Methodology

A repeatable pipeline from raw data to analytical output

Turbo Frog develops and uses its own software, analytical methodologies, algorithms, machine learning models and data-processing techniques.

  1. 01

    Intake & assessment

    We review sources, formats, volume and quality, then define the target output and the measures used to judge it.

  2. 02

    Cleaning & validation

    Records are validated, deduplicated and normalised. Anomalies are flagged and resolved against documented rules.

  3. 03

    Structuring

    Data is mapped into a consistent schema, with derived attributes added where the analysis requires them.

  4. 04

    Classification & modelling

    Proprietary taxonomies, statistical methods and machine learning models assign categories and produce scoring outputs.

  5. 05

    Analytical review

    Results are reviewed against expected distributions and sampled manually before anything is released.

  6. 06

    Delivery & iteration

    Processed datasets, classifications, scores and reports are delivered in the agreed format, with recurring runs where needed.

Principles

How we work

Reproducibility

Every transformation is defined in code and can be re-run on new data.

Transparency

Deliverables are accompanied by documentation of methods and limitations.

Proportionality

Model complexity is matched to the problem, never applied for its own sake.

Confidentiality

Client data is handled strictly within the scope of the engagement.