Skip to content

Methodology

The definition, denominator, period and minimum sample behind every figure on the site. If a formula changes, this page and its version number are updated.

Metric definitions

Every rate on brand profiles and in the ranking is shown with its numerator and denominator. Only complaints that passed moderation and have an approved text are counted; drafts, cases under review and unpublished submissions are not.

Response rate
Share of complaints published in the period that received a public reply from a verified brand representative. Denominator: complaints published in the period.
Median first response time
Median time from publication to the first brand reply. The median is used instead of the mean so a single very late reply does not distort the result.
Resolution rate
Resolved / (Resolved + Unresolved). The denominator only contains concluded cases decided by the complainant. Cases marked “No user response” are excluded and shown separately; withdrawn cases never count as resolved.
Satisfaction
Average, sample size and distribution of 1-5 ratings given only by complainants for their own published cases. There are no brand ratings without a case. The rating reflects the complainants' resolution experience, not the brand's overall quality or a payment guarantee.

What does “resolved” mean?

A case only counts as resolved when the complainant confirms it. Even if the brand says “we paid” or “we solved it”, the status does not change without the user's confirmation.

Silence is not a resolution: when the deadline passes the case is marked “No user response” and is excluded from the resolution-rate denominator. If a resolved case is reopened, the previous outcome is not deleted; it stays on the timeline.

Periods and minimum samples

The default period is the last 12 months; the ranking can also be viewed for the last 3 months. The period is based on the complaint's publication date. All times are stored in UTC and shown in Istanbul time.

Rates are not shown without a denominator of at least 5 cases; a satisfaction average needs at least 5 ratings. Below the threshold we show “Not enough data” instead of 0.

Ranking formula

Only brands with at least 5 concluded (resolved or unresolved) complaints in the period are ranked. The others are listed separately under “Not enough data”.

The score is the lower bound of the 95% Wilson confidence interval of the resolution rate (z = 1.96). A 100% rate from very few cases carries a wide interval and the lower bound penalises that uncertainty, so a small sample like 2/2 does not outrank a broad one like 40/45.

p̂ = resolved / n, n = resolved + unresolved, z = 1.96

score = ( p̂ + z²/2n − z·√( p̂(1−p̂)/n + z²/4n² ) ) / ( 1 + z²/n )

1st tie-breaker
Wilson lower bound of the response rate, computed the same way (higher first).
2nd tie-breaker
Median first response time (shorter first; last if there is not enough data).
Full tie
If every measure is equal, brands share the same rank number.

The ranking is recomputed every hour and the page reads that computed snapshot (“last updated”). Formula version: 1. The satisfaction average is shown in the table but is not part of the score.

Verification badges

Each field on a brand's identity card is verified separately and shows its verification date. Unverified information is the brand's own statement.

Legal entity and licence
Record checked against an official source, with the check date. Licence information is not a payment guarantee.
Verified representative
An active representative whose authority to reply for the brand has been reviewed. When revoked, the badge and reply rights are removed immediately.
Sponsor
Shows a commercial relationship; it is independent of representative or licence verification and is not a trust signal.

Corrections and appeals

If you spot an error in a metric or a case, you can report it; corrections are recorded and the snapshot is recomputed when needed. Appeals against moderation decisions and correction rules are explained on the “Appeals and corrections” policy page.