1 · Define the problem
One problem, stated so precisely that a stranger could measure it. The advisor on the right reads as you type.
State the gap, not the missing solution. “District hospitals lack surgeons” already assumes the answer is surgeons.
2 · Evidence
The advisor scores structural strength, numbers, sources, baselines, not factual truth. Verifying the facts remains your job.
3 · Instrument & theory of change
The instrument is the vehicle. The advisor checks whether the vehicle you chose can carry the problem you defined.
4 · Delivery chain & first 90 days
A plan without named owners is a wish. Every step needs a person, not an institution.
5 · Review & export
This is the working draft of your Implementation Compact, with the advisor’s findings attached. Export it, then take it to your ministry sponsor.
Concepts library
The full set of concepts the advisor can raise, ordered by how often they kill real policies. Open any row for the full card. Read them before you draft: most failed health policies fail in one of these ways, and none of them is new.
The implementation gap
Policy is remade at the point of delivery. Nurses, clerks and district officers with too little time and too many rules will adapt, ration and reinterpret whatever the circular says. Lipsky called them street-level bureaucrats, and their coping decisions, repeated daily, become the real policy.
Before you design, consider
- Ask what your instrument costs the person who must comply, in time, money and risk, and fund that cost explicitly.
- Pair every mandate with a support or enforcement mechanism you can actually staff.
- Pilot the paperwork: if the form takes 20 minutes, the form is the policy.
How the advisor works, and what it cannot do
Read this once. It is short and it is honest.
What it does. The advisor reads your text for structural signals: whether the problem names a population, a place, a magnitude, a timeframe and a data source; whether your evidence contains numbers, rates, baselines and named sources; whether the instrument you chose is the kind of instrument that usually carries your kind of problem; and whether your wording matches patterns that has repeatedly punished, production without absorption, targets without safeguards, incentives that reward the wrong thing.
What it does not do. It does not verify facts. A confident false number scores as well as a true one. It does not know your country's politics, and it cannot see the room you will present in. It flags questions, not verdicts, a flag means "policy experience says look here," never "this is wrong."
Why rule-based. The advisor runs entirely in your browser. Nothing you type leaves this page, no server, no account, no analytics. That is deliberate: fellows work on politically sensitive national material, and a tool that uploads it is a tool ministries would rightly ban. The cost of that privacy is that the advisor is pattern-matching, not reading. It will occasionally flag something irrelevant and miss something important. Your cohort, your faculty and your ministry sponsor remain the real reviewers.
Saving. Your draft saves automatically to this device only. Export to Word before switching devices.