Measure what your grants actually change

Agents bring your grants, your outcome models and public data into one place, and show what each dollar is estimated to achieve and why. No new data infrastructure or data science team required.

Explore a live impact model

Agents built this from GiveWell's public grants record and its published cost-effectiveness models for four programs.

It answers one question: what did GiveWell's grants buy, and where does the next dollar do the most good?

Pick a program to see the places it covers and what drives the estimate.

What runs behind the example

The model is the part you see. Behind it is a managed data pipeline that agents built.

  1. Sources

    A funder's own public records.

    • GiveWell grants table
    • 4 cost-effectiveness models
    • Your grants and outcomes
  2. Ingest and catalog

    Agents load every grant and every cell of the models' summary tabs.

    • 375 grants
    • 2,563 model cells
  3. Link and check

    Agents tag grants to programs and rebuild each place's estimate from its inputs. The build stops if any place does not match.

    • 114 places
    • 4 programs
  4. Live application

    The model reads from the checked tables, and every number traces back to its source.

    • Impact model
    • Audit trail

Every figure shows the day the data was read. When a model is updated or new grants are published, each stage can be run again.

Using your own data

The example runs on public data. Here is what changes when you bring your own.

Grants
In this exampleGiveWell's public grants since 2014
With your dataYour grants, investments or programs
Outcomes
In this exampleEstimated lives saved per $1M
With your dataThe outcomes your mission is measured on
Drivers
In this exampleCost per child, coverage and deaths averted
With your dataYour cost, reach and effect assumptions
Places
In this example114 places in 16 countries
With your dataYour grantees, regions or cohorts
Decisions
In this exampleWhere the next dollar does the most good
With your dataWhat to renew, scale or wind down

Your data stays private to your organization. Nothing is public unless you publish it.

Why you can trust the model

An impact number is only as good as the model behind it. This one rebuilds every estimate from its inputs and says whose estimate it is.

114 of 114

Places whose estimate was rebuilt from its inputs

Each one matches GiveWell's published figure. If one did not, the build would stop.

Highest estimate in each program, lives saved per $1M

  • Seasonal malaria pills1,089
  • Vitamin A doses1,050
  • Vaccine incentives731
  • Nets364

These are GiveWell's model estimates, not measured results. This is not an evaluation of GiveWell or its grantees, and governments and other funders work in the same places.

Estimates are labeled
Every outcome figure says whose estimate it is and which version of the model it came from.
Every number is traceable
Each figure links to its source and the calculation behind it.
No causal claims
The model shows what the funder's own analysis estimates. It does not claim to measure what happened.

Build an impact measurement engine for your organization

Start with your grants and the question you need answered.

  1. Connect your data

    Upload your grants and outcome data, or connect your grants system. Agents add the public data.

  2. Ask your question

    Describe the outcomes you care about. Agents build a first version you can use right away.

  3. Review and share

    Check the numbers, comment and approve the model before your board relies on it.