Back

How We Score

The Republic Score uses an AI-powered, time-adjusted system. We evaluate "are they doing what's expected at THIS point in time?" — not "have they delivered everything?"

Key Concept

A government that's only 9 days old shouldn't get a D for not having delivered everything. The time-adjusted system understands each commitment's complexity and timeline — quick wins are expected in the first week, but structural reforms take years.

1. Every Commitment Has Its Own Timeline

AI classifies each of the 109 commitments into one of 4 complexity tiers:

Quick Win1–30 days

Executive orders, directives, appointments — things a PM can do immediately.

e.g.: Form cabinet with 18 ministries, issue anti-corruption directive

Medium30–180 days

Policy reform, program launches, budget allocations — requires coordination across departments.

e.g.: Launch digital land registry, reform procurement rules

Long-Term180–365 days

Legislation, infrastructure, institutional reform — requires sustained effort over months.

e.g.: Build new highways, pass constitutional amendments, education reform

Structural1–2+ years

Fundamental institutional change — deep reform that takes years to implement.

e.g.: Federal restructuring, tax system overhaul, judicial reform

2. We Measure 3 Types of Effort

Every signal (news article, speech, budget document) is AI-classified into an effort tier:

Intent

Speeches, announcements, plans, committee formations — talk, not yet action.

Action

Budget allocations, policy changes, bills tabled, contracts signed — concrete steps.

Delivery

Infrastructure completed, services launched, laws enacted — tangible results citizens can see.

3. Weights Shift Over Time

In the early days, intent signals (speeches, plans) are valuable. As time passes, only delivery matters.

PhaseIntentActionDelivery
First 2 weeks60%35%5%
Month 1-230%55%15%
Month 3-610%50%40%
Month 6-125%25%70%
Past deadline2%13%85%

4. Every Commitment Gets a Trajectory

Ahead
More progress than expected
On Track
Meeting expectations
Behind
Less than expected
Overdue
Past deadline
Too Early
Not expected to start yet

5. Overall Republic Score

The overall score is a weighted average across complexity tiers. The weights shift based on how far into the term:

First 2 weeks:Quick-wins 70% · Medium 25% · Long 5%
Month 3-12:Quick-wins 10% · Medium 25% · Long 35% · Structural 30%
After 1 year:Long 35% · Structural 40% · Medium 20%
Grades
A ≥80B ≥60C ≥40D ≥20F <20

6. Our Data Sources

AI automatically collects and classifies signals from 80+ sources every 12 hours. No human intervention needed.

80+ RSS feeds
17 YouTube channels
28 Facebook pages
X, TikTok, Telegram, Reddit
Government portals
Parliament records
Kathmandu PostRepublicaKantipurOnlineKhabarSetopatiNepal PressKhabarhubNews24NTV+70 more

Data Confidence

We don't show scores without sufficient verified data. "Too early" commitments score 50 (neutral) and are excluded from grading.

Sufficient
50+ signals
Partial
15-49 signals
Insufficient
<15 signals

Transparency

  • All scoring code is open-source — anyone can inspect the logic
  • AI models: Qwen 3.6+ (free classification), GPT-4.1-mini (deep analysis)
  • Score auto-recomputes every sweep (2x daily)
  • Admins can override AI timelines — overrides are flagged
  • Total AI cost under $5/month (free models used first)

Current Limitations

  • AI occasionally misclassifies signals — admin review corrects this
  • Budget execution data is still limited
  • Citizen voting just launched — sample size is growing
  • Some commitments have "insufficient" data confidence — more signals needed
📴