A forward by Mek, Open Library’s program lead:
This year, through Google Summer of Code, the Open Library Team had the privilege of collaborating with Tanishq Sangwan to help readers avoid dead ends and discover more of the books they’re looking for. This collaboration was particularly fruitful because of Tanishq’s strong product sense, his commitment to metrics, and his ability to move quickly from idea to prototype. Tanishq has an impressive ability to cover a lot of ground while consistently bringing thoughtful, ambitious, new ideas to the table. He’s also a clear and proactive communicator, making him an especially well rounded and effective collaborator. I’m excited for the world to see what Tanishq accomplished this summer, and even more excited for countless readers to experience the benefits of his work firsthand.

This year, Open Library published a call for Proposals to tackle one of its most persistent problems: patrons registering, borrowing (or trying to), and disappearing. My proposal, “Personalizing the Patron Experience,” set out to fix that drop off by giving new patrons a taste-driven onboarding flow and a personalized dashboard to land on.

Going in, we thought the gap was mostly about personalization – that patrons weren’t returning because there was nothing tailored to bring them back. Once we looked deeper into the real funnel, instead of assumptions, the picture shifted: a large share of patrons were dropping off way earlier, before personalization could even matter. 

Why now?

In 2023, more than 500,000 books were removed from the lending library. That single event created a lasting mismatch between the value patrons expect from Open Library and the value they are able to access when they show up. A library catalog matters so long as it is useful every time someone visits it – and for a growing share of visits, it wasn’t.

My proposal, built with feedback from my mentors, focused on the post-registration onboarding experience: the sequence of moments right after a patron signs up, when they either find a reason to stay or quietly leave.

Finding a Deeper Problem: A Drop-off Funnel Analysis

Our initial hypothesis was that focusing on the post-registration onboarding experience was the best point of intervention to help patrons connect with more value. Before racing forward and writing code based on this assumption, we evaluated our hypothesis against analytics about the current journey of Open Library patrons.

Our findings from this research helped us shape a revised plan that would ultimately lay groundwork to make our onboarding work as impactful as possible:

1. Patrons fail to find a readable book more often than they succeed.

Analytics show that for every successful “Borrow” click on Open Library, patrons clicked “Locate” twice, attempting to access a book that wasn’t available to borrow. That means patrons who came looking for a specific book were unable to read it roughly two-thirds of the time. This is a problem we needed to address first because no amount of post-registration onboarding will compensate for an initial experience that fails to connect most patrons with a book that is useful to them.

2. Registration itself works against onboarding. 

Before patrons could begin an onboarding flow, they first had to make it through a complicated, high-friction, confusing registration process that brought them to archive.org for activation. Just as bad, the registration process did not remember or preserve the reason a patron registered an account to begin with. In the 1/3 of cases where patrons found a book of interest, they would have to survive a confusing registration process and remember why the registered before accessing that value.

3. Day zero is a dead end. 

After registering, a patron lands on a virtually empty “My Books” page. It’s a ghost-town: There’s nothing to do, nothing to return to, and, unsurprisingly, very few patrons find reason to re-engage after that first visit. This is the crux of the retention problem: onboarding doesn’t fail at signup, it fails in the days immediately after.

Starting with the ability to measure outcomes

Before building any patron-facing feature, we needed a way to actually know whether what we shipped was having a positive impact – not just a hunch that it probably helped. So the first thing built wasn’t a UI feature at all: it was a micro A/B testing framework (#12789) to give the team a way to run experiments and get data-backed answer without guessing.

Helping more patrons succeed

In a library where two-thirds of book clicks are leading to dead ends, the immediate question shouldn’t be how we can optimize for the minority of cases that succeed, it’s: how can we flip the calculus so the majority of patrons succeed. In service of helping patrons succeed in finding books of value to them, we brainstormed four solutions.

Fixing registration so every success counts

For many years, after a patron signed-up, they would receive an activation email that redirected them to archive.org and then required them to return to Open Library and then login. This process was confusing because it spanned two domains. During our collaboration, Mek switched this flow so that activation emails brought you directly to Open Library and automatically logged you in. But what then?

Prior to my work, the registration process would forget the action the patron was trying to accomplish, making success less likely. To address this problem, I designed a banner (#9409) that remembers a patron’s registration intent and allows them to seamlessly continue after they’ve registered. Since implementing this feature, more than 10% of registered patrons choose to use this to continue where they left off.

Giving patrons reasons to return

As a result of the above interventions, we expect our metrics to show a greater percentage of patrons are succeeding to connect with book they love — whether that’s taking advantage of search inside, finding suitable alternative books when their first choice is unavailable, or having a registration process that works with them instead of against them.

The final step (and our initial goal for GSoC) is to give patrons reasons to return. This means every time a patron visits their “My Books” page, they should encounter value. We outlined two efforts that dovetail with our original onboarding plans:

1. The “Continue Reading” carousel: a consolidated view of books the patron may wish to continue reading.
2. The “Build Your Library” section: a collection of ways for patrons to add books to their reading log.

We were able to complete the “Continue Reading” initiative during the course of GSoC. We produced specifications and mockups for the “Build Your Library” phase, which we plan to continue working on as follow-ups.

The “Continue Reading” carousel (#13256)

Over the past several years, the Internet Archive has moved from a default two-week loan period towards more flexible and equitable models that allow patrons to borrow book for the duration of session, so long as the book is being actively used. On one hand, this means fewer patrons are leaving books sitting unused on their digital desks. On the other hand, it means going to your “My Books” page is less likely to show active loans to continue reading. To address this problem, we’ve been building a “Continue Reading” carousel with two phases:

1. Phase I: Merge Active Loans & Loan History (#13272).
   • Because many loans expire after about an hour and then vanish into a separate history table, this phase makes it so that both your active loans and your loan history get merged into a single coherent view, making it easier for you to continue reading where you left off.
2. Phase II: Enhance “Continue Reading” by adding open-access books (#13273, #13274).
   • One gap of our current system is that open-access books don’t get recorded under loan history and so never show up under your My Books page. Similar to our “recent searches” history, we’re first using local storage and then exploring a dedicated database table to help patrons keep track of the open access books they’ve began reading. We intend for these settings to be manageable via the loan / reading history page.

Next steps: The “Build Your Library” section (#13255)

After GSoC, we plan to continue development on the second phase of improvements for the “My Books” page, to give patrons clear calls-to-action and new experiences for adding books to their library. The goal it to give patrons the tools to, “Make Open Library Your Library“. Rather than designing a blocking onboarding flow that is only done once or even skipped, we decided to try adding a section to the My Books page that increases discovery of our import options (Goodreads, barcode scanner, search) so that patrons can add books any time they visit their library. As more patrons use these import options and grow their reading logs, their activity contributes towards book discovery for everyone. The library grows not just through its books, but through the collective reading experience of its readers.

The idea of the new “Build Your Library” section is to meet patrons where they already are, based on which of three groups they fall into, and give each a seamless action to bring their reading into Open Library at once rather than book by book:

• New readers: a flow to explore genres and bulk-add books to shelves and lists directly from that exploration, instead of searching for each book individually, opening its page, and adding it one at a time.
• Readers with a library elsewhere: the import tool already exists but is underutilized. This means improving that experience and expanding it to cover other popular platforms, so a patron can export from wherever they already track their reading and import straight into Open Library.
• Readers with a physical collection: the barcode scanner already exists but is underutilized too. This means improving it for bulk scanning, so a patron can log a shelf of physical books in one sitting instead of one at a time.

How patrons discover what to add next, how shelves get organized and recommended, and how the page evolves as a patron’s reading history grows are all part of this epic too. This is the epic I’d point a future contributor toward first.

This unlocks two things down the line. Once patrons have a real library instead of an empty shelf, there’s finally enough signal to build genuine recommendations – curated to taste, not generic – since bulk-adding from these three flows is what generates that signal in the first place. It also revives the existing activity feed, which is underused today for the same reason: it only has something to show once patrons are active. As more patrons bulk-add their libraries, the feed becomes more valuable to the whole library.

Impact

Already, several of our interventions have improved the experience of thousands of patrons across the Open Library.

• Preserve Intent banner: ~500 clicks/day against total shown, a ~12% click-through rate – patrons redirected through registration are following the banner back to what they were originally trying to do.
• Preview / Search Inside visibility: ~13,800 previews and ~1,100 Search Inside opens per day since the buttons became more prominent.
• Unavailable-book alternatives banner: ~2,500 clicks/day, patrons routing to a readable alternative instead of a dead end.

Having fun along the way

One of the parts I enjoyed about working on Open Library was the autonomy to work on features that I thought were useful. One example was the “Stopped Reading” bookshelf (#12400) that many patrons had been requesting for years. There was another case where I took the initiative to improve the design the highly visited “My Loans” page (#12912) to make it easier for patrons to navigate their books.

What I’ve learned

The instinct going in was to treat this as a personalization problem – better empty states, taste capture, recommendations. Open Library already has plenty of features like that, loved by a smaller set of patrons – it’s easy to keep adding to that pile. The funnel numbers pushed me to stop designing from “what’s a good addition” and start from the patron’s actual seat: what’s stopping them from getting anywhere at all? That shift mattered more than any single feature – it’s why the summer became about removing friction patrons were already hitting, not adding something new to discover.

Looking back, a lot of the highest-impact work shared the same idea: don’t ask patrons to do something new, save what they’re already doing. Preserving intent through registration (#9409) means a patron who clicked “borrow” doesn’t have to re-find that book after signing up – we just remember what they were already trying to do. The Continue Reading epic (#13256) applies the same idea to reading itself, picking up on what a patron’s already reading instead of asking them to curate a shelf. Neither asked patrons to learn something new – they just stopped losing track of what was already in motion.

I also came away with a much better appreciation for measuring before shipping. We didn’t get to actually run experiments through the A/B testing architecture yet, but having it in place – ready to give real answers instead of hunches – is something I’m genuinely excited to put to use.

Finally, I’m proud that we we’ve been able to move quickly and prototype many features that span the patron experience, however, with a large open source project, it’s been a learning that we can only move as quickly as the review process and the feedback we’re able to get from the community. There are several efforts are in review or mid-flight, worth tracking post-GSoC:

• #13113 – Consolidating buy options (open)
• #12914 – Locate → Check WorldCat options (open)
• #13281, #13282, #13283 – the three Continue Reading phases (all open)

None of this ends with GSoC. Both epics – Continue Reading and Build Your Library – are far from done, and I plan to keep working on them past the program, seeing the remaining phases through rather than leaving them as a hand-off.

Acknowledgments

Thank you to my mentor, Mek Karpeles, and the Open Library staff and community – they’ve been extremely helpful and supportive throughout, even before GSoC. The drop-off funnel analysis that shaped nearly every decision in this project came out of that early framing, and the steady feedback across a summer of shipping into a live, high-traffic library catalog made all of it possible.

A special thanks to Lokesh Dhakar and Ray Berger, who encouraged and supported me in getting started with Open Library back in February – which eventually led to spending the whole summer building with the team.