Predictive analytics that turns
data into what’s next.

Operational minimum: 5,000-10,000 historical rows spanning 12-18 months (for seasonality). Below that, the model won't generalize well. For churn/LTV: ideally 24+ months of data. For forecasting: minimum two complete seasonal cycles.

Default: XGBoost / LightGBM (95% of cases are tabular data, and they win there). Time series: Prophet + XGBoost ensemble. Recommendation: matrix factorization + deep learning (when traffic is high enough). Deep learning only when justified (NLP/CV/recommendation). We don't sell 'deep learning AI' when XGBoost solves the problem at 1/10th the cost.

REST API with <50ms latency (FastAPI + Redis cache). Output goes directly into your CRM (Salesforce/HubSpot via custom field), ESP (sequence trigger), ERP (forecast injection), and executive dashboard (Looker/Metabase/Power BI). Webhooks fire on critical changes (rising churn risk).

Monthly recalibration is included in the retainer: refit on the last 30 days of new data, A/B test against the current production model, and deploy only if the lift is significant. Continuous monitoring for data drift (Evidently AI) plus alerting when input distributions shift meaningfully.

Pilot build (one end-to-end model): 9,500-18,000 EUR one-time. Monthly operation (hosting + recalibration + monitoring): 1,200-3,500 EUR/month. Additional models after the pilot: 4,500-8,000 EUR each. For context: average ROI on predictive campaigns is 4x.

Yes. We only model on data you have legitimate consent for. PII is anonymized wherever possible (hashed email, generalized geo), and pseudonymized in training data. A DPO impact assessment is run when scoring has significant effects on individuals (automated decision-making, Art. 22).

No. Power BI Forecast / Quick ML are auto-ML features for exploratory work; they are good for uncovering insights but weak as a production system. We build models tuned to your domain, with automatic recalibration, data-drift monitoring, and output delivered directly into your business systems via API.