In his book Super Crunchers: Why Thinking by Numbers is the New Way to be Smart, Ian Ayres outlined how statistical methodologies are challenging expertise and intuition in a multitude of fields including seemingly unlikely areas such as film making. In the book, he discussed the approach used by a company called Epagogix to selecting movie scripts which would most likely produce a profit at the box office. The firm developed a neural networkto based on the analysis of numerous successful and unsuccessful scripts. The neural network tuned the weights of the various input factors to a point where, according to Epagogix executives, it could pick winners eight out of ten times. By Hollywood standards – or any other entertainment industry for that matter – that is a phenomenal success rate.
Interview of Ian Ayres, author of Super Crunchers
The claims were both lauded and challenged in the popoular press, but venture capitalsts were sufficiently impressed to invest in the company and start using its methods in movie production. Given the length of time it takes to produce films, it will be awhile before the methodology is able to show us the money – or not.
But the story raises an interesting question for publishers. Could this approach be used to select books for publication – especially in the fiction arena? The first reaction to such a proposition might be a dismissive, defiant NO – such a thing is impossible. How could an algorithm, a mindless piece of software make a judgment about the merits of art and the reaction of its human consumers? Actually, if we stop and think about it, this may not be such a stretch.
- Track record – Publishers have a poor history of selecting books that will be profitable. Estimates of profitability range from about 1 in 10 to 3 in 10. This by itself is a clear indication that the human powered title selection process is deeply flawed (at least from a business perspective). Perhaps an algorithm could do better or at least no worse.
- Distractions, distractions – Much of human intuition is geared toward protecting ourselves from danger and figuring out the behavior of our fellow humans. These serve us well as a species, but not so well when it comes to analyzing in a brutally objective manner those elements that make titles successful. The reason a publisher takes on a title may have more to do with relationships than business considerations.
- Short memories– Any kind of statistical analysis starts with a meaningful collection of data. The human memory is an amazing, but in many respects fallible tool. It is hard to keep in mind thousands of samples of successful and unsuccessful books – we usually just remember the outliers on either end of the spectrum. So we develop rules of thumb that may be biased to the outliers and perform poorly for the bulk of books published.
- Useful judgments– What we’re really good at is figuring what are the right factors to take into consideration in the first place; not sifting through mounds of data to assign the weights to these factors. Book publishers can build predictive models based on factors they judge to be the most important. Then back test the models and see which factors really are significant. Once the key success factors are identified, sample data sets can be fed into models like that developed by Epagogix to tune the weights for each factor and start making predictions.
Will publishers adopt such analytical methods in selecting their titles? Maybe – but probably not. As Ayres pointed out in his book, when Epagogix approached one major Hollywood studio about their algorithm and presented their evidence for its effectiveness, the firm was turned down cold. When Dick Copaken, CEO of Epagogix, asked the studio executives why they wouldn’t use the tool even it picked 8 out of 10 winners, they replied that it would interfere with their long standing relationships with agents, agencies, actors, producers and directors. “We wouldn’t be invited to the right parties. Our wives would hate us.”
Opportunity could be knocking for publishers who are weary of the pursuit of elusive best sellers, and need new thinking to survive in this dreary economy. Perhaps the most counter intuitive idea we would have to swallow is that to find better art, we may to have sublimate part of our humanity.
- Book Publishing’s Month of the Long Knives
- Can’t All Books Be Above Average?
- Petabytes vs. Puny Books
- Dark Mysteries of the Amazon Sales Rank
- Why Publishers Buy (or Good Buy, Bad Buy)
- Publishing Analytics
Share this Post