Relevance

Data is chosen and collected with consumer borrowers in mind. Data fields are limited to metrics that inform decision-making and risk-mitigation, as well as contact information. Most of the raw data must be analyzed by Thick Credit to drive insights, except for simple fields like phone number and email.

Collection      

All data is continually collected in the field by human agents. Most data comes directly from lender websites and representatives contacted via email or phone. In some instances, automated collection techniques are used after the process has been proven and vetted by a human agent.

Accuracy

Accuracy is reinforced by dual-authentication and continual updating of evolutionary figures, such as interest rates that may evolve with the prime rate.

Completeness

In some instances, lenders refuse to communicate data points or require an application. If a reasonable substitution through peer-comparison or nearest-neighbor techniques can be used, we use these or other best practice to complete the data. Otherwise, we remove the lender from any analysis where the empty data would corrupt results. The sample is large enough to remain representative despite the removal of a small number of records.

Recency

The data is updated continually to include new data and new offers entering the market.

Cleaning

Cleaning is not required for this dataset because the data is collected raw in the field and not repurposed from other sources.

Privacy

The dataset is hosted natively offline and on hard drives inaccessible via internet channels.

Validation

Outliers do not impact raw data, only analysis. We use a conservative framework that eliminates outliers with a material impact on analytical metrics helpful to consumers.

For example, not all offers impose annual fees, so an average of this metric across all lenders would misrepresent the figure due to a superficially high denominator. We exclude them to ensure consumer borrowers have the most representative figures available.