Results Dashboard

Track how your experiments are performing.

Two Main Views

Overview: See high-level performance across all experiments All Experiments: Dive into individual variant data

Overview Tab

Track your overall optimization performance:

• Live Page Combinations: How many variants are running
• Conversion Rate (30D): Your overall conversion rate (last 30 days)
• Conversion Lift vs Baseline (30D): How much better Dalton performs vs. your original pages
• Sessions Analysed: Total sessions tracked

What is "Baseline"?

Baseline is a holdout group (20% of traffic by default) that sees your original pages without any experiments. This lets you measure the overall impact of all Dalton optimizations combined.

The remaining 80% of traffic sees your experiments (the "Optimized" group).

AI Suggestions Banner: Dalton continuously generates new optimization ideas. Click "Review suggestions" to see proposed variants and new experiments.

Conversion Chart: See how Dalton's variants perform vs. baseline over time. Toggle between chart/data views and filter by time period.

All Experiments Tab

See detailed performance for each variant:

• COPY: The variant text being tested
• SPLIT: Traffic distribution bar (multi-armed bandit allocation)
• SESSIONS: Number of sessions for this variant
• CVR LIFT: Conversion improvement vs. baseline (green = better, red = worse)
• SIGNIF: Mini bell curve showing statistical significance
• ACTION: Pause button and current status

Understanding Status

Learning Phase

The algorithm is gathering data. All experiments currently stay in learning phase indefinitely (this will change soon).

Check statistical significance: Look at the mini bell curve in the SIGNIF column to see how confident you can be in the results.

Controls

• Toggle view: Switch between graph and table
• Show history: View past performance
• Pause: Stop the experiment temporarily

When to Pause

Trust the Algorithm

Pause experiments only when you need to make urgent page changes.

Multi-armed bandits automatically reduce traffic to underperforming variants, so there's no need to pause based on performance. Give the algorithm time to settle (2-4 weeks) before making decisions.