Data Science Proof of Concept

Testing the viability of ideas in a fast paced controlled environment.

Reaktor Data Science PoCs

We’ve been running Data Science and AI proof of concepts projects with msot of our partners in roder to help them accelerate their data transformation roadmaps.

There are lot of benefits for small isolates PoCs:
They are a good tool to quickly diagnose the quality of your data and understand in the short and long run your data capabilities and gaps

They help deliver good value in short period of times with very modest investments


  • They are not production ready
  • It’s experimental, therefore every project has a reasonable amount of uncertainty.
  • Fear not, we are here to support and guide you though the process


What are data science PoCs good for?

  • Validate the viability of an idea
  • Clear understanding of current data capabilities
  • Short turn around of insights


What’s the process for running Data Science PoCs?

The process is pretty standard and can be interrupted at each phase, where each of them give valueable insights as we move forward:

  • Business Case Calculation (Definition of KPIs and RoI)
  • Use case evaluation (Feasibility)
  • Data Assessment
  • Experiment Design
  • Machine Learning Model Design
  • Machine Learning Model Training
  • Experiment Execution


What is the main result?

Surprisingly, the most valuable deliverable of a Data Science PoC is not the model, but the insights on the data. More often than not, the optimal data points identified tend not to be available. This gives huge surgical insights on how to improve the data management across multiple areas.


How long it takes.

A PoC from start to finish might be done in 4 weeks, but there a few areas that can cause huge impacts in the timeline:

  • Data Preparation
    • Sometimes obtaining access to the data and piecing it together can easily become a project of its on. In that case it’s important to decide if the PoC should move forward to be stopped.
    • This can be easily identified in the beginning during the data assessment phase avoiding potential risks for the project.
  • Experiment Execution
    • Some experiments can be validated in a week, but others might require a lot more.
    • This is usually identified in the firs phase of the PoC when the use case is being put together


Common areas that we’ve been exploring

  • Bespoke recommendation engines
  • Churn prediction

Oops! We could not locate your form.