This page explains the words you'll see on the AI Dashboard, one at a time, with a simple everyday example for each. For the full technical breakdown (formulas, calculations), see the detailed Metrics Guide.
Jump to a term
What is a Session?
A Session is one complete conversation with the AI assistant — from the moment it starts to the moment it ends.
What starts and ends a Session?
| Action | Effect |
|---|---|
| Developer opens Claude Code and types the first message | Session starts |
| Developer keeps asking follow-up questions | Same Session — each exchange is a Turn |
| Developer closes the window or starts a new project | Session ends |
| Developer opens Claude Code again later | A brand-new Session begins |
Full Example
A developer opens Claude Code at 9 am to fix a login bug — that's Session 1. They ask several follow-up questions, then close it at 10 am.After lunch they open it again to add a new feature — that's Session 2.
By end of day: 2 Sessions, even though they used the tool all day.
What are Tokens?
Tokens are just a way of measuring how much text was exchanged with the AI — similar to counting words. It includes both what the developer typed and what the AI wrote back.
How is a Token counted?
A token is roughly 4 characters or about ¾ of a word. Short common words like the, is, fix each count as 1 token. Longer or unusual words may count as 2 or more tokens.
| Text | Characters | ≈ Tokens |
|---|---|---|
the | 3 | 1 token |
bug | 3 | 1 token |
login | 5 | 1 token |
Fix the bug on the login page | 30 | ≈ 8 tokens |
| A typical AI code reply (50 lines) | ~1,500 | ≈ 375 tokens |
Full Example
You type: "Fix the bug on the login page" → about 8 tokens (Input).The AI replies with an explanation and 50 lines of code → about 375 tokens (Output).
Total for this one exchange: 8 + 375 = 383 tokens. A busy day with many exchanges easily adds up to tens of thousands of tokens.
What is LOC Gen (Lines of Code Generated)?
This is simply how many lines of code the AI actually wrote or changed in the project.
What counts as a line?
Every line the AI adds or edits using its Write or Edit tools is counted — whether it is brand-new code, a bug fix, or a small tweak to an existing line.
| What the AI did | Lines counted |
|---|---|
| Added a new 20-line function | 20 |
| Fixed a bug — changed 5 lines in another file | 5 |
| Added a 50-line HTML page | 50 |
| Deleted a block (deletions are not counted) | 0 |
Full Example
In one session the AI: added a 20-line feature, fixed 5 lines in another file, and created a new 50-line page.LOC Gen = 20 + 5 + 50 = 75 lines generated for that session.
What is a Turn?
A Turn is one back-and-forth — the developer asks something, the AI replies. A single Session can contain many Turns.
How Turns build up in a Session
| Who speaks | What they say | Turn count |
|---|---|---|
| Developer | "Fix the bug on the login page" | Turn 1 |
| AI | Finds the bug, fixes the code, explains the change | |
| Developer | "Now also add a loading spinner" | Turn 2 |
| AI | Adds the spinner and updates the CSS | |
| Developer | "Looks good, done" | Session ends |
Example
Developer asks a question, AI answers, developer asks a follow-up, AI answers again — that's 2 Turns inside one Session.
Avg Turns (Average Turns per Session)
This tells you, on average, how many back-and-forth messages it took to finish each Session in a given week. A lower number is better — it means requests were clear and the AI got things right quickly.
Avg Turns = Total Turns in the week ÷ Total Sessions in the week
Example — Week 2026-W23
This week had 5 Sessions, and across those 5 Sessions there were 24 Turns in total.Avg Turns = 24 ÷ 5 = 4.8
This lands in the "Efficient" range — most sessions that week were wrapped up in just a handful of messages.
What is a Tool Call?
Every small action the AI takes behind the scenes to get the job done — like opening a file, searching for something, or running a command.
Common Tool Calls and what they do
| Tool | What it does |
|---|---|
Read | Opens and reads a file to understand the current code |
Edit | Makes a targeted change to an existing file |
Write | Creates a brand-new file |
Grep | Searches across files for a word, function name, or pattern |
Bash | Runs a terminal command (e.g. install a package, run tests) |
Glob | Lists files matching a pattern (e.g. all .html files) |
Full Example — "Fix the bug on the login page"
Turn 1 — developer sends the request. The AI then:1.
Grep — searches for "login" across all files → Tool Call 12.
Read — opens login.html to read the current code → Tool Call 23.
Read — opens auth.js to check the related logic → Tool Call 34.
Edit — fixes the bug in auth.js → Tool Call 4Total: 4 Tool Calls to complete one Turn. A complex task across many files can easily reach 20–30 Tool Calls.
What are Iterations (Avg Iterations)?
The average number of Turns it takes to finish a Session. A lower number means the instructions were clear and the AI got it right quickly. This is the same idea as Avg Turns — just the name used on the Weekly Summary table instead of the Weekly Drill-Down.
Example
One task was finished in 3 Turns, another took 9 Turns — the average across both is 6 Iterations.
How Avg Iterations is calculated
Avg Iterations = Total Turns in the week ÷ Total Sessions in the week
Example — Week 2026-W24
This week had 7 Sessions, and across those 7 Sessions there were 28 Turns in total.Avg Iterations = 28 ÷ 7 = 4.0
The number is then shown as an easy-to-read tag:
| Avg Iterations | Tag | Meaning | Tip |
|---|---|---|---|
| Below 5 | 🟢 Efficient | Clear, precise prompts — right first time | Keep it up — your prompts are working well |
| 5 – 9 | 🟠 Moderate | Some back-and-forth — room to improve | Try giving more context and examples upfront |
| 10 or more | 🔵 Complex | Hard task or prompt needed several tries | Break large tasks into smaller, focused requests |
Applying it to Week 2026-W24
4.0 is below 5, so this week gets the 🟢 Efficient tag.
What is the Score?
A single number out of 100 showing how effectively someone is using the AI tool overall — how often they use it, how much it produces, how clear their requests are, and how much it explores the project.
How the Score is calculated — 4 parts added together
The Score is made of 4 smaller scores, each rewarding a different good habit:
| Part | Max points | Rewards | How to earn full points |
|---|---|---|---|
| Usage | 30 | Using the AI regularly | Open Claude Code every working day — even a quick task counts |
| Output | 25 | How much code the AI produces | Let the AI write and edit files, not just answer questions |
| Prompt Quality | 25 | Getting answers in few Turns | Give clear, specific instructions with enough context upfront |
| Tool Diversity | 20 | AI using many different tools | Let the AI explore freely — searching, reading, running commands all count |
Total Score = Usage + Output + Prompt Quality + Tool Diversity (out of 100)
Full Worked Example
A developer's month: Usage = 18, Output = 20, Prompt Quality = 17, Tool Diversity = 15.Total Score = 18 + 20 + 17 + 15 = 70 out of 100
Their weakest area is Usage (18/30) — opening Claude Code on a few more days each week would push their score above 75 and earn them the 🏆 AI Champion badge.
What is a Badge?
A simple label based on someone's Score, so performance is easy to see at a glance without reading numbers.
How the Badge is decided — by Score range
| Score range | Badge | What it means |
|---|---|---|
| 75 – 100 | 🏆 AI Champion | Uses AI every day, produces lots of code, clear and precise prompts |
| 55 – 74 | ⭐ Power User | Strong regular usage with good output — nearly there |
| 35 – 54 | 🌱 Active Learner | Growing habit — using the tool but still finding their stride |
| 15 – 34 | 🚶 Getting Started | Early days — occasional use, light output |
| 0 – 14 | 🧭 Needs Coaching | Rarely used this month — a conversation may help |
How to move up a Badge
| To improve… | Do this |
|---|---|
| Usage (max 30 pts) | Open Claude Code every working day — even a short task counts as a Session |
| Output (max 25 pts) | Let the AI write and edit code rather than just answering questions |
| Prompt Quality (max 25 pts) | Give clear, specific instructions — fewer follow-up messages = higher score |
| Tool Diversity (max 20 pts) | Let the AI explore freely — searching, reading files, running commands all count |
Full Example
A developer's month: Usage = 18, Output = 20, Prompt Quality = 17, Tool Diversity = 15.Total Score = 18 + 20 + 17 + 15 = 70 → falls in the 55–74 range → Badge: ⭐ Power User.
To reach 🏆 AI Champion (75+), they need just 5 more points — one extra session per day for a week would do it.