What Counts as a Promise

A statement qualifies as a trackable promise if it meets all three conditions: it is attributable directly to the official (not a surrogate) in a campaign or inaugural context; it is forward-looking (an expressed intention, not a description of existing policy); and it is verifiable — it has at least one condition that can in principle be confirmed true or false.

Statements of value without action components ("I believe in education") are logged but not assessed. They carry no verifiable condition, so they cannot be measured on delivery — but they are not discarded. Each is recorded on the assessment as a Clarity-1 Stated Value, marked Not Converted, and displayed alongside the assessed promises without a delivery record attached. This keeps the value-versus-record gap visible: an official does not bank rhetorical credit for a value simply because it was too vague to measure. Promises contingent on federal action outside the official's jurisdiction are excluded. Where a promise appears at multiple specificity levels, the more specific version is retained.

There is no cap on promise count. All qualifying promises are assessed. Capping promise count introduces editorial bias — selecting 6 from 30 means 24 are excluded on undisclosed grounds. Every assessment displays a mandatory disclosure: N promises tracked · estimated M identified · V stated values logged (not assessed) · X excluded for non-verifiability. Here estimated M is the reviewer's count of candidate promise-statements surfaced in the source sweep before qualification filtering; it is an estimate of the field, not an audited total, and is disclosed to show the share of statements that survived to assessment.

Promise Types

PolicyLogic recognizes three promise types, each with its own delivery scoring track. All three feed into the same D0–D4 delivery buckets — the distinction is in how the bucket is determined.

Assessment Framework

Each promise is assessed on three independent axes. The axes combine into a Promise Record. Promise Records are aggregated into a final Grade. Maximum per promise: 25 points.

Axis 1 — Delivery (max 12 points)

Delivery measures what the official actually accomplished relative to what was promised. It is the primary axis and carries the most weight in the final Grade.

Their Role modifier (0.0–1.0): The official's direct causal contribution to the outcome is assessed and applied to Delivery points before aggregation.

See the Their Role lookup table below for anchor values by office-action combination.

Axis 2 — Difficulty (max 5 points, earned proportionally)

Difficulty rewards ambition — but only when delivery occurs. Zero delivery earns zero Difficulty points regardless of how ambitious the promise was.

Axis 3 — Impact (max 8 points)

Impact measures what was at stake. It is not modified by delivery — a failed promise on a critical issue still carries high impact stakes. This is intentional: if you promise something that matters and don't deliver, the Grade reflects how much it mattered.

Promise Record

Assessment Calculation

The final Grade is calculated from two ratios combined in a fixed 60/40 split. This split is published on every assessment and does not vary by official, party, or jurisdiction.

Worked Example

Their Role — Lookup Table

The Their Role modifier is the single most judgment-dependent element of the methodology. The following anchor values reduce inconsistency across AI runs and human reviewers.

Clarity Scale

Clarity measures how specifically a promise was stated at the time it was made. It is assessed as of the original statement — it cannot be improved retroactively by subsequent clarifications.

Assessment begins at Clarity 2. Clarity 1 exists as a record-only tier: a pure values statement with no action component is logged and displayed but never enters the assessed set, so it receives no Delivery, Difficulty, or Impact points. It is marked Not Converted on the assessment. This is why a Clarity-1 statement can appear on an assessment yet carry no delivery record.

Time Pressure Adjustment

Every promise is assigned a timeline bucket based on the official's stated or implied delivery window. Delivery is adjusted for whether a promise is early, on-track, or overdue. Reductions are a fixed step schedule, not continuous; the maximum penalty is 40%. Being early never reduces delivery — an official is not penalized for outpacing their own timeline.

Behavioral Flags

Behavioral flags override or cap standard assessment when the official actively distorted the promise. Each flag requires a written rationale in the assessment JSON.

When more than one flag applies to the same promise, a cap (maximum) always takes precedence over a floor (minimum), and the lowest applicable cap controls. Example: if Externally Blocked sets a D2 floor and Scope Reduced caps Magnitude while Redefined caps delivery at D2, the delivery result resolves to D2 — the cap binds. A flag that sets delivery to a fixed value (Reversed = D0) overrides both caps and floors.

Transparency Flags

Transparency flags are applied at the assessment level and communicate limitations to readers without altering the result. They appear as labeled badges on the assessment.

Data Sources

• Tier 1 — Primary: Official government records, signed legislation, executive orders, budget documents, Federal Register publications. Required for D3 or D4 delivery results.
• Tier 2 — Authoritative: Major news organizations with named sources, academic and policy research, nonpartisan watchdog reports (CBO, GAO, CRS). Acceptable for confirmation; not sufficient alone for contested outcomes.
• Tier 3 — Supporting: Other credible reporting, official statements, press releases. Requires corroboration.
• Inadmissible: Anonymous social media, partisan advocacy material, press releases from the official's own office as sole support for a positive outcome.

Changelog

Version 2.1 — Clarifications, one substantive rule change, and a terminology standardization. No change to the axis structure, point maxima, or the 60/40 weighting.

• Terminology standardized to neutral, procedural language. "Assessment" replaces "scorecard"; "delivery" and "record" replace evaluative uses of "score." The per-promise total is now the Promise Record and the composite value driving the bands is the Assessment Input. "Grade" is reserved for the single final end-of-term letter. The production banner now describes the pipeline and verification process rather than attributing output to a named model. Neutrality note: the final letter Grade is retained in this version; a delivery-ratio-only presentation was considered and may be revisited, as a letter Grade is the most evaluative-looking element on the page.
• Values statements now logged, not discarded. Pure values statements enter the record as a Clarity-1 Stated Value, marked Not Converted and displayed without a delivery record, rather than being screened out silently. Keeps the value-versus-record gap visible.
• Band boundaries made contiguous. Bands now carry one-decimal edges with explicit inclusive-lower / exclusive-upper convention; Assessment Input is stated as a 0–1 value scaled ×100 and rounded to one decimal before banding.
• Weighting claim corrected. The prior claim that the 60/40 split prevents low-ambition gaming is replaced with an accurate statement of delivery weighting and a disclosed design limit.
• Difficulty–Role interaction stated. Difficulty Earned scales with unmodified Delivery Points; Their Role does not compound into the Difficulty axis. Documents existing behavior.
• Time Pressure schedule rebuilt. Now a fixed step schedule to a 40% maximum, with overlapping boundaries removed and the early-promise penalty eliminated.
• Negative-promise worked example added, covering Their Role on avoidance, the D4 evidentiary standard for proving a non-event, and unmodified Impact on a kept negative promise.
• Disclosure line expanded to report logged Stated Values and to define estimated M.