» » » Early Evidence Is Often Too Early And Not Really Evidence

Early Evidence Is Often Too Early And Not Really Evidence

I wanted to share an interesting article on market experiments.  While the article is geared towards start-ups in product development, the ideas about choosing marketing experiments carefully is pertinent. So is the thought that market data that comes in early needs to be recognized as early data and not definitive.

How does this relate to Art?

If you look at large art websites and POD sites that selectively market their artists, “popularity” is a key feature used in search ordering and promotion of pieces.  As artists, collectors and aficionados, we have all experienced art that is easy to like and quickly popular.  But are those our favorite pieces?  Are those the pieces that are purchased, recognized, exhibited?  It’s often the case that the easy instantly likeable pieces are the ones that are noticed quickly, but the more intricate and challenging works are the ones that grow on their audience.  Substance and substantiveness grip the mind and imagination and won’t let go, while the easy “like” is quickly forgotten in the constant stream of pretty pictures.

What can we learn from this?

There are two takeaways here for the arts:

1.  If you’re developing a website, collective or other activity that involves selectively promoting work from a group of artists, be very careful about choosing the work that rises to the top based one early popularity data.  You could be left with a site that is defined by easy imagery.  There needs to be a method of bringing less popular work up to the front that still distinguishes between the less popular and slow to like and the work that shouldn’t have been posted in the first place.

2.  Artists whose work isn’t instantly popular shouldn’t worry so much about the early data.  Just make sure your career development strategy includes financial planning and allows you to pay your bills while your slow-growing work builds a following.  Anything substantially new and original meets with initial resistance.  It’s human nature.

Early Evidence Is Often Too Early And Not Really Evidence.

via Early Evidence Is Often Too Early And Not Really Evidence.

 

10 Responses

  1. robpixaday
    |

    Wow! so true, about “easy imagery” and popularity! It’s soooooo tempting to create for the parameters of “the popular,” when there’s no money coming in.
    Terrific post!

    • nerdlypainter
      |

      Thanks Rob!
      And from what I’ve seen at festivals, the easy “popular” route doesn’t work very well. Why? Supply and demand is a push and pull. There are two words there. The people racing to try for cute, easy and “popular” restyling are seeing an apparent demand. But they aren’t considering the huge oversupply of “popular” styles, nor that vanilla styles start to see competition from mass produced and generic “popular: decorating ideas.

  2. nannus
    |

    Popularity does indeed say very little. This holds for other things as well. You can, for example, easily write a blog article that gives you many likes in half an hour and takes 10 minutes to write. Just combine three lines of text, a haiku and a nice photograph and the orange stars will start showing. I once wrote such an article that essentially said I was too tired that day to post anything substantial. Lots of Likes! If, on the other hand, you write a scientific or philosophical article that contains some real substance, you will work on it for hours or days and then it will take a lot of time before anybody takes the time to read it or even like. But on the long term, these articles bring you new followers and interesting discussion. One should strive for quality instead for popularity..

    • nerdlypainter
      |

      I see a lot of people striving to create easy “popular” work online and offline. With art, “popularity” that comes easily is often almost a ruse, tricking the artist onto a path that is neither authentic nor a road to success. People don’t often acquire original art based on a quick response. It’s a major decision for many collectors. So “popular” styles can work against you. There is also an issue of supply, demand, and visual exhaustion. There may be slightly more demand and interest in certain “popular” styles, but there’s also so very much of it out there that the supply oversaturates the demand by several orders of magnitude – the supply of simple works aimed at quick popularity is large enough to exhaust the viewer.

    • nerdlypainter
      |

      This article initially caught my eye because I’ve been fascinated by the way some works percolate through the large art sites, and the way people’s favorites seem related to their favorite artists’ favorite artists. I spoke to a friend about analyzing the data, Simson Garfinkle. He’s a computer Science professor with a quite a bit of insight into data mining and what software is already available to help someone mine the data. I also floated it past the hubs, who has done a lot of algorithm development and is familiar with statistical algorithms (he’s a big fan of “R”). Both computer whizzes seemed to find it odd that I’d want to mine the data on art preference, because the believed that the rather obvious first steps I’d proposed had already been done. They have not – the state of search and recommendation for art online is so appallingly primitive that it took some careful explaining to describe just how “not present and not available” it is.

      So I’ve been trying to learn Orange, and noodling with Knime. Orange is appealing because it plugs into Python, Knime has a layer of tools that dip right into “R”. I don’t know if I’ll be able to single handedly tame the online visual firehose, but Orange is already generating some unexpected correlations when I cluster my own work.

      It’s already known that the social networking functions of the internet can create little “bubbles” of online reality, and this seems to happen very quickly with art networks and art websites.

  3. nannus
    |

    The dynamics of the art market seems to be an interesting area of research. I guess it is quite complex. It has outside influences (like the “financial crisis” or the development of the internet (e.g. what changes are caused by online-platforms and direct sales, as oppose to the traditional way through galleries?)). I guess there are also different types of buyers with different motivation (investment/speculation, status/fashion, or buyers who simply like certain things). Investment- or status-buyers would maybe not even buy things based on what they personally like. I would expect a lot of complexity here. As a cultural phenomenon, the art market should also be something that is historically developing and does not have unchangeable laws.
    However, if you do data mining here, where do you get the data?

    • nerdlypainter
      |

      I’m selling prints through a start-up called Turning Art. They use a subscription model. People subscribe to have a certain number of prints in their home at a time, they try them, then decide to swap them for new ones or to keep them and buy them. There are several levels of interest that can be tracked (anonymously). How often a piece is put into a queue, how often it is viewed in search, how often it is brought into a home, etc.
      Right now I have the data on my own pieces as numbers, but no identifying info on the people picking them. Not even an anonymized code. But I can use the data from my own dashboard to quantify levels of commitment. I can also create indexes for various attributes of my own, familiar, work.

      I’ve been using the data that i have to look for selection trends among my own pieces, adding data to the table describing (guesstimating) the complexity, colorfulness, level of abstraction, etc. of the pieces I have online. It’s interesting that some of the correlates I’d expected weren’t strong and there were others that I had not expected. I also have figured out some of the data I need to look for – for example, it’s probably important to normalize for how long something has been online.

      With these graphical modular programs for data mining, the experiments are cheap to do, so it makes sense to get some data and start poking it, then iteratively start to build better experiments and insights.

      • nannus
        |

        Sounds very interesting! If you could predict what people like, base on what they liked before (at least some correlations) you could create a selling platform (for yourself and for others).

        • nerdlypainter
          |

          Or just a more natural way to navigate all of the visual information on the internet. I don’t see that yet from any of the art sites. Google makes a lot of noise about image searching, but their engine is heavily weighted towards keywords as far as I can tell from experiments in using it.

        • nerdlypainter
          |

          Most of the big Art sites have artist accounts and general accounts where people can “watch” someone or select and create lists of “Favorites”. Some also include sharing options and publicly track how often a piece is shared. This information is all generally hidden behind a tab and isn’t used for much of anything, but it’s also not private. So the data is there. The data is everywhere. Making sense of it in a way that is meaningful … now that’s the challenge.