The problem

A foodtech company's R&D team evaluates certain characteristics of each new formulation from its input data. The evaluations had to get faster without losing the ability to check how each result was reached.

Before and after

Starting point

Each evaluation was worked through by hand in spreadsheets and took about a day.

What I did

  • Built a Python tool that reads the formulation's input data from Excel or CSV files.
  • Fitted a statistical regression model that predicts the characteristics, each with a confidence value.
  • Made every prediction explain how it was reached, so the R&D team can check the reasoning before acting on it.
  • Kept it purely mathematical, with no language model, so the same inputs always give the same result.

Outcome

  • An evaluation that took about a day of spreadsheet work now takes only as long as preparing and uploading the input data.
  • The R&D team sees a confidence value and the reasoning behind every prediction, so uncertain results stand out.

My role

  • Sole developer, from the input format to the explained output
  • Statistical modelling with confidence values and explanations
  • In use by the customer's R&D team

Technology

  • Python
  • Statistics
  • Regression modelling
  • Excel
  • CSV

Related service

Reporting & Data

Have a process or software problem worth checking?

A short conversation is enough to decide whether an audit, a small build, or no custom software is the right next step.