Novus AI Stats
Read Novus AI Stats metric quality labels
Learn how Novus AI Stats marks each metric exact, source reported, derived, estimated, or unavailable so estimates never masquerade as facts on the dashboard.
Each metric in Novus AI Stats is marked exact, source reported, derived, estimated, or unavailable. That labeling is the honesty contract: a derived session count and an exact token total from the source should never look the same on the chart.
This tutorial teaches you to read those labels before you treat streaks, time spent, or project totals as ground truth.
Contents
Two ways to finish
Know the five labels
Exact through unavailable — what each claim means.
Read without over-claiming
Use labels when comparing providers or projects.
- 1
1. Treat the label as part of the number
A big token total with an estimated label is a different claim than the same digit marked exact. Train your eye to read the quality tag in the same glance as the value.
When a metric is unavailable, that absence is information — the source did not provide enough signal, and AI Stats refuses to invent a confident substitute.
- 2
2. Map the five quality states
Exact means the value comes directly from a countable field. Source reported means the provider stated it. Derived means AI Stats computed it from other local fields. Estimated means a model or heuristic filled a gap. Unavailable means no honest value can be shown yet.
Estimates are allowed; pretending they are exact is not. The UI exists so those two never share the same visual weight by accident.
Estimates may appear — they must never masquerade as facts. - 3
3. Compare providers with labels in view
When ChatGPT chats sit beside Codex tool runs, volumes are not automatically comparable. One stream may give exact session counts while another only supports derived time spent.
Prefer questions the labels can answer: “which provider has more exact sessions?” beats “which AI did I use more?” when time is only estimated on one side.
Label before narrative
Write the quality label into any note or screenshot caption. Future-you will thank present-you when an estimated streak no longer looks like a certified record.
Frequently asked questions
Quick answers to common questions about this topic.
What are the metric quality labels?
Exact, source reported, derived, estimated, and unavailable — so estimates never masquerade as facts.
Why is a metric unavailable?
The import did not provide enough signal for an honest value. AI Stats leaves the gap visible instead of inventing confidence.