Jane Friedman

Reader Analytics from Jellybooks: Crunching the Numbers to Improve Book Marketing and Sales

Note from Jane: This is an extended version of a piece that originally ran in The Hot Sheet, an email newsletter for authors that I run in partnership with Porter Anderson.


Last year, the New York Times dubbed JellyBooks “Moneyball for Book Publishers.”

If you’re not familiar with Jellybooks, here’s the short version: They research consumer reading behavior, and that research is typically paid for by publishers. While companies like Amazon and Apple can track reader usage and data, that data isn’t typically shared with publishers. So Jellybooks gathers willing readers and secures their permission to collect and report on their anonymized reading data to publishers.

Earlier this month, Jellybooks announced they would focus exclusively on reader analytics. Since 2012, they’ve been running the company primarily as a service to help readers discover books, but they’ve now decided to pivot entirely to understanding how people actually read books. Andrew Rhomberg, founder of Jellybooks, says they can answer questions such as:

Rhomberg says Jellybooks typically addresses these issues prior to publication, but sometimes they study why a hardcover hasn’t sold well and how performance can be improved for the paperback.

Jellybooks has already run dozens of reader analytics campaigns spanning hundreds of titles for all types of publishers, and they have provided ebooks for free to thousands of readers. Rhomberg was kind enough to answer my questions about some of their findings.

Jane: When clients hire you, are their needs related predominantly to marketing and promotion of books—to better figure out who to market a book to? I  noticed that, at the time of the New York Times article, people were worried that JellyBooks insights could affect acquisitions or editorial decisions about how writers craft their stories. So I guess I’m asking: Are any publishers hiring you for that type of work?

Andrew Rhomberg: We are still very much focused on the publicity, marketing, and sales side of the publishing industry. Reader analytics is used to measure the strength of reader engagement with a book. … Usually we address [this] before a book is published, but sometimes we also look at why a hardcover hasn’t met its sales targets and how this could be turned around for the launch of the paperback.

Very recently, I was invited to give a presentation at a leading literary agency and the question did come up: Could reader analytics suddenly feature in acquisition decisions? My answer was: very unlikely.

First, if there is no manuscript or only a rough draft, then what are we testing? Acquisition editors have to make decisions on what the potential of a book is, while what reader analytics measures is the reaction of the audience to the final work.

Second, the time it takes to conduct a reader analytics campaign is a couple of weeks, but the typical acquisition process works on much more compressed timescales, so there is a mismatch here.

Third, we at Jellybooks are not keen on finding ourselves wedged between authors and agents on the one side and publishers on the other side. Therefore we are not the least bit enthusiastic about applying reader analytics to this area even if it was a fit.

There are always exceptions. We have often seen great book for which nobody abroad seemed to show any interest. Increasingly we hold detailed data on audiences and reader preferences in multiple languages and territories. What excites us is: How we can help a great book get translated, find a foreign publisher and reach as many readers across the world as possible? This is an area where reader analytics can potentially help a lot, not as a gatekeeping tool, but as tool to make good work really shine and spread its wings. We prefer acting as evangelists for books, not judges of books.

We found that literary agents were excited about the tool because they wanted to maximize the sales of a book. Using reader analytics and A/B testing, it is possible to quantitatively measure what is the best cover, the best title, the best description, and the best way to position a book. The goal is to make great books, and for authors to succeed and reach maximum potential, and the way we work with publishers is very much focused on this aspect. Obviously, not every book has the same potential or should be marketed the same way and a large part of our work is to help find the right approach, target the right audience, and use the appropriate channels to reach a book’s natural audience.

It is correct that editorial still views us with some degree of suspicion, but that’s mostly editors who have never had first-hand experience of the tool. Those who do value it for helping them better understand why some things work. They absorb the findings into their knowledge base and it helps them make better judgements.

Reader analytics also helps cut through internal debates between editorial and marketing when a book is not meeting its forecast. We help people around the table understand what factors are responsible for a book not reaching its intended audience. Reader analytics data is highly actionable, but also provides common ground for debate between different parties within a publisher.

Do you have any specific takeaways to share about the influence of cover art on reading or recommendation behavior?

When picking a book, readers are much more influenced by the cover than they realize. We have routinely asked readers why they pick a book and then compare the answers with results from A/B tests where we could objectively measure the influence a cover has on the actual decisions that test participants made. How did a cover influence readers to pick one book and not another?

The answer is that covers influence readers greatly in their choices and reader are not even aware of it. There are no hard rules why one cover works better than another with respect to “pick me.” It’s not simply about standing out; a cover has to be appropriate for the targeted audience, fit the title, match the description and much more.

The cover should raise expectations, but not create misleading expectations. There is a fine balance. We call it “truth in advertising,” because cover, title and description are the central advertising message for a book. A/B testing is a wonderful tool for measuring what works, but more often than not the outcome of those tests is surprising. It’s just incredibly difficult to predict why one cover will work and another will not, because visuals have such a huge impact in the selection process of how people choose books.

What we have found in countless A/B tests is that covers have a negligible influence on whether somebody will finish a book or not—assuming of course the cover wasn’t wildly misleading, then you do see an effect. Completion rates seem to be driven almost exclusively by content. On the other hand, the speed with which somebody finishes a book is indeed influenced by the cover. Good covers pull people back and give people a reason to finish the book faster, which helps in sustaining a viral cycle with a fast turn-around time. Think of a virus that needs 14 days to become infectious versus one that needs only 5 days. Which virus is going to spread faster? It’s the same with books.

The most important finding we’ve made, though, is that the word-of-mouth potential for a book—the probability that somebody will recommend a book to others—is heavily influenced by the cover. This surprised many publishers, but it makes perfect sense. People in general, and readers in particular, are very much aware that they are being judged by the recipient based on what they recommend. Our social standing is influenced by what people see we read and recommend and, in that context, covers matters greatly. Again this is mostly something people do subconsciously without even being aware of it.

Now, it is important to note that a cover that performs well when readers make a selection for themselves will not automatically perform well when they recommend the book. The two effects are distinct and reader analytics provides a tool for measuring these two influences independently of each other, so we look for the cover and package that optimizes for both situations.

This is also an area where we often have discussions with publishers who argue, “All I care about is that people buy the book, that it sells in the bookshop.” And we respond, “No, you need to also pay attention to the word-of-mouth factor, because the bulk of sales will come from people recommending the book to others.”

In addition, as an industry overall, we need to focus on readers being happy with the books they buy. Only then will we grow the industry and only then will we effectively compete with other entertainment options, such as games and virtual reality and games. Those guys do lot of product testing. They pay close attention to how engaging a game is, if people stay engaged, if it is fun, and much more. If books want to compete with games in the 21st century, then author, agents and publishers will need to pay just as much attention to whether readers are entertained, informed, and content with their book purchases, as people are who play computer games.

Do you think your testing is applicable across all genres? And has one genre been the favorite of yours (or your clients) for testing?

We have worked with a wide range of fiction, from romance and crime to thrillers and horror, women’s fiction and literary fiction, YA novels and fantasy, Man Booker winners and unabashedly mass-market fiction. One category we haven’t done yet—more an accident of history than reader analytics not being suitable for it—is science fiction.

We have found that completion rates in each genre can fall anywhere between zero and one hundred percent. Every genre has super-engaging and less engaging books. There seems to be no rule that romance is more likely to be finished than literary fiction and vice versa. However, romance is generally a much faster read, thrillers are read faster than literary fiction, biographies are the slowest.

Generally speaking, we focus on fiction, any area of fiction. It is nonfiction which we haven’t quite cracked yet, because reading behavior for nonfiction is very different to fiction and completion rates matter less here. Reader analytics in its current form is still useful for nonfiction with a strong story arc, like a Malcolm Gladwell, but for other books it is much more difficult to measure reader behavior because nonfiction books are often purchased or downloaded to be read “some day.” So that means they have a high degree of optionality. This is something that still vexes us. It is a tough nut, but we hope to crack nonfiction, too.

Can you share a story from any recent projects that resulted in a key marketing takeaway or insight?

Well, I stated that reader analytics is not an editorial tool, but there is one thing we have noticed when using reader analytics: a lousy ending that is too abrupt and leaves people hanging or wondering, “Why is this ending now?” This has a very negative effect on the recommendation factor for a book. Readers want some level of closure. Authors should avoid being too clever with cliff hangers. The last 10 or 20 pages really need to seal the deal.


Note: If you would like to experience reader analytics as a test reader, JellyBooks is running a test reading campaign for 10 books with Sourcebooks until end of August that is open to US residents. You can claim one or two books in exchange for sharing your reading data with Jellybooks and Sourcebooks.

If you enjoyed this post, then you’ll want to check out The Hot Sheet, the essential industry newsletter for authors.