Rating methodology

Data sources

Every rating starts from data published by the financial regulator of the country in question: official rates, registries of authorized entities, maximum permitted fees, and monthly bulletins. When a data point comes from a secondary source, we mark it explicitly with the source URL and the consultation date. We do not use unverifiable data, we do not accept payment for inclusion, and we do not suppress legitimate products at the request of the rated entity. Each cluster (microcredit, interest rates, insurance, calculators) carries the dataObservationDate field, which records when the data was observed at its original source.

Scoring weights

The final score combines five dimensions: real total cost (35%), regulatory transparency (20%), channel accessibility (15%), documented approval speed (15%), and verifiable reputation (15%). Real total cost uses the comparable rate each regulator requires — TEA in Ecuador and Peru, CET in Brazil, CAT in Mexico, TCEA in Peru, CFT in Argentina, CAE in Chile — so cross-product comparison is mathematically consistent. Regulatory transparency measures whether the lender publishes all costs before disbursement. Accessibility evaluates geographic and digital coverage. Speed uses self-reported times verified by test applications. Reputation weights formal complaints recorded with the regulator.

Conflict-of-interest policy

Comparamigo may receive a commission for applications referred to some rated lenders. That fact is declared on every comparison page with the notice "Comparamigo may receive a commission for this referral". The commission does NOT affect the score: the scoring algorithm receives no economic input — only regulatory and verifiable data. To audit this, the scoring code lives in pure modules with automated tests that run on every deploy. There is no code branch that modifies the score based on commission.

Review cadence

Each cluster has a documented review cadence and a public dataObservationDate field. If the observation date is older than 90 days, a WARN alert appears in the editorial dashboard; if older than 180 days, the cluster fails the freshness deploy gate and the page is flagged "data outdated" until refresh. This runs automatically in every build via scripts/seo-freshness-check.mjs and covers microcredit, rates, insurance, and calculators. Target cadence is quarterly for rate data and annual for regulatory methodology, unless the regulator publishes an update sooner.

Correction procedure

When an error is detected — whether by internal review, by a rated entity supplying evidence, or by an applicant reporting an incorrect data point — the flow is: (1) verify against the primary regulatory source; (2) correct the affected page with a date stamp; (3) public entry in /about/corrections summarizing the error and the correction; (4) notify the reporter if contact details were left. Corrections are never silenced. The /about/corrections page is the public log of the process.

Algorithmic transparency

The code computing the scores is Comparamigo proprietary, but its logic is documented on this page and in /about. We do not use opaque models or unintelligible black boxes: every criterion reduces to a formula expressible in the public data we cite. If you request the exact computation detail for a specific page, we can supply the number-by-number breakdown — contact is in the /about footer. The methodology is updated only via public correction in this document; there are no silent updates.

Editorial review methodology (v1.0)

Editorial lender reviews use their own methodology, complementary to the general framework described above and published as a SoT in src/config/methodology_v1.json. Every lender receives a score from 1.0 to 5.0 across four weighted criteria: Transparency (30%) — clarity of rates, fees, and APR disclosed before signing; Speed (25%) — declared and observed time from application to funding; Support (25%) — availability and response times of the support team plus the lender response rate to public reviews; and App Quality (20%) — stability and verifiable ratings on the App Store and Play Store. The overall score is composed as a weighted average and translated to a visible tier: Excellent (≥4.5), Good (≥4.0), Fair (≥3.5), Poor (≥3.0), Bad (<3.0). When a criterion lacks verifiable public evidence, it is marked "unrated" and excluded from composition — we prefer an incomplete score to a fabricated one. Two editors must score independently before publication; discrepancies greater than 0.5 trigger a third review.

FAQ

• How do I know Comparamigo is not biased by the commission?

  The scoring algorithm does not receive commission information; it operates only on regulatory data. The code branch is public and audited by automated tests. If you find something inconsistent, the correction procedure is described above.

• Why does Comparamigo sometimes not rate a lender?

  For one of three reasons: the lender is not in the public registry of its country regulator, published data is insufficient to apply the methodology, or the last observation is older than 180 days. We mark this explicitly in all cases.

• Can entities pay to improve their rating?

  No. This is prohibited by our conflict-of-interest policy and verified by the architectural separation between the scoring module and the commission module. Any documented attempt would be published in /about/corrections.

• How is the methodology itself updated?

  Updates to this methodology are published as a new dated version on this page, leaving the previous version visible. The policy applies retroactively to all pages until the next recompute.

• Can I request the breakdown for a specific page?

  Yes. The breakdown is available on request to the editorial contact in /about. This is part of our algorithmic transparency policy.