. This distortion will consolidate established artists at the expense of emerging artists and favour galleries that represent established artists, not least because of the growing cadre of buyers who see art as an investment,” Hooper explains. Human advisors might serve to correct this, as Kraussl says, “the good advisors will survive, those who know what the art is really about, instead of those only know how to spot a bargaining deal.”
LIMITS OF BIG DATA
For all its power, big data comes with its own set of limits. In an industry where collectors are often driven by very personal desires, it is impossible to predict what the next Jean-michel Basquiat, Claude Monet or Jack Pollack will sell for. Few would have predicted that a 1982 painting by Basquiat would soar way above its US$60 million estimate to reach a record-breaking US$110.5 million at a Sotheby’s sale, nor that Zao Wouki’s 29.09.64 would punch three times its US$6 million estimate to score Us$19.7million at Christie’s.
“It’s emotional, it’s impulsive, and you never know how two people will be bidding against each other. More than 10 per cent of it [art buying behavior] is irrational, because love is. In the case of Basquiat and Zao Wou-ki, a single jump in price altered the whole curve,” notes Wei.
“IN SUCH A MARKET ENVIRONMENT, SOME PEOPLE DO MAKE BIG PROFITS, AKIN TO AN INSIDER IPO MARKET. MY RESEARCH, MY ANALYSES, MY INDICES, BASED ON BIG DATA, BRINGS TRANSPARENCY TO THE MARKET, AND NOT EVERYONE IS HAPPY ABOUT THAT”
– Roman Kraussl, finance professor
Zaowou-ki’s 29.09.64went underthe hammerfor Us$19.7million atchristie’s Hongkong, morethan threetimesits Us$6million estimate.