The book you list as your #1 book is most certainly not an 'academic' book. Just because a book is written by an academic, does not make it an academic book. Just the byline gives it away ("A Guide for Anyone Who Works with Data"), but the description even more so: "This book describes, simply and in general terms, the process of analyzing data. ... This book is a distillation of their experience in a format that is applicable to both practitioners and managers in data science. "

In other words, "we've written down our experiences as applied statisticians". Which is fine and actually it sounds like it would be useful for me, but it's not "academic publishing".

That has little to do with the fact that they are ebooks, and everything to do with the fact that they are about data science. Just about the sexiest 'academic' topic right now, and in extremely high demand outside academia. For the vast majority of academic books, there is simply little or no demand.

Data science is not a niche academic field by any means. It's a high-growth employment sector with probably tens of thousands of practitioners interested.

Fair point!

Of the 30 books on the front page of your bestselling list, the closest things to an 'academic book' I see are all texts designed to accompany Coursera courses.

Those aren't academic books, those are introductory text books.

I should have been more precise: yes, they are introductory textbooks which accompany courses. I was defining academic as "it's a textbook of any level" not "it's graduate level only".

Leanpub has also been used to publish work based on dissertations (which is as academic as it gets), and those works obviously have far fewer sales.

I would wager that the cost of printing is not what is setting the high prices for these books. Short-run publishing is quite economical these days. It's the overhead cost of running these organizations -- sales forces to sell to libraries, editors, proofreaders, marketing folks, etc.

Switching to e-book will save very little of that money.

Sure, it allows for self-publishing, but a self-published Ebook is not going to look great on your CV, and the royalties are not going to matter much if only a few hundred people are interested in the topic.

I think Leanpub has a ways to go before it can be useful for academic writing in my field (mathematics and theoretical computer science).

My personal experience: I've tried using Leanpub and found issues with the way it handles mathematics. For example, the last time I checked (maybe around June) the TeX support in their Markdown was not developed enough to handle aligned equations, which is a big sticking point for me and I think for most mathematicians.

The standard process for academic writers is to typeset an entire book themselves in LaTeX regardless of the publisher (they tend not to publish-as-they-write, though that could conceivably change). Leanpub is unlikely to convince book authors in math by forcing them to convert to Markdown, especially for book writing when cross-chapter labels for theorems and exercises are common. If Leanpub wants to attract academic writers in math they need a sturdy tool that converts a LaTeX book to the Markdown flavor that Leanpub needs. But that's a crap-ton of work (especially considering third-party libraries like tikz) with an unclear contribution to Leanpub's bottom line.

Of course, the alternative is to post only a pdf format, but that kind of defeats the point of Leanpub which is that they generate multiple formats. A service like gumroad would be cheaper per transaction.

You're correct that we're more expensive per transaction than Gumroad, which is an excellent service. Now, for some books, the amount earned above the minimum price is more than enough to make up the difference: our price sliders work really well to earn extra money for authors. The Leanpub vs. Gumroad question is actually a really interesting one; there are other considerations on both sides. (Gumroad is what I recommend for authors who don't find Leanpub a good fit. I know that authors like Nathan Barry make a killing using their own sites for the marketing and Gumroad for the selling.)

You're also correct that one of the main points of Leanpub is the one-click "make all the formats and sell them" approach. We do offer PDF upload for authors who find Leanpub's workflow preferable to more DIY sites like Gumroad.

Specifically, in terms of LaTeX support: we do support LaTeX math, but you're correct that there are some limitations. I'm not sure if the work we're doing on Markua (https://leanpub.com/markua) will help with this or not. I think that it will help with some of it (making equations into figures and referencing them from other places), but not others (probably the alignment isn't good enough). I'd really appreciate it if you could contact me with specifics if you want...

One thing that's immediate: the only math typesetting language serious math writers ever use is LaTeX, and I think in terms of "math mode" support it would be enough to fully support whatever subset of tex syntax is supported on sites frequented heavily by mathematicians, specifically mathoverflow.com (I believe it's MathJax with some extensions for special scripts and diagrams like [1]). It seems like the technology to get TeX->HTML is already there at mathoverflow, so I imagine it wouldn't be too hard to avoid reinventing the wheel.

The other things are less math specific and more organizationally specific. If I were writing a serious academic math text, it would be full of theorems, corollaries, and exercises which I would refer to in later chapters by a label. For example you might put some simple but tedious computation as an exercise in an earlier chapter and then when using the result in a later proof you can say "The claim that BLAH is Exercise II.4.3." Leanpub appears to currently only have support for figure labels, but in math we often want labels for these other things too, which we can then reference across the entire book. In TeX, anything that is assigned a number can also be assigned a label.

Serious math texts also have a need for bibliographies at the end of each chapter and at the end of the book, and if there's anything a writer doesn't want to do it's alphabetize and manage labels for a hundred references. We have BibTeX for this, but Leanpub appears to have nothing. One could write a tool that translates bibtex output to markdown, but again it's a lot of work.

[1]: http://mathoverflow.net/questions/159655/commutative-diagram...