> For the complete documentation index, see [llms.txt](https://ambitious-impact.gitbook.io/ambitious-impact-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://ambitious-impact.gitbook.io/ambitious-impact-docs/other/our-prioritization-funnel/stage-3-shallow-dive.md).

# Stage 3: Shallow dive

In this stage, a reviewer will examine an idea for one and a half to two hours, focusing on its evidence base, chance of success, and cost-effectiveness.&#x20;

Depending on the number of ideas available and team capacity, this exercise aims to select between 20 and 35% of the ideas for a narrower shortlist.&#x20;

At this stage ideas should not be discarded without spending 2-5 minutes thinking if there is a close alternative idea that could work.&#x20;

If an idea does particularly badly at a specific earlier aspect of this stage (e.g., the strength of evidence base), it may be worth discarding without continuing the research.&#x20;

{% stepper %}
{% step %}

### Strength of evidence base&#x20;

The template explains much of this process, plus the criteria.&#x20;

The main reviewer for an idea should spend no more than 50 minutes reviewing the evidence base for an idea. They will find evidence to support and evidence against the promise of an idea – all should be noted.&#x20;

A second researcher ensures the quality of the evidence review by spending 5 to 10 minutes providing feedback on scoring, the process followed, alternative views, areas for future research, etc.&#x20;
{% endstep %}

{% step %}

### Quantitative modeling&#x20;

Design a CEA / BCR spreadsheet template. For comparing interventions of different types, we should hold some factors that are highly uncertain and highly likely to affect the BCA constant. Here are examples from different rounds ([1](https://docs.google.com/spreadsheets/d/1yvvB_TF4LMA-KvUYAX6khZAv42AWv9WdN7lTt9wwMz0/edit#gid=1094884305), [2](https://docs.google.com/spreadsheets/d/1tRtVNbEoqAbbdRI620YXjVMHluaJWoUEuTwyKSoJx6A/edit?usp=drive_link), [3](https://docs.google.com/spreadsheets/d/1k1iFwNksekDiKHWOJG3KmHeUN_oTi36Ym2V3MRHrH1k/edit#gid=194345489)).&#x20;

We can allow some of these factors to change with good reason. For example, scale up costs by a % for international or developed country interventions. Aim to avoid the CEAs varying by orders of magnitude based on speculative numbers.

* We should pre-set options where possible to support easy decision-making.
* For key factors to consider, see [CEA Uncertainties](https://docs.google.com/spreadsheets/d/1mmvIFZiTgi4bGOZIbOwqSIs3zmkEvV1AuzE5IqEKWlc/edit#gid=943856405) and see existing CEAs
* The [2024 human-focused template](https://docs.google.com/spreadsheets/d/1k1iFwNksekDiKHWOJG3KmHeUN_oTi36Ym2V3MRHrH1k/edit#gid=194345489) is split into two sections: health interventions and (economic) consumption interventions. Each has multiple suboptions
* Health ideas – impact can be modeled using:
* Deaths averted: Interventions that reduce mortality
* Prevalence: Interventions that are assumed to reduce the prevalence of a condition, especially conditions that cause disability, typically as a result of treatment
* Incidence: Interventions that prevent new cases of a disease
* Other (e.g., total DALYs)
* Consumption ideas – impact can be modeled as:
* One-off interventions (usually aiming for a policy change) vs continuous interventions (e.g., continuously delivering a service)
* ‘Top-down’ (where the beneficiaries are a whole group of people, such as a city or a state) or ‘bottom-up’ (where the number of beneficiaries grows proportionately with the charity’s activities)
* Having only immediate effect (while the intervention lasts) vs having a longer-lasting effect (e.g., after a policy is passed or after skills are successfully built).

As a baseline comparison, include an example cost-effectiveness estimate of existing promising interventions, e.g., FWI 2.0 or corporate campaigns for egg-laying hens in HIC.&#x20;

A second researcher reviews.

If ideas do significantly worse than the comparison baseline. Assess if there is a key uncertainty or reframing of the idea that could be useful to research that might make the idea look more promising and research that. Or cut the idea.
{% endstep %}

{% step %}

### Decision meetings and additional research

The round lead may hold meetings across stage 3 to advance the timeline and eliminate ideas earlier that clearly fail the bar in one of the three aspects being examined.&#x20;

The round lead (or others) will chair a final decision-making meeting to reduce the list to between 20% and 35% of its current size. Scores from all three aspects are included in a spreadsheet and discussed to lead to conclusions for each idea. Depending on the topic area, scores for each sub-stage may be weighted differently (E.g., you may want to weight the one-line CEAs less than the evidence review, or vice versa).&#x20;

Additional research: For each “probably yes,” aim to resolve the key uncertainty with <2h of research.
{% endstep %}
{% endstepper %}

<br>


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://ambitious-impact.gitbook.io/ambitious-impact-docs/other/our-prioritization-funnel/stage-3-shallow-dive.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.