How the Kindness City Index Is Calculated

Methodology version 6

The Kindness City Index ranks American cities by verified acts of kindness, adjusted for the size of each city. A town of 2,000 and a city of two million each earn a fair chance to rank. This page explains each step, each term and the reason behind it.

We count verified acts of kindness

Definition. An act is a specific contribution to a community, such as a funded program, a donation or a rescue, with a date and a place. An act counts once, however many posts describe it.

Rule. Before an act counts, it passes six screens. It is documented; it serves others; it reaches people in the city credited; it occurred within the past 36 months; it is free of unresolved controversy; and any foundation behind it is validated.

Why. The screens keep the Index anchored in real, local and recent service. Anyone can trace each act to its source.

Stronger evidence earns more credit

Rule. An independent record, such as confirmation from the school or clinic that received the help, earns full credit. Two independent news reports earn 90%, one earns 75%, and material a person or foundation publishes about itself earns 60%.

Why. Proof matters. Independent confirmation is the surest evidence that an act happened as described, so it carries the most weight.

Recent acts count most

Rule. An act earns full credit for 12 months and half as much in each year after, until it leaves the Index at 36 months.

Why. The Index reports kindness happening now. It also honors service that continues to shape a city over the years.

Community response shows how far kindness travels

Definition. Resonance measures the community's response in the first 24 hours. It combines Lift and Spread in equal parts, each ranked against acts from cities of similar size.

Lift. Lift measures how strongly people respond. Likes count once, comments twice and shares three times, because shares carry an act to new audiences. The total is divided by the posting account's usual engagement: a neighborhood page that usually draws 40 interactions and draws 400 has a Lift of 10.

Why Lift. Measuring each post against its own usual audience lets a small community page compete with a large television station.

Spread. Spread measures how widely an act travels: the effective number of local accounts engaging with it. Ten outlets sharing engagement evenly count as ten; ten outlets echoing one dominant post count as about one.

Why Spread. One account is a mention; many accounts are a movement. Kindness that travels through many local voices reaches more of a community.

Each act earns a weight

Rule. Each act's weight equals its evidence credit, times its recency credit, times 0.5 plus its Resonance. An act with typical community response counts as one act, a standout as up to 1.5 and a quiet one as 0.5.

Status of Resonance. Future feature. Resonance is not in the current computation: every act's weight today is its evidence credit times its recency credit, with Resonance held at the neutral 0.5. See Implementation notes.

Example. A recent act confirmed by two news reports, with Resonance of 0.9, weighs 0.90 × 1.00 × 1.40, or 1.26 acts.

Why. Quiet acts count. Community response adds credit, within firm bounds, so a single post stays within its share of a city's score.

A neighborhood of helpers outranks a single prolific poster

Rule. Each scored person counts toward the city, and each person contributes up to three acts: their three strongest.

Why. Breadth wins. A city's rank reflects how many of its residents serve, and prolific individuals earn their own recognition on the Kindness Leaders board.

Cities are compared per resident, adjusted for size

Rule. A city's score is its weighted acts per 100,000 residents, adjusted with a cushion equal to five typical acts at the national rate. The cushion steadies small towns, where a single act would otherwise swing the rate.

Example. A town of 2,000 with one act weighted 1.26 has an unadjusted rate of 63.00 per 100,000 residents. With the cushion, its rate is 3.10, strong enough to lead its peers and measured on the same footing as large cities.

Why. Small numbers mislead. Statisticians have long observed that small populations swing to both extremes by chance, and the cushion gives credit for real kindness while holding chance in check.

Cities compete with peers of similar size

Rule. Cities rank within three size groups: Big Cities, with 500,000 residents or more; Mid-Size Cities, with 100,000 to 499,999; and Hometowns, with fewer than 100,000. Within each group, the top 10% are Leading, the next 20% Strong, the next 30% Growing and the rest Emerging.

Why. Each group crowns its own leaders. Fans follow a fair race among peers.

People and single acts earn their own recognition

Kindness Leaders. Kindness Leaders ranks people by their weighted acts over the past 12 months, with athletes and foundations on one board and local heroes on another.

Moments. Moments celebrates a single act that catches fire, with a Rising badge on the map and a chance at Moment of the Week.

Kindness Ratings. Player cards show each person's Kindness Rating, from 0 to 99. A private individual's rating appears only with that person's written consent.

Status of Kindness Rating. Future feature. No Kindness Rating is computed or displayed today, on any card. See Implementation notes.

What stays outside the score

Rule. Cheers do not affect the Kindness City Index score or city rankings. However, in Kindness Matchups, each cheer earns one matchup point, subject to the per-game cheer cap. Cheers also help determine the weekly Fan Favorite. Sponsors receive visibility, and rankings remain independent of any sponsor.

Safeguards protect the Index

Rule. Engagement counts only from organic posts by accounts at least 90 days old. Lift divides by the posting account's usual engagement or 20 interactions, whichever is larger, so a tiny account's single burst stays in proportion.

Engagement-integrity review. Before publication, editors review the engagement behind acts at the top of the rankings for bots and bought engagement. The review covers acts above the 99th percentile of Lift, the acts behind each Leading city, the top of the Kindness Leaders boards and each Moment of the Week.

What reviewers check. Reviewers examine the age and history of engaging accounts, sudden bursts of activity, repeated or generic comments, and engagement arriving from far outside the community. Editors also check each video for staging.

Outcome. Engagement judged inauthentic is removed, and the act's Resonance resets to the neutral 0.5. Deliberate manipulation removes the act from the Index. Each decision is logged, and the person or organization may request a second review.

Why. Leading ranks carry the most influence and attract the most attempts to game them, so they receive the closest scrutiny.

Updates, versions and ties

Rule. The Index refreshes each Monday. Each change to this method carries a version number, an effective date and an entry in the public change log. Cities with equal scores rank by the number of residents with verified acts, then alphabetically.

Methodology version 6, effective November 7, 2026.

Definitions

Kindness Matchups

Kindness Matchups tie the Index to the NBA schedule. Each scheduled game pairs the two teams' home markets. From 48 hours before the start through 48 hours after the final whistle (counted in calendar days, as the note below explains), each side earns 10 points per verified act, 3 points per nomination and 1 point per cheer. Cheer points are capped separately for each city in each game at that city's act and nomination points combined. Totals are reported per million residents, so a large city and a small one compete on even terms. The higher score takes the Kindness win. A scoreless game, with no points earned on either side, is a draw, and no tie-break is applied. If both cities scored and matchup points per million are equal, the city with the higher published Kindness City Score wins; that score is a separate Index measurement, not points earned in the matchup. If the Index scores are also equal, the city with more verified acts in the window wins; if still equal, the matchup is a draw. Results post as provisional when the window closes and become final at the Monday 12:00 PM ET processing cutoff if the window has closed by then; a window closing after that cutoff becomes final the following Monday. The data counted in a result is frozen at that cutoff: a nomination counts only if it was approved by the cutoff, cheers are the counts captured across the window, and a verified act counts only if it is in the verified record at that cutoff; an act verified later does not change the final (it counts toward the Index and toward any not-yet-final matchup whose counted days contain its date), unless a documented correction is issued under the revision policy below. The Index scores recorded with the result are the ones published on this site when the result is finalized. If processing runs after the cutoff, it publishes the frozen result, and the Matchups page shows both the scheduled cutoff and the actual finalization date. Official season records count final results only.

Matchups run under their own rule version, currently 48-48-v11-cities, with each version recorded in the change log below, alongside Index Methodology version 6.

Two notes on how the count runs on this site. The Matchups page displays each active window's opening time and estimated closing time. The underlying act records carry a date, not a time of day, so eligibility is applied by America/New_York calendar day: the two days before the game, game day, and the two days after it. An act on a counted day counts in full, whatever the hour. Nominations are submitted on the Matchups page during an open window, with the date and place of the act, its city, evidence and the submitter recorded; once an administrator approves a nomination, it scores 3 points for its city in the matchup whose counted days contain the date the act happened, provided the approval lands before that matchup is finalized (a later approval is recorded but does not change a final result), and an act that is independently verified joins the verified record and scores 10 points as an act instead, by the same date. A nomination reports an act that has already happened (a future act date is rejected at submission), and once approved it is credited to every matchup whose counted days contain the act's date and that has not yet finalized, including a matchup whose window opens after the nomination was sent. Duplicate nominations are not counted. One separation matters: Matchups count each qualifying verified-act record at a flat 10 points, whatever its Index weight. The evidence, recency and resonance weighting in this methodology, and its limit of three acts per person, apply to the Kindness City Index only and never change a matchup score; the same verification standards decide what enters the verified record for both. Current fixtures, results, scoring breakdowns and season records are on the Kindness Matchups page.

Three integrity rules run under the count. Winners are decided on full-precision points per million; figures are rounded to one decimal for display only, so two cities that both display 0.0 are not treated as tied unless their exact values are equal. When the Index tie-break decides a matchup, the Index scores used are the ones published at finalization and are recorded with the result, so a later Index refresh cannot change it. An act counts in every matchup whose counted days contain its date, so one act can score in more than one matchup when windows overlap; the same applies to nominations, which follow the date of the act they report, and cheers are counted separately inside each matchup's own window, so a cheer in an overlapping span can count in both, under each matchup's cap. And a final result is never silently changed: if a scoring rule changes after finalization, the result is marked Revised and shows its original score, its revised score, the revision date and the reason, with the rule versions used.

Population source

City populations are U.S. Census Bureau city and town population totals from the April 1, 2020 estimates base, as published in the Bureau's Vintage 2023 release (the City and Town Population Totals tables). Matchups use city-proper populations, not metropolitan-area populations, on the same basis for all twelve league cities. The population figures are fixed for the season and are not updated mid-season. Later Census vintages revise the estimates base slightly (Vintage 2025 lists Houston at 2,299,649); the season keeps the Vintage 2023 figures below so no denominator changes mid-season.

Implementation notes: how the published scores are computed

The cushion, exactly. A city's Kindness City Score is Score = (W + 5) / (P + C) x 100,000, where W is the city's weighted acts, P is its population, and C is a cushion population, C = 5 / R x 100,000. R is the national rate: the total weighted acts in all ranked cities divided by their total population, times 100,000, recomputed at every weekly refresh. At the current refresh R is 0.595590 weighted acts per 100,000 residents and C is 839,504 residents. Worked example, Houston: W is 95.70, P is 2,300,833, so the score is (95.70 + 5) / (2,300,833 + 839,504) x 100,000 = 3.21, the published figure. The town example in the text above uses an R of 2.5, which makes C exactly 200,000: (1.26 + 5) / (2,000 + 200,000) x 100,000 = 3.10. The cushion C is the same for every city; it does not change with a city's population, and its steadying effect is largest on the smallest populations, which is its purpose. Figures on this page are rounded for display; the calculations themselves use full precision. Every act behind these totals, with its evidence credit, recency multiplier, weight and whether it counts toward its city, is published on the Act-Level Audit page.

Resonance today. Resonance is held at the neutral value 0.5 for every act in the current implementation, because the first-24-hour response data it is built from, Lift and Spread, is not published by the platforms the acts are sourced from, and this site does not estimate it. Each act's weight today is therefore evidence credit x recency credit x (0.5 + 0.5), which is evidence credit x recency credit. The Lift and Spread definitions above describe how Resonance will be computed when that response data is collected; until then no published Index score depends on them, and every score on this site can be reproduced from the evidence and recency rules and the cushion equation above.

Cheers. One counted cheer per visitor per city, enforced on the server through a long-lived anonymous cookie: a second click withdraws the cheer rather than adding another. The cookie identifies a browser, not a unique person: two people sharing one browser count once between them, and one person on two browsers can count twice; no account or identity check stands behind it, and the per-game cap in Matchups bounds what that can add. Every cheer and withdrawal is stored as a timestamped event (time, city, plus or minus one; no identity), so each matchup sums exactly the events inside its own window, including across overlapping windows; counts are also snapshotted weekly. Cheers never move the Index ranking; in Matchups they earn points only under the per-game cap.

Kindness Rating. No Kindness Rating is computed or displayed anywhere on this site today. The 0 to 99 definition above is retained for a future player-card feature, and no rating will be shown until its calculation is published here.

Evidence and attribution audit (November 2026). Every act behind the Index was re-checked against its source: what happened, where it happened, who benefited, and whether the source supports it. A city is credited only for a documented local activity (class A) or a documented direct benefit to its residents (class B, including a metro community of that city). A person's team, hometown, residence or organization is never by itself a reason to credit a city (class D), and an act whose location cannot be established is held out of the scores until it can be (class E). Of 280 records reviewed, 24 were re-credited to the city where the act actually happened or whose residents it benefited, 18 were excluded (honors for work done elsewhere, activity in another country, and nationwide programs with no city beneficiary), 5 were duplicates of another record and were consolidated, and 5 are pending further evidence; 252 acts are counted. Acts are also labelled by date precision: an act known only to the month or year still counts toward the Index but is never assigned to an individual Matchup window. Six cities whose acts all failed the attribution test (Greenwich, Indianapolis, Minneapolis, Newport, Sacramento and St. Louis) currently hold no counted acts; their published scores are the cushion floor the formula gives every ranked city, and they will rise only when a qualifying act is verified. One population was added: Newport News, Virginia (186,247, U.S. Census Bureau, April 1, 2020), after an act credited to "Newport" was found to have taken place in Newport News. No Matchup final changed: none of the removed or re-credited acts fell inside a final's counted days, and the corrected record stands as rule 48-48-v10-evidence under the revision policy. Every record, its class, its evidence and the reason for each correction is published on the Act-Level Audit page (snapshot evidence-audit-2026-11-10).

Change log