Analytics
Track delivery velocity, quality, build costs and cycle performance with per-project analytics dashboards and an AI analytics chat.
The analytics dashboard shows how your AI-powered development process is performing. Track trends across cycles, identify bottlenecks, measure quality outcomes and monitor AI costs.
When AI handles the implementation, lines of code and commit counts don't tell you much. Bynary's analytics focus on what matters: are features being accepted, how long do cycles take, and what is the defect rate?
- •Outcome-based metrics — Track acceptance rates, not just output volume
- •Cost visibility — See what AI operations cost per cycle and per project
- •Trend analysis — See how quality and cost change over successive cycles
The analytics dashboard with velocity charts, quality metrics, and cycle filtering.
Open Analytics
Filter by Cycle
Explore the Tabs
Ask the AI
Key Features
Overview
AI validated pass rate over the last 10 cycles, with fix rounds, human-caught defects and cost versus estimate.
Planning
Features per cycle, tasks per feature, conversation length and planning duration.
Development
Build duration, token usage, fix round frequency and task completion rate.
Quality
Acceptance, defect and missed rates, and issues per cycle.
Findings
Out-of-scope findings raised by the build agent, supervisor and reviewers, and how they were triaged.
Velocity
Features per cycle, cycle duration trends and delivery cadence.
Breakdown
Per-cycle detail at feature and task level.
Tags
Pass rate by tag, feature count, shipped count, escaped defects and total cost per tag.
Custom Reports
Charts you've saved from the AI analytics chat.
Bynary logs every AI call with its input and output token counts and a cost calculated from your model pricing. The costs page breaks spend down by team, project, cycle and pipeline stage. Super admins manage model pricing in Settings → Model Pricing.
The Overview shows the AI validated pass rate: the percentage of done whens the AI validator marked as passing during the build. It is paired with human-caught defects: issues a reviewer raised that the validator missed. A high pass rate with few defects suggests the AI's self-checks can be trusted. A high pass rate with many defects suggests the AI is over-confident and reviewers are catching real problems.
Tags group features across cycles, such as Auth, Billing or Onboarding. The Tags tab rolls metrics up by tag, so you can see which areas are consistently well built and which need extra fix rounds.
When do analytics become available?
Analytics fill in as cycles complete. You need at least one completed cycle for meaningful data, and trends become more useful after three or more.
How are costs calculated?
Input and output token counts are multiplied by the model-specific rates in Settings → Model Pricing.
Are analytics real-time?
Costs are logged as builds run. Dashboard charts reflect completed operations and update when you open the page.
We've thought carefully about these areas. Some are intentional design choices, others are on our roadmap.
The analytics dashboard is per project. The separate costs page covers spend across projects.
If you update model pricing rates, historical cost calculations are not adjusted retroactively.