For Libraries

Find out what actually works — in your library system, for your patrons.

Public libraries are one of the best-positioned institutions in the country to run rigorous experiments. High civic trust. Voluntary participation. Rich administrative data. We help library systems design, run, and publish low-cost pilots — and share findings across the field.

Start a conversation →Library experiments in the registry

Why libraries

The first-wedge institution.

The Experiment Society's first partnerships are focused on public libraries — not because library problems are trivial, but because libraries are among the few public institutions where the conditions for rigorous experimentation already exist.

High civic trust

Libraries consistently rank among the most trusted public institutions in the United States. That trust makes randomized pilots politically feasible — patrons are not surprised that their library is trying something new.

Rich administrative data

Card-holder records, circulation data, program registration, and branch visit logs give libraries outcome data that most government agencies spend years building. You already have the measurement infrastructure.

Voluntary participation

Library programs are opt-in by nature. Assignment to message variants, reminder schedules, or outreach sequences is low-stakes and consistent with how patrons already experience the library.

Replicable across systems

A finding from the Columbus Metropolitan Library travels to Hennepin County, Denver, and NYPL. Libraries share enough operational structure that results generalize — more than almost any other civic institution.

Pilot templates

Four experiments ready to run.

These are not hypothetical. Each is designed to fit within normal library operations, require no new data systems, and produce a publishable result in 8–12 weeks.

Program attendance sprint

Which outreach message produces the highest attendance at a free library program?

Design

Randomly assign registered card-holders to receive one of three message variants — informational, social proof, or personalized — for a 6-week program series.

What you measure

Registration rate; attendance rate; return visits; opt-outs and complaints.

Timeline: 8–10 weeks from design to published report.

Prior evidence: A social proof variant ('127 families in your neighborhood have already signed up') increased attendance 23% vs. a standard informational message in a 2015 NYPL pilot.

Summer reading engagement

Do weekly SMS reminders to caregivers increase summer reading challenge completion?

Design

Randomly assign enrolled children to receive weekly progress reminders (sent to parent or caregiver) vs. no SMS contact. Messages include reading goal, current progress, and a single encouragement.

What you measure

Challenge completion rate; books logged; branch visits during summer.

Timeline: Runs across summer program cycle; results available September.

Prior evidence: NYPL found +18 pp completion rate from weekly SMS reminders in a 2015 RCT. Effect was largest for children who had previously attempted but not completed the challenge.

Late fine waiver outreach

Which communication approach most effectively restores lapsed card-holders after a fine waiver program?

Design

Card-holders with blocked accounts randomly assigned to receive: (a) standard notice, (b) simplified one-step waiver instructions, or (c) personalized letter with branch staff name. Measure reactivation within 90 days.

What you measure

Account reactivation rate; checkout activity within 60 days of reactivation; opt-out rate.

Timeline: 10–12 weeks.

Prior evidence: Chicago Public Library's fine elimination (2019) restored 180,000 blocked card-holders. Message framing variants were not tested — that's the gap this pilot fills.

Digital literacy program enrollment

Does proactive outreach to seniors without internet activity increase enrollment in digital literacy programs?

Design

Card-holders aged 60+ with no recent digital resource checkouts randomly assigned to receive a targeted invitation to a digital literacy workshop vs. standard branch-wide promotion only.

What you measure

Workshop enrollment; program completion; digital resource checkouts in 90-day follow-up.

Timeline: 12 weeks.

Prior evidence: Segmenting outreach by behavioral signal (low digital activity) rather than age alone is a design improvement over most existing library digital inclusion efforts.

Registry Evidence

What library experiments have already shown.

Chicago, USA · 2019

Library Fine Elimination

+240% returned books in month following elimination; significant increase in new cards from low-income zip codes; no increase in unreturned materials

Salt Lake City, USA · 2019

Library Fine Elimination — Return Rate Impact

Late returns fell from 9% to 4% after fine elimination

New York City, USA · 2015

New York Public Library Summer Reading Text Reminders

Challenge completion: +18 pp (55% vs. 37% control); books logged: +1.4 books per child; library visits: +0.9 visits per child; effect largest for children who had never completed the challenge before

Process

Four steps. Your staff runs the program. We handle the science.

Conversation

A 45-minute call with your team. We help you identify a question you genuinely don't know the answer to and that you have the data to test.

Design

We draft an experiment charter: question, intervention, comparison, primary outcome, sample size calculation, randomization method, and analysis plan. You review and approve.

Run

Your team implements. We are available for questions. We handle the analysis once data collection closes.

Publish

A public report goes into the registry within 6 weeks of data collection closing — positive, null, or negative result. Your system is credited. You retain all data.

Cost: Initial library partnerships are free. We treat them as learning investments and use the published results to build the evidence base. Larger or multi-branch pilots involving dedicated analyst time involve a fee we discuss upfront.

Get Started

Run your first pilot this quarter.

A 45-minute call is enough to identify a testable question and sketch a design. We'll tell you honestly if the question is worth the effort — and what a realistic result might look like.

Start the conversation →