Skip to content ↓

26/04/10

A Year in the Life of a Book Recommender

We launched the Bookseer in June last year.

Bookseer Home Page - Find your next book

The site had a single purpose: to recommend new books to readers who have just finished something. It would be as simple as Google: a search box asking for the last book you read (and enjoyed). The site would then go and query Amazon, Librarything and other booky API services, and deliver you some recommendations for what to read next. Very simple.

We (well, James) built it quite quickly, and I blogged the results after the pretty insane first month.

And now the dust has settled, I thought I’d update that with some stats for the first 10 months.

Who Visits Bookseer?

Since launch (June 1, 2009) until April 21st, 2010, we’ve had 739,565 visits, 592,969 of which are absolutely unique. 1,839,702 pageviews have been recorded (2.49 pages per view) and a bounce rate of 40.41%.

Across all visitors, there is an average stay of 1m 27 seconds, and 80% of our visitors are first-timers here. Here’s where they come from:

  • 519,249 are from the USA (mostly California, then New York, then Texas)
  • 47,879 are from Canada (really evenly spread out actually)
  • 42,143 are from the UK (overwhelmingly, London)
  • then it’s Brazil, Australia, Russia, Germany, Ireland….

What is also interesting is the difference in usage shown by the different type of visitors. There are two main types of visitors: those who are “referred” to the site by a link on another site (e.g. Stumbleupon) and those who find the site through search.

People who search organically for “what should i read next” and find Bookseer.com have a significantly lower bounce rate (27%), spend twice as long on the site (3m 23s) and view almost four pages, when compared to the average. We think this shows that these visitors have a genuine interest in finding another book.

Encouragingly, these visitors have increased a lot in the past few months. There is also some amazing loyalty: 5.47% of users visited the site more than 9 times. 2% visited the site over 100 times!

Most Popular Books

Without further ado, here are the most popular recommendations asked of the Bookseer in the past 10 months:

  1. Twilight (15,260)
  2. Harry Potter (11,438)
  3. The Help (6,945)
  4. 1984 (6,510)
  5. The Road (4,286)
  6. On The Road (2,195)
  7. Breaking Dawn (3,559)
  8. The Time Traveler’s Wife (3,381)
  9. Pride and Prejudice (3,176)
  10. The Bible (2,859)
  11. Fight Club
  12. The Lovely Bones
  13. Enders Game
  14. Catcher in the Rye
  15. My Sisters Keeper
  16. The Lost Symbol
  17. The Perks of Being A Wallflower
  18. Atlas Shrugged
  19. The Girl With the Dragon Tattoo
  20. Water for Elephants
  21. Brave New World
  22. The Book Thief
  23. Lolita
  24. The Catcher in the Rye (note double entry with above)
  25. The Great Gatsby
  26. Lord of the Rings
  27. The Dispossessed

and many more. Note that these appear to be genuine searches (rather than being distorted, for example by referrals from popular sites direct to results pages). The only title to receive more than 1,000 page views as a landing page is The Dispossessed, just outside the top 20.

Obviously the two big titles are the Twilight and Harry Potter series. I find it interesting that whilst people search on a book name (”Twilight”) for Stephenie Meyer, they do a generic one for J.K. Rowling: “Harry Potter” is, obviously, not the name of a book.

I have a little theory that many people come to the site and test out their favourite book and see what comes back. In fact, I’m certain that’s what lots of people do, and this is why we have such a diverse spread of titles (over 104,000 different books have been entered into the Bookseer).

I’m delighted that the top 10, whilst influenced by Meyer and Rowling, contains some of the greatest works of literature from the 20th and 21st Century. The Road? On the Road? 1984? What a literate bunch of people with great taste inhabit the site.

I am also amazed and thrilled that we have had 104,886 different pages viewed in the period since launch. Ignoring the home page, I think that means we have had 104,883 different titles entered into the system.

And, whilst you might think lots of them have just been entered once (by the author?!) it is not until the 44,008th title (”كتاب السير”) that we start to see single searches. It’s pure long-tail stuff.

Oh and The Bible is quite a fun one, in terms following the author names attached to it:

  • God
  • Jesus
  • King James
  • Various
  • Jesus Christ
  • Peter
  • Moses
  • Anonymous
  • Richard Dawkins
  • The Apostles
  • Many authors

Why (and how?) do People Come to Bookseer?

96% of our traffic comes from referrals, and only 4% comes from search. This is a bit disappointing, and we should work on that.

Search falls into two camps – branded and non-branded keywords.

  • Branded – 70% of search terms are around ‘book seer’ (or a derivative).
  • Non-branded / organic search: ‘what should I read next’ or ‘i’ve just finished reading’ (and derivatives) account for 30% of search (1.23% of visits)
  • .

As of today, bookseer is ranked no 8 for ‘what should I read next‘. The most common “non-branded” search terms, i.e. stripping out variants on “bookseer” are:

  • what should i read next
  • what to read next
  • what book should i read
  • what book should i read next
  • what should i read next?
  • what should i read
  • i just read what should i read next

Which is great as this demonstrates the need that the bookseer is trying to fulfil. What’s also interesting is the impressive growth of these terms in the last 6 months:

PastedGraphic-1

Referrals

Stumbleupon sends a whopping 56.47% of traffic our way. I know very little about stumbleupon but that’s a huge slice of our visitors. Thanks to whoever is involved in that!

In terms of referrals, Lifehacker sent a chunk of traffic (16,745) but pretty much all in one day. Facebook sent 11,114, and Twitter 6,539.

I find this last particularly interesting as on March 5th of this year, Tim O’Reilly (founder of O’Reilly publishers, and with just under 1.5m followers on twitter) tweeted Bookseer:

Twitter -Tim O'Reilly-http---bookseer.com-recom ... (20100324)

I follow Tim, and dreaded it when I saw this – expecting it to fireball our server – but in fact only 210 visits came from Twitter on March 5th. Not sure what this suggests about our site, or Twitter, but I would have expected Tim’s followers to be bookish types who might follow the link. As it is I think it’s 0.015% of Tim’s total readership.

%age of visitors who actually submit a search

Many websites have a majority of visitors who come to the home page and then go away again without actually doing anything. So we thought we’d compare people who come to the site without actually putting in a book, and those who ask for a recommendation. Kind of a test of how good the home page is:

visitors-who-search

The number of visitors who actually search for something remains at a steady 50%. Not too bad, really. In the above chart, non-searches is visitors who didn’t look for a book, and searches is the ones who did: they’re roughly equal!

(Those who don’t search either just come to the site and “bounce” off again, without engaging any further; or who enter the site via a link to a results page [example])

Popularity by Region

We did some data wrangling to breaks out visits by Territory then Page Title, to see if there is a difference in popular titles in a given country. We extracted top 10 title data for each country in turn, for any time period, and thus saw trends for top titles in each country for any time period.

As an example, here’s a table of successful searches by country (USA):

top-10-countries

Drilling down in the US gives us a list of top 10 book titles; we need to filter these by result pages to get:

top-10-US-titles

Clicking on Twilight gives us the trending graph:

twilight

…and we can see that searches for Stephenie Meyer’s vamprance (rompire?) were rather popular in the lead-up to the US film release date in November last year.

Obviously the USA is a huge share of our audience – if there is interest we’ll show UK only titles up here by comparison.

Next Steps

I hope you found this interesting – we did. And it’s made us think a bit, and try a few things out.

So, here’s what we’re going to try and do next:

  • Work on the page ranking to try and improve bookseer.com’s search visibility on a search for the non-branded, organic terms such as “What should I read next
  • Consider social features, such as adding Twitter and Facebook linking on each page, so that people can ‘like’ the recommendations. Compliment this by creating a facebook page for BookSeer.
  • Use the Bookseer Twitter account more. We’ve been having some fun with this, making recommendations inspired by topical events.
  • Consider a mobile app? Some ideas have included the ability to take a photo of an ISBN and get back recommendations immediately. So for example if you’re in a bookshop, you “scan” a book with your phone camera and get back recommendations. Last I heard, Amazon had banned the use of their data in mobile apps

Got any other requests for data or features? Let us know below.

Visit the Bookseer
Follow the Bookseer on Twitter
More about the bookseer

Watch a short film about the data we capture from ebook apps.

Posted by Peter Collingridge in Apt Studio work, Long Tail, Publishing, analytics, bookseer.

New Feature: Instapaper reading // Mills & Boon New Voices Finalists Announced

No comments yet.

RSS feed for comments on this post. TrackBack URL

Leave a comment