---
description: "FAQ on Dalton results and traffic — multi-armed bandit allocation, statistical confidence, and traffic requirements."
---

# Results & Traffic FAQ

Quick answers about traffic requirements, how the multi-armed bandit algorithm works, statistical significance, and reading results.

## How does the algorithm work?

Dalton uses multi-armed bandits. Traffic starts equal, then shifts toward winners automatically. Poor performers get less traffic quickly.

## How long until I see results?

Typically 2-4 weeks. Higher traffic = faster results. The dashboard shows time estimates.

## Why don't I see results yet?

Dalton needs data before it can report meaningful results:

- **~250 sessions** and **~100 conversion events** to start showing reliable data
- Until then, traffic splits equally and numbers will fluctuate

How fast you hit these thresholds depends on your traffic volume. Early numbers are noise—a variant might show +50% one day and -10% the next. Wait for significance indicators before drawing conclusions.

See [Understanding the Learning Phase](/optimization/results-dashboard#understanding-the-learning-phase) for details.

## How do I know when results are significant?

Look for 95%+ certainty in the dashboard. Dalton labels "High Performers" when variants hit significance.

## What's the baseline?

By default, Dalton keeps 20% of traffic on your original page to measure against. The remaining 80% goes to experiment variants. You can adjust this split in your settings.

## How much traffic do I need?

Minimum 5K monthly sessions per page. Below that, expect 6-8 weeks for results instead of 2-4.

## What if an experiment performs poorly?

The algorithm automatically reduces traffic to underperformers. You can also stop any experiment instantly—changes revert immediately.

## Can I test multiple languages/countries?

Yes, but each needs separate setup. Start with your highest-traffic market.

## Why don't my numbers match GA4 exactly?

Different attribution windows and sampling can cause small differences. The relative comparison (variant vs baseline) remains accurate.