What-if simulation

What-if simulation answers a single question: if I make this change, how much would my visibility move? It runs the change through the calibrated predictive model instead of the live engines — so you get a predicted score in seconds, at no credit cost, before you build anything.

How simulation works

Lokrix takes your property’s current feature vector, applies the proposed change to the relevant features, and re-runs the predictive scoring model. Because the model is calibrated (Platt/Isotonic) and validated with Brier score, ECE and AUC, the predicted probability is meaningful — not a guess dressed up as a number. Every simulated result carries a 95% confidence interval, the same statistical honesty Lokrix applies to live measurement.

Why it costs no credits

Live engine runs are the dominant cost, metered against your credit balance (1 credit ≈ 1 engine response). Simulation never queries an engine — it only evaluates the model — so you can explore dozens of hypotheticals freely and reserve credits for verifying the changes you actually decide to ship.

Predicted vs. verified

A simulation is a prediction, and Lokrix labels it as such — provenance is always clear, just as data is labelled live, cached or mock across the product. To confirm a change landed in the real engines, run a fresh live scan afterwards; the observed result then re-anchors the model for the next simulation.

• Simulate to choose between options and set expectations — free, instant, model-based.
• Verify with a live scan once the change is live — real engine responses, metered in credits.

Simulation is the engine behind both recommendation uplifts and the path to #1.