It’s important to do more than one estimate of your titles’ sales. Why? Because each set of data you use will give you a different number, and all of them will be wrong in different ways. If you average them, you get a better chance of a good prediction. Of course, if you could pick the best estimate from the bunch, you’d be closer, but which one is it? I don’t know, and you probably don’t either.
The previous 3 parts of this series dealt with the use of experience, marketing plans, and Amazon data on comparable titles to predict the sales of either that comp or of your book. Part V will discuss using sales of comps to predict sales of your title, and then combining all of the various predictions into an overall prediction of sales for your title.
This part of the series covers projecting the sales of comparable titles from other sources of data, and combining all the various projections of comparables’ sales. The best sources of data besides Amazon are Bookscan, Ingram, and perhaps the data your friends inside other publishing companies are willing to share. (That last is an old industry tradition that is fading slowly as the industry expands and changes.)
So, you have a list of carefully chosen comparable titles. They’re all from companies that are distributed in similar ways, and they all have similar marketing muscle behind them. They’re all aimed at the same audience, and are intended to fill similar needs.
Now what? Well, Ingram has the iPage facility. Use it to get a good idea of what Ingram’s volume has been for each of those comparables over the past few months or a year. Bookscan also offers good data on sales.
Ingram is, of course, the primary wholesaler to the book trade. You should know roughly what fraction of this type of book’s sales will go through them. For many trade books, it will be something like 50%, but for others, it will be much more or much less.
Bookscan records approximately 70% of the sales by general bookstores. Obviously, the publisher will be selling more to the bookstore than the store has moved out of the doors, but this is still a solid number. It may not be very helpful to you, though, if you’re doing books that don’t move through bookstores and similar retailers.
If you have those numbers for a “bookstore book,” or even for one that will sell a significant share of the total through stores, you can make a pretty good estimate of the total sales for that comparable title.
To go from the fraction that went through a channel, say Ingram, to the total sold, is simple. Divide the sales by the fraction. For example, let’s say that half of your books generally go through Ingram. And that you know your comparable title sold 1,000 copies through Ingram. Divide that 1,000 by 1/2, and you get 2,000. That’s your estimate of how many copies that comparable title sold — based upon the Ingram data.
For Bookscan, it’s only slightly more complex. You divide Bookscan’s report by .7 to arrive at the estimated bookstore sales, and then divide that result by the fraction that you expect to distribute through bookstores. That result is the Bookscan-based estimate of your comparable title’s total sales.
If you have an estimate of the sales through Amazon (from Part III), you can gross that up to an estimate of total sales for that title, by again dividing by the fraction of total sales that you expect to make through Amazon. (For trade books from a mid-sized publisher, that might be .15, for example. For niche non-fiction from a micro-publisher, that number might be as high as 75 or 80%.)
To make ANY estimate better, you can combine a number of different versions that are based upon different data sources. If you average them, you should help remove any random error. Here, you’d take all three estimates of total sales for each comp (based on Ingram, Bookscan and Amazon numbers), add them together, and divide by 3. [Yes, I know you know how to average. Just being complete.]
CAVEAT: This doesn’t always work. If Murphy is out to get you, all the errors will be in the same direction, perhaps because of some underlying and confounding variable that’s not properly understood.
My response to this issue: if my judgment says that the results don’t make sense, I either dig further to find the problem, or ignore the estimates. Do NOT, NOT EVER, ignore that little voice in your head that says something’s wrong.
Any questions out there? Come on, ‘fess up. I’m pretty sure my writing isn’t so incredibly lucid that everyone is still with me.
For that matter, does anyone have a better way?