The customer-specific model layer, and why Starter unlocks it

Free uses the global model — trained across all Veritus traffic. Starter adds a customer-specific model layer on top.

As you confirm and overturn verdicts in the dashboard, those decisions become training labels. Each night, a model layer specific to your traffic is retrained and weighted alongside the global model (customer models trained on 30+ days of your data carry meaningful weight in the final probability). The result is scoring tuned to your audience — if VoIP numbers or a particular country are normal for your customers, the model learns that instead of penalising them.

This is one of the main reasons growing SaaS and e-commerce sites move from Free to Starter: the engine starts adapting specifically to their signups.