Credit Score Research Methodology

1. Scope and collection period

The study describes self-reported answers submitted through the Credit Score Check & Report AI app. All valid submissions in the observed collection period were retained for pooled analysis. Timestamps are interpreted with a UTC assumption; the originating timezone is unverified.

Results are pooled across July 2025 - November 2026 rather than presented as time trends. November coverage is partial at the analysis snapshot.

No attempt was made to infer unique people. Repeated answers, intentional caching and experimentation can generate repeated logged submissions.

2. Shared questions and response mapping

Shared questionnaire prompts and predefined answer options were reviewed locally. The observed questionnaire included seven main questions and three additional questions in some submissions. The analysis uses exact, manually curated question-specific mappings. These are observational research mappings, not a verified official definition of every historical questionnaire version.

Different or ambiguous wording is not merged merely because it appears similar. Missing, malformed, conflicting, unresolved and unrecognized responses are excluded from the eligible denominator for the affected question or combination. Their submissions remain in the overall collection.

In particular, mapped answer bands should not be read as exhaustive measurements of a continuous range. A missing band in the observed option set is not evidence that no submissions would fall within it.

3. What the three percentages mean

Each numerator counts submissions selecting the relevant mapped option or options. Each denominator includes only submissions with valid mapped answers to every question used in that finding.

Percentages are calculated as numerator divided by the eligible total and rounded to two decimal places. Because eligibility differs, the three percentages are not parts of a single whole and should not be added together.

The combined findings require all named answers to occur in the same submission. They do not imply independent behaviors, causal relationships or statistical significance among unique people.

4. Repeats, caching and completeness

Identical questionnaire payloads are expected because the app uses caching and may be used repeatedly. Submissions were not deduplicated based on answers, estimated scores or identical payloads.

Having a frequently observed set of questions does not prove that a submission matches a complete official questionnaire. The study uses the valid-response eligibility rules above, rather than claiming that every historical questionnaire was complete or identical.

5. Interpretation and important limits

• Self-selection: people who choose to use the app may differ from other consumers. The sample does not support population prevalence estimates.
• Self-reporting: answers may be inaccurate, hypothetical or experimental. They are not verified against credit files or financial accounts.
• Repeated submissions: frequent use and caching can give some patterns more weight. Submission counts are not counts of unique people.
• Questionnaire context: unclear wording, unresolved options or changing questionnaires may affect who enters each eligible total.
• Geography: recorded country information has unverified provenance and default behavior. The most commonly recorded country codes were India and South Africa; codes are not verified places of residence. A default value is not assumed to mean the United States. No national prevalence claims are made.
• Estimated scores: app estimates are not official bureau scores. Answer-score relationships may be built into the scoring logic. The public findings here describe behaviors, not independent score effects.
• Association: combinations of answers are descriptive. No causal or independent-person significance claims are supported.

6. Privacy and what is shared

This public-facing study contains only the three reviewed aggregate findings and their methodological context. It does not provide raw questionnaires, individual answers, request identifiers, record-level profiles or downloadable source data.

This research pipeline processed records locally and did not send individual records to external AI services. The study describes aggregate research and does not offer a raw dataset for download.

7. Responsible use of the findings

Use the findings to discuss the behaviors reported in this app sample and to frame educational questions. Preserve the relevant count, eligible denominator, collection scope and self-selection caveat when quoting a result.

A suitable formulation is: “In KIQQI’s pooled app submissions with valid answers to both questions, 6,453 of 41,771 submissions (15.45%) reported always paying on time and utilization above 50%.” It would be inappropriate to replace “submissions” with “consumers” or claim that this is a national rate.