Methods

Every measure follows the same discipline, then documents its own specifics.

What qualifies as a measure

• An official promise, citable to a primary document.
• A numeric target and a deadline.
• Public data to score it with.
• A metric defined before the outcomes are known. Each measure's methods page is published with its baseline, not after the results.

Shared rules

• Comparisons are like-for-like: year-over-year for seasonal data, target-vs-actual for countable commitments.
• Status vocabulary is fixed, including the words for falling short: met, on track, early, baseline-setting, announced, behind, missed. No grades. The negative statuses are pre-registered now, before any results exist, for the same reason the metrics are.
• Every methodological judgment is documented on the page it affects.

Per-measure methods

• Buses - speed baselines, weighting, exclusions, and how each corridor was matched to bus routes, with sources.
• Childcare - pre-registered before first data (September 2026).
• Groceries - rules locked before any store exists (first scheduled 2027).
• Heating oil - permit counts from the city's environmental database, registered July 2026.
• Noise cameras - camera counts from DEP annual reports, registered July 2026.

The Record

Past promises are scored with the same rules on the Record. One difference: a metric can't be registered before results that already exist, so each Record metric is defined from the promise's own text before the score is computed, and the definition is published with the page.

The data itself

The datasets this site builds are public, including a queryable API over 1.1 million NYC buildings. Documented on the data page.