Statistical Data Analysis

Artfinder: "We are in many ways OkCupid for art"

Artfinder is an online marketplace for original artworks founded by Jonas Almgren in 2013. Its aim is to make it easy for artists and buyers around the world to connect through the “sharing economy” and to disrupt a lagging art market.

In February Artfinder launched a new machine learning algorithm to help provide better daily art recommendations to its users based on technology from ‘graph database’ firm Neo Technology. We catch-up with Almgren to learn more about his vision for the art market and how the new algorithm works in practice.


Can you describe your vision to create a “mass market” for art?

Artfinder is changing the way that art is bought and sold, allowing art lovers to directly connect and buy from artists in over 100 countries around the world. It is a truly global marketplace and, for the first time, affordable original art can be found and acquired with ease and convenience regardless of where you are located.

Our mission is to bring the world of art to everyone, to create a mass market without any mass production, a market in which artists play the lead role and directly benefit from all the sales.

With hundreds of new artists and thousands of new artworks added every week, we are now growing faster than ever. This is one of the reasons that personalisation has become incredibly important.


How technologically “behind” do you feel the art market is?

The art market is being disrupted, just like Uber disrupted the taxi business. The momentous shift towards the so-called “sharing economy” - in which artists can directly reach a global audience and a global audience can identify their own “winners”, independently of galleries and traditional gatekeepers - will revolutionise the market itself. When a young marketplace like Artfinder can sell 10x more art in a month than 90% of galleries sell in a year it is clear that a very basic shift is taking place.

In your new algorithm, how do you rate art as similar in order to make daily recommendations?

Since art and taste is very subjective, you need to shift the focus from the artwork to the person looking at it. The traditional machine learning approach is based on classifying the artwork along hundreds of dimensions (which can be done partially automatically, and partially manually) and match “similar” artworks based on these characteristics.

However, in our tests, we have found that taste often doesn’t follow the path of “similarity” based on a few hundred criteria, but on minute details that engage the viewer, and pull them in, getting them interested. These details are often not covered by classification criteria, even when rather complex, and so fails to retrieve even the best recommendations.

But if we instead look at human preference and find people with similar taste, we can more accurately recommend artworks that will inspire the viewer. The more overlap we find between two viewers in terms of taste, the more accurately we can suggest more art.

For the future, we will analyse more users’ activity on the site, so that we can fine-tune our preferences map in more subtle ways. This can be especially important for users that don’t like to express their likes in public by “favouriting” art works.

It’s also important to realise that preferences can change based on intent, i.e., when we shop for a gift we might look for something different than when we search for a large artwork for our living room.

And we have not given up on machine learning. There are some interesting, sophisticated, fully-automated image classification techniques that might come in handy, especially when building recommendations for new users who have yet to express their own individual preference.

Art preferences are quite subjective, how can you measure accuracy?

We are able to measure the quality of our recommendations by analysing user behaviour on the site. We [also] look at how many users engage with the recommendations we make and use this data to test different versions of the algorithm.


Do you see as your closest rival site, with this new algorithm acting as a core differentiator? is one of several online sales platforms for art. Our key differentiator, from them and from all other sites, is that we only sell original art - i.e., not posters reproduced from paintings, which is actually mostly what you’ll find on these ‘larger’ sites. Our reasoning is that we’re making it possible for everyone to own original art - art which directly puts money back into the pockets of the artists producing it and open editions of prints reproduced from paintings don’t do that.

A focus on innovative technology has always been front and centre of our business model - on making the experience of being on the site not just seamless, but joyful and exciting for the user.

We are the first site in the art space to deliver a completely personalised home page, which is a big step in a traditionally curator-led, advisory market. It is a new way of shopping for art online.


How many artist applications do you get, do you reject many on the basis of quality?

We’re gaining velocity all of the time. In February we had over 500 applications and of those we accepted around 80%. We do reject on the basis of quality, but this is more about the quality of the application - e.g., poor photography of the artwork and poor artwork descriptions of the artwork - than a value judgement on the artwork itself.


How global are your buyers and artists, are some regions growing faster than others?

Our community of both artists and buyers is very global. Our current split is roughly a third from the UK, a third from the US and a third global, out of which our next biggest market is Australia. We are growing very quickly in the US and will soon see sales there overtake those in the UK, which will be a real milestone for us.


How much has the site grown since 2013, how fast is it continuing to grow?

We launched the marketplace in 2013 with 59 artists and 464 artworks and since then it has grown to over 6,200 artists and 195,000 artworks. With so much art, we surely have something for every buyer, but it’s also easier to drown the buyer in too much choice. We need to get ever more precise. We are in many ways OKCupid for art.


I noticed that you recently partnered with some art suppliers to get your artists discounts on materials. Are you working on other partnerships to bring the whole community together?

Absolutely, one of our recent new appointments is focussing on growing partnerships that support both customers and artists, globally.


What ambitions do you have for Artfinder short medium and long term? And what role will new technology play?

It’s all about making art accessible to all, so we will continue to explore ways to make it even easier to find art you love. This could mean making the technology easier to use, but it often means implementing such sophisticated technologies that no one even realises they are there. And when so much is subjective and taste based, the shopping experience has to be more inspirational than informational.

Our main competitors are eBay and Amazon. This is where most online buyers find art today. But we believe they both present an incredibly weak experience for art buying and will let many prospective customers down even before they have purchased any art. We want to change that. Buying art should be pleasurable and exciting.

Going forward, we will continue to invest in new technology (as seen above), the team (we have doubled in size over the last six months) and in our global presence. Of all regions, the US is growing fastest and we’re thrilled to be ranked #1 in our category for customer satisfaction by TrustPilot in the US.  


« C-Suite Talk Fav Tech: Draven McConville, Klipboard


Crowdfunding Revisited: Mindsense seeks to redefine your inbox »

Recommended for You

Future-proofing the Middle East

Keri Allan looks at the latest trends and technologies

FinancialForce profits from PSA investment

Martin Veitch's inside track on today’s tech trends

Amazon Cloud looms over China: Bezos enters Alibaba home ground

Lewis Page gets down to business across global tech


Do you think your smartphone is making you a workaholic?