The Weight

Why one thing is at the top of your list and another is halfway down.

Deciding what goes first is the only editorial power this thing has. So the rule is written down, it runs the same way for everybody, and every number that goes into it comes from a document you can go and read.

Five factors, multiplied. Multiplication matters: any one of them going near zero pulls the whole score down. That is deliberate. A rumour with nowhere to take it should never out-rank a documented finding with a vote on Tuesday, however many people it would touch.

WEIGHT = HARM × (1 + REACH) × EVIDENCE × WINDOW × LEVERAGE

1. Harm

How bad it is for one person. Two parts: what kind of harm, and whether it can be undone.

What is at stakeWeight
Liberty — detention, deportation, custody10
Body — force, medical neglect, death10
Home or livelihood — eviction, job, licence7
Due process — a case closed, a hearing denied6
Privacy and dignity — surveillance, records4
Can it be undoneMultiplier
Permanent — deportation, death, a building demolished×1.0
Lasting — a record, a firing, an eviction×0.6
Recoverable — a fine, a delay, a fee×0.3

Irreversibility is the heaviest idea in the formula. A thousand parking fines are not one deportation, and no amount of scale should be able to say otherwise.

2. Reach

How many people, on a log scale — ten thousand people matter more than ten, but not a thousand times more. Then a concentration multiplier between 1.0 and 2.0: is this landing on one group far harder than on the county as a whole?

REACH = log10(people + 1) × concentration

Concentration is what makes this a civil-rights instrument rather than a complaints board. A policy that touches everybody a little and one group enormously is not a neutral policy, and the ranking should say so.

3. Evidence

The same ladder used everywhere else in the system. This one is a gate, not a nudge.

GradeMultiplier
In the record — the institution's own document says it×1.0
Corroborated — two independent sources agree×0.8
Alleged — in an official complaint, not proven×0.5
Contested — under review, disputed on the record×0.4
Unverified — we have not checked it×0.15

At 0.15, an unverified claim cannot reach the top of anybody's list. That is the point. The fastest way to discredit a movement is to send people out repeating something that turns out to be false.

4. Window

Is there a lawful way in, and when does it shut? This is the factor nobody else computes, and it is the difference between being informed and being useful.

WindowMultiplier
Open, closing within 14 days×1.0
Open, closing within 45 days×0.7
Opens later, date known×0.4
Nothing scheduled×0.15

Something terrible with no open window drops to Watch — and the app says so plainly instead of manufacturing urgency. Turning people out for a meeting that is not happening is how movements burn through the people they need.

5. Leverage

Would showing up plausibly change the outcome? Four sub-factors, multiplied together.

FactorRange
Margin — was the last comparable vote close, or lopsided0.4–1.0
Grantable — can this body actually give what is being asked, or only forward it0.5–1.0
Reachable — elected and local, appointed, or a federal contractor0.4–1.0
Crowding — nobody working it is a bonus, a crowded fight is a discount0.8–1.2

What comes out

A number from 0 to 100, scored against the strongest live item on your list — not against all imaginable injustice, which has no denominator. The number decides the order. What you see is a band and the reasons behind it.

BandMeans
Act nowOpen window, solid evidence, and your presence plausibly matters
Act this monthReal, with a door that is open but not closing this week
WatchSerious, but there is nothing to attend yet. We are holding the clock

You are never shown a bare decimal as though it were a measurement of suffering. A score of 82.4 implies a precision nobody has.

Rules the formula is built to keep

It ranks decisions, never people. Every row is an institution acting in its official capacity. There is no score attached to a private individual, ever.
Unverified cannot rise. The evidence gate is multiplicative, so no amount of scale or outrage lifts something we have not checked.
No window, no mobilisation. Where there is nothing to attend, the app says there is nothing to attend, and offers the two things that can be done without a meeting.
Scale never buries irreversibility. Harm is multiplied, not added, so a permanent consequence to a few outweighs a recoverable one to many.
Every input is checkable. People count, evidence grade, window date, vote margin — each traces back to a document with a date on it.

Why it is published

The instinct is to keep this secret, because it is the most valuable thing here. It should be public anyway.

A ranking that decides where a civil-rights movement spends its Tuesday nights will be accused of bias the first week it matters. The only durable answer is: here is the formula, here are the inputs, check our arithmetic. A secret score invites the accusation and has nothing to answer it with.

And the formula is not really the moat. Anyone can copy these numbers in an afternoon. What is hard to copy is the thing underneath it — the daily record of who decided what and when, the window calendar built county by county, and the calibration: the accumulating history of which predicted-leverage scores actually moved a vote. The weights get better every time an outcome comes in. That history is the asset. The equation is just the part you can read.

Working draft. Numbers in the demo are illustrative; the structure is not.
← Back to the app · Everything else