Analytics Insights
Analytics Insights analyzes your exported project's performance data and tells you what's working, what isn't, and what to do next. It interprets engagement metrics, identifies drop-off points, compares variants, segments results by platform or device, and surfaces patterns across your project portfolio.
This specialty focuses on analysis and interpretation. It reads your real performance data and translates it into specific, actionable recommendations for your next iteration.
Before diving in, it helps to understand the core metrics Analytics Insights works, read the Analytics Documentation here.
What It Covers
Analytics Insights addresses four main areas of performance analysis.
Funnel Diagnostics and Friction Points
Find out exactly where users are dropping off or getting stuck. Analytics Insights can identify which scenes underperform, which layers get the most or least interaction, and why performance degrades over time.
"Where is the biggest drop-off occurring?"
"What are my most and least interacted-with layers?"
"Why does performance decrease from 1-week to 1-month?"
Variant Comparison and A/B Analysis
Compare how different variants perform against each other. Analytics Insights can explain why one variant outperformed another and suggest how to improve capture rates on specific elements like the end card.
"Why did Variant A have better performance than Variant B?"
"How can I increase captures on the end card?"
Platform and Device Segmentation
Break down performance by partner, platform, operating system, or device type. Analytics Insights can tell you which partners are driving the most sessions, engagements, or captures — and whether your playable performs differently on iOS vs. Android or on newer vs. older devices.
"Which partner is driving the majority of captures?"
"Does the playable perform differently on iOS vs. Android?"
"Is there a specific dynamic where my performance is better?"
Portfolio Trends and Benchmarking
Zoom out and look at patterns across your full project history. Analytics Insights can correlate metrics, identify what your top performers have in common, benchmark against your vertical, and rank projects by opportunity.
"How do my results compare to others in my vertical?"
"What common characteristics do my top performing projects share?"
"Based on this month's performance data, what have we learned and what should I do next?"
How to Use It
Open CRAFTSMAN INTELLIGENCE and ask a question about your project's performance. Analytics Insights works with your exported project data, so make sure the project has been exported and has accumulated performance data before asking.
For a quick diagnosis, ask a broad question and let the assistant find the biggest issues.
"Where is the biggest drop-off occurring?"
For targeted analysis, specify the metric, variant, partner, or time range you're interested in.
"Which partner is driving the most sessions this month?"
"Why did Variant A outperform Variant B on capture rate?"
Template Prompts
Click the Analytics Insights card in the chat panel to see template prompts you can send with a single click:
What are the top reasons for mid-playable drop-off?
How does performance differ between iOS and Android?
Which partner drives the highest Capture Rate?
What is the correlation between Duration and Captures?
How can I improve Engagement Rate?
Tips
Start with the funnel. If you're not sure where to begin, ask "Where is the biggest drop-off occurring?" — this gives you the most impactful insight first.
Be specific about metrics. Analytics Insights understands sessions, engagements, and captures. Using these terms in your prompts helps you get precise answers.
Compare variants early. If you're running A/B tests, ask for a variant comparison as soon as you have enough data. Don't wait until the campaign ends.
Use this to close the loop. After building with the Builder Agent, export and gather data, then come back to Analytics Insights. Feed what you learn into Creative Strategy to plan your next build.
Ask for a monthly summary. At the end of each month, try "Based on this month's performance data, what have we learned and what should I do next?" — this gives you a clear starting point for the next cycle.
Last updated