FeudFix Intelligence Index — Methodology

Every metric is derived from a single source: anonymized Resolved Case Summary records generated when a FeudFix user confirms their dispute has been resolved. Each summary is produced by an AI model from the case's correspondence trail, then stripped of all customer PII (names, emails, addresses, phone numbers) before storage.

A Resolved Case Summary captures only:

• A neutral 2–3 sentence description of the dispute pattern (no PII)
• The confirmed resolution method (refund, replacement, repair, credit, apology, no resolution, other)
• A neutral 2–3 sentence outcome summary (no PII)
• Short key factors that influenced the outcome (no PII)
• The resolution date and an inherited quality score

Timing and dispute-category data are pulled from the originating Case record (case creation date and product/service field). No raw case documents, correspondence text, or customer-identifying information ever enters the index calculation.

A score is never published for a company with fewer than 5 resolved-case precedents. Below that threshold the company's confidence level is set to insufficient_data, all metric fields are set to null, and the public Intelligence Index directory excludes the company entirely. This prevents unreliable scores from being displayed or indexed by search engines.

Once the sample-size threshold is met, the following are computed over up to the 200 most recent resolved-case precedents for that company:

Resolution Rate

round( (resolved count ÷ sample size) × 100 )

The share of precedents whose resolution method is anything other than no_resolution, expressed as a percentage. A case resolved by refund, replacement, repair, credit, apology, or "other" all count as resolved.

Dispute Difficulty Score

max( 0, 100 − Resolution Rate )

A 0–100 inverse of the resolution rate: a company that resolves 80% of disputes scores 20 (easier); one that resolves 30% scores 70 (harder). The score is descriptive of observed outcomes, not a judgment of company conduct.

Median Resolution Time

median( days from case creation to resolution )

The median number of days between the originating case's creation date and the resolved-case summary's resolved date. Cases missing either date, or with a negative elapsed time, are excluded from this calculation. Reported in whole days.

Most Common Dispute Category

mode( case.product_or_service )

The single most frequent product/service value among the company's precedents. Blank values are ignored. If there is a tie, one is selected deterministically by sort order.

Confidence reflects how much observed data backs a published score. It is based solely on sample size — not on the score's magnitude or the company's reputation.

The index for a given company is recomputed in two situations:

• Real-time on resolution: the moment a user confirms a case as resolved (or reopens one), the company's metrics are recalculated immediately, so public pages reflect the change without a manual refresh.
• Daily catch-up: a scheduled job recomputes every company with at least one precedent once per day, to pick up any changes not captured by real-time events.

The "Last Updated" timestamp on a company profile reflects the most recent successful recalculation.

• No resolution outcomes count against the rate. A case marked resolved with a method of no_resolution is included in the sample size but excluded from the resolved count, lowering the resolution rate. This reflects the observed reality that the dispute was closed without a remedy.
• Reopened cases are removed. If a user reopens a previously resolved case, its precedent summary is deleted and the company's index is recomputed — the case no longer counts toward the metrics.
• Missing timing data is ignored. Precedents whose originating case lacks a creation date, or whose resolved date predates the creation date, are excluded from the median-time calculation but still count toward the resolution rate and difficulty score.
• Self-reported outcomes. Resolution method is chosen by the user who owned the case; FeudFix does not independently verify it. Users have no incentive to misreport, but the data is observational, not audited.
• Sample bias. The index reflects disputes that were brought to and resolved through FeudFix — not all disputes a company handles. A company that quietly resolves most complaints before a customer finds FeudFix may appear more difficult than its overall record suggests.
• Not a legal or credit rating. The Difficulty Score is a descriptive statistic about observed resolution outcomes among FeudFix users. It is not a measure of legality, compliance, financial health, or good faith, and should not be cited as such.