Stories, evidence, and learnings from the organizations and Frontline Workers delivering verified services through Connect.
Trained Frontline Workers carry Kangaroo Mother Care into the home, closing the dangerous gap that opens when families of small and vulnerable newborns leave the hospital.
Read more →
A new grant will scale and rigorously evaluate Connect, delivering Vitamin A and oral rehydration solution to children where coverage is lowest.
Read more →
GiveWell has awarded Dimagi more than $1 million to scale verified child health delivery across northern Nigeria, funding over 300,000 visits in 18 months.
Read more →
Across seven countries, Frontline Workers delivered more than 100,000 verified services, a foundation for scaling to millions of visits a year.
Read more →
The first year of designing and testing Connect: 77,000 clients reached, $161,000 paid to Frontline Workers, and the learn, deliver, verify, pay model proven in the field.
Read more →Every entry here is published in a funder report, an independent evaluation, or a public partner document, and has been judged honest, scoped, and worth putting on the public record. Underlying reports are available to funders and researchers on request.
In adversarial testing we paid Frontline Workers to generate fake data with financial rewards for those who did best. A model trained on three data fields hit AUC 0.91, detecting 100% of fake Frontline Workers while only flagging 2.5% of real Frontline Workers as suspicious.
Without a grid, Frontline Workers cherry-picked easy houses. With a grid forcing them to go door-to-door, the hardest-to-reach children stopped being the ones who got missed.
That figure is Connect’s all-in cost per visit, including commodities, LLO payments, setup, and product cost. The biggest driver of the drop was setup costs falling from ~$0.40 to ~$0.10 per visit as we moved past early-stage inefficiencies. We report cost-per-verified-outcome rather than cost-per-worker-trained because it’s the number that actually matters.
Past reputation, team size, references, none of it told us if a partner organization would actually deliver. So we contracted small first. Of 37 partners, 24 ran trial runs. 10 underperformed. 3 were selected to scale based on their performance.
Locally Led Organizations independently mobilized imams, pastors, village elders, and community-recommended Frontline Workers. The platform didn’t teach this, they already knew.
Remote coordination worked across Nigeria, Kenya, Uganda, Tanzania. It didn’t hold in the Central African Republic. Staffing turnover and connectivity issues exited both contracted partners.
If missing the window cost a worker money, we’d see suspiciously perfect compliance instead. So we treat it as a guardrail, not a payment criterion. We coach and integrate hospital discharge notifications to close the gap.
The headline 50% number understates the gap. The mean is 6.05 days. One delivery opportunity doesn’t collect hospital discharge dates at all, so 100% of those records are blank on this metric, which means the real-world bottleneck is data plumbing, not worker effort.
The Ministry is participating in working groups, quarterly reviews, and site visits, and contributing data to Uganda’s five-year national neonatal plan. The next step: government co-funding at $10-20 per case.
That covers partner engagement, the Kangaroo wraps, scales, thermometers, pulse oximeters, training, payment per verified case, and supervision. We picked it for our first individual-giving experiment because $60 is the smallest unit at which one donor can fund a complete intervention with verified delivery.
Workers who passed our digital test still struggled with nuanced in-home counseling. No single layer of training was sufficient. This improved with 78% of workers scoring above 80% during their first observed visits. Average performance increased from 85% to 91% in the second observed visit.
A 50-caregiver baseline showed near-universal endorsement of responsive caregiving. Attitudes weren’t the gap. Knowledge and observed teaching practice were, and that’s where we saw movement. We doubled the visit cadence in the next round.
Autonomy encouragement requires consistent behavior change across many interactions, not a single coaching session. It’s the sub-domain least moved by the current visit structure. We aim to expand from 3 to 8-10 visits per child specifically to give this target more runway. We don’t know yet if it will be enough.
Nothing matches those filters.
Whether you want to hear more about a program or dig into the evidence behind it, we are glad to talk. Reach out and we will share more.
Get in touch →
