Against Forecasting: A Case for More Agility in Book Publishing
For the past 60 years book publishers have thought of themselves and behaved like packaged goods businesses. It has been sufficient for publishers to adopt, often at a lag of up to 15 years, best practices (like inventory turns, just in time delivery, and sales forecasting) from traditional businesses even when those ideas haven’t served publishers or readers very well. Going forward the publisher-as-goods-packager model will break. It is now required we not lag but leap ahead of other industries by 15 years. Can publishing survive the time crunch? Agility rather than stability will be the new paradigm. Looking to the software industry might help.
It is widely understood that the book publishers and book retailers are behind the times. We have been in catch-up mode for say, the last 40 years. The reasoning goes that if only we adopt best practices from other ‘best-in-class’ companies then the sky will open and the industry can be saved.
Bupkis.
Inventory Turns Are For Accountants
Ten years ago I sat on a conference call where a participant preached measuring inventory turns as the salvation for the publishing supply chain. We had the same conversation every week for two years and never managed to get our inventory at an average turn rate much higher than two. Inventory turns are a common metric in dry-goods retail. The participant was from dry-goods retail. Using turns to manage book retail is useless. Let me explain why.
Photo by Smudie
Inventory turns are a financial metric – total dollars received versus total dollars sold. To be useful to a bookstore manager, at the very least, turns need to be measured at a unit level not the dollar level AND the isbn level not the store level. And even if it was measured per sku per store you would need to allow for returns and reorders (not to mention stock outs). If you did all that you would be tracking the sales of the book so closely, you might as well track the sales of the book itself and forget the turns altogether. At that, what would it teach you? Are you going to manage Dan Brown any differently once you learn the Lost Symbol turned 675 times? I don’t think so. Oh look here comes the trade paper. You better return all of Jane Austen to get your turns in line. Using inventory turns to manage your business is not only misleading, it is dangerous.
Title Forecasting is Catnip for MBAs
A similar dogma exists with sales forecasting. The reasoning goes like this: 1) our industry is mature. 2) in mature industries you control costs 3) stricter forecasting is a great way to control costs. 4) book publishing needs more sophisticated forecasting algorithms to help control costs.
Nope.
Sales forecasting in the book trade is useless. Let me explain why.
Forecasting is only effective when the model is stable, when a sales history is well established, and the time lines are short. As Nassim Nicholas Taleb so forcefully points out in his book The Black Swan forecasting models don’t allow for new information. They filter out randomness. They deny the existence of black swan events. Forecasts, by definition, are self-fulfilling. At worst they are frauds.
By necessity books are almost exclusively forecasted with no past sales information. Book sales almost entirely rely on happy accidents (aka black swans events). Not only is forecasting sales a waste of time for most books most of the time, it too can be dangerous. The models I have encountered deal with volatility in one of two ways: the book sales are modeled as if they were on rails (no new information allowed) or the volatility (aka potential demand) is zeroed out as forecast noise. In both cases the ‘forecast’ is suppressing reorders. It is done to protect against the downside risk of too much inventory. No book forecasting model I know of is engineered to capitalize on potential upside.
/Polemic
No forecasting? Be afraid of random events? I am not saying you should lock yourself in your stock room and never make a decision again. You have to submit a purchase order with some number on it. So what is the alternative?
Book publishers and book retailers should focus on agility not accuracy. Being smarter is a delusion. Get faster.
Zara, the Spanish clothiers, do not forecast their inventory. Imagine if book publishing worked like fast fashion? Write it. Print it. Jump on it if it works. That is impossible you say. No it’s not.
I am amazed at how many Canadian book publishers under deliver on orders. A bookstore will order 100 copies and the publisher will deliver 50. Come on. Yes the book publisher owns the risk. Yes the book publisher has to endure baroque negotiations with their US head office but it would be a lot easier if the publisher and retailer did some collaborative planning. Shrink the supply chain. Get faster.
In the Black Swan, Taleb offers the more extreme ‘dumbbell’ solution. Spend a small amount of your time and attention on managing the safe bets in your publishing portfolio. Spend the rest of your time preparing to pounce and then pouncing on the riskier bets with the largest potential. Leave all other titles to fend for themselves. Or better yet, don’t publish them at all.
The End of Inventory Management
I have spent the last six months learning about agile methods for software development and that has prompted thoughts not only of a leaner, faster book supply chain but a future where the supply chain shrinks to nothing. In a world where publishers sell directly and instantly to readers, it makes a lot of sense to have a rapid response task-force (author, editor, marketer, designer) to not only respond to sales demand ex post facto but modify and create new content in response to readers’ needs.
If a CEO of a major publishing house familiarized themselves and then adopted a ‘go-agile’ philosophy, I think that would constitute an almost bullet-proof strategy for dealing with the forthcoming wave of digital change. The question is what would be lost in such a transition?
related: Mark Leslie’s thoughts on PubFight, the forecasting simulation game from BookNetCanada
related: Chris Dixon, in a great post on commoditized forecasting algorithms points out that studies have shown that naive bayes is as good or better than fancy algorithms in a surprising number of real world cases. He says “it’s relatively easy to build systems that are right 80% of the time, but very hard to go beyond that.” This post is well worth your time.
8 Comments