Datum Nexus Book a consultation

Home/Services/Predictive modeling

Scores your team will actually trust.

A model that nobody understands does not get used, whatever its AUC. We build on features people can explain, validate against decisions rather than test sets, and monitor for drift once it is live.

Book a consultation

What this looks like

Four things we are usually asked for. Most engagements combine two or three of them.

Churn and retention scoring

Identify accounts at risk early enough for someone to do something about it, with the reasons attached to each score.

Risk and fraud models

Classification tuned to your actual cost of a false positive against a false negative, not to a generic accuracy metric.

Propensity and segmentation

Who is likely to convert, upgrade or respond, at the segment granularity your campaigns run on.

Monitoring and retraining

Drift detection, scheduled retraining and a rollback path when a new model underperforms the one it replaced.

How an engagement runs

The same four stages on every project, whether it runs six weeks or six months.

Discovery

A two-week sprint with your team to map the problem, audit the data, and work out whether this is worth building at all.

Scope and estimate

A written plan with what we will build, what we will not, the timeline, and what it costs. No deck, no retainer trap.

Build

Two-week iterations with a working demo at the end of each. You see progress in the product, not in a status report.

Handover

Documentation, a walkthrough, retraining and deploy scripts, and a support window while your team takes over.

What clients get out of it

Figures from real engagements. We will walk you through the full context on a call.

Explainable

Per-prediction feature attribution shipped with every score.

Cost-weighted

Thresholds set against your real cost of each error type.

Drift alerts

Monitoring on input distributions and output stability from day one.

Tools we work with

We pick from these based on what you already run. Nothing here is mandatory.

  • scikit-learn
  • XGBoost
  • LightGBM
  • SHAP
  • MLflow
  • Feature stores
  • A/B testing

Tell us what you are building.

Send a few lines about the project. You will get a reply from an engineer within two working days.