GiveWell’s latest webinar took a close look at a critical step in our research process: how we’re working to evaluate past grants to understand what happened, why, and how we can use what we learn to improve our grantmaking over time.

Our research spans the full lifecycle of grants. Before we approve funding for a grant, our research team reviews academic evidence, builds cost-effectiveness models, and consults with experts to estimate the program’s expected impact. When programs near the end of their funding, we investigate whether a grant achieved the impact we initially projected. In 2025, our growing research capacity enabled us to expand this final step for select grants and publish comprehensive evaluations that we call “lookbacks.”

During the webinar, GiveWell co-founder and CEO Elie Hassenfeld moderated a discussion with Program Directors Alex Cohen and Julie Faller and Senior Program Officer Adam Salisbury. They covered four examples of grant lookbacks—discussing what we’ve learned and how these lessons are informing our work—and answered questions from the audience live.

Elie, Alex, Julie, and Adam discussed:

• Cash incentives that are more cost-effective than we estimated: New Incentives aims to increase coverage of routine childhood vaccinations in northern Nigeria by providing small cash incentives to caregivers who bring their children to health clinics for vaccinations. GiveWell has directed over $140 million in grants to support the program since it became a Top Charity in 2020. One of GiveWell’s first lookbacks was on our initial $17 million grant to New Incentives. We found the program was about twice as cost-effective as we initially estimated—saving around 8,000 lives. The program reached roughly twice as many children as expected, at about half the cost per enrolled child. This difference from our initial estimate illustrates the uncertainty in our cost-effectiveness estimates and the value of ongoing monitoring, reassessing our analysis, and collecting additional information beyond what’s included in our models—such as an implementer’s track record, conversations with experts and local officials, or qualitative reasons to think a program might outperform its initial estimate.
• Following the data to a difficult decision on chlorine dispensers: In 2022, GiveWell made a grant of around $65 million to support Evidence Action’s Dispensers for Safe Water program, which installs chlorine dispensers for households to treat their drinking water. Our lookback found that roughly a third as many people were using the dispensers as we had estimated. This discrepancy was identified after a separate GiveWell-funded study in Kenya suggested chlorination rates were far lower than Evidence Action’s routine monitoring indicated, prompting GiveWell and Evidence Action to commission independent surveys in Uganda and Malawi. As a result of this finding, which we believe was caused by several contributing factors, we decided not to renew our funding, though we continue to support chlorine dispensers in other contexts where we believe they will be highly cost-effective. This lookback exemplifies the importance of GiveWell’s growing ability to gather and analyze additional sources of monitoring data.
• Strengthening our data and evaluation of insecticide-treated net distributions: The Against Malaria Foundation (AMF), a GiveWell Top Charity, supports large-scale distributions of insecticide-treated nets. We looked back on a grant to AMF in 2023 to support this work in the Democratic Republic of the Congo (DRC). The lookback found that the net campaigns reduced the risk of child death by about 25%, leading the grant to be about as cost-effective as we originally estimated. Learning from previous lookbacks, we collected and analyzed new data sources to better measure the grant’s results. For example, researchers analyzed the timing of children’s births relative to the date of different local net campaigns as a natural experiment, examined three independent quantitative sources, and commissioned qualitative research involving in-depth interviews with people involved in DRC’s net distribution system.
• Challenges looking back on grants for government advisory support: Between 2016 and 2022, GiveWell directed $13 million across three grants to support Results for Development’s program to improve treatment for children with pneumonia in Tanzania by providing advisory support—that is, technical assistance—to the government. Our lookback found that the share of health facilities stocking amoxicillin-DT, used to treat pneumonia, rose from approximately 48% to 90%, and that financing and procurement fully transitioned to the government. However, it is more challenging to know what would have happened in the absence of technical assistance—in this case, because the government was already working toward this goal. Though our confidence isn’t as strong as for other lookbacks, the team gathered some indirect measures of impact—such as stock rates for amoxicillin-DT growing faster than other medicines not targeted by the grant. This lookback taught us the importance of interrogating the theory of change for future technical assistance grants to understand how activities are addressing bottlenecks and to identify reliable predictors of success, such as existing government relationships.

During the audience Q&A, the panelists fielded a range of questions from the more than 200 attendees. These included examples when qualitative data has informed our grantmaking decisions, how GiveWell has been using AI tools to responsibly extend research capacity, and how we distinguish between predictable and unpredictable program disruptions when evaluating grants.

Panelists also discussed how they expect lookbacks will influence GiveWell’s research in the future. Adam shared how lessons from technical assistance lookbacks, like the one for Results for Development, are already informing active grant investigations. Alex discussed the additional potential for lookbacks to help GiveWell be a better grantmaker over time by teaching us how programmatic changes can increase a grant’s impact. Julie highlighted how lookbacks are setting us up to learn by helping us make better decisions at the start of a grant, such as external monitoring choices.

To close the event, Elie addressed a final question about GiveWell’s long-term success. The world has a huge amount of need—approximately 4 million children die in low- and middle-income countries each year, and most of those deaths are preventable.¹“An estimated 4.8 million children died before reaching their fifth birthday in 2023—deaths that were overwhelmingly preventable.” UNICEF, “Levels and Trends in Childhood Mortality,” 2025.

“In 2024, an estimated 4.9 million children under the age of 5 years died worldwide, equivalent to one death every 6 seconds.” World Health Organization, “Child Mortality (under 5 years),” 2026.

Our World in Data, “Number of Child Deaths by Region,” 2026, shows 4.0 million child deaths in 2023 in sub‑Saharan Africa (~2.8M) and southern Asia (~1.2M), where nearly all countries are classified as low- or middle-income (see map in Metreau, Young, and Eapen 2026). We’re very grateful for donors’ support that has allowed us to make larger and larger inroads against these significant problems. We expect that in 2026 we will direct the most funding that we ever have to save and improve lives.

Through lookbacks, we are learning from past grants to continuously improve our grantmaking portfolio and to direct donor funding where we believe it can help people in need the most.

Visit our All Grants Fund page to learn more about how you can support this work.

This webinar was recorded on June 9, 2026, and represents our best understanding at that time. A transcript of the webinar is available.