Mind Analytics

Mind Analytics shows how people use your Minds: the tokens they spend, the messages, conversations and outputs they produce, and the answers that went wrong. Use it to see who uses which Minds and how much, and to review answers that need attention.

Open Mind Analytics

Open Observability > Mind Analytics.
At the top, pick Mind, User, Role or Insights.
Set Period, and optionally a model in Token usage.

Choose a mode, period and model

Setting	What it does	Default
Mode: Mind, User, Role, Insights	Sets whether the figures and the side list cover Minds, people, roles or the insight tabs.	Mind
Period	Sets the time window: All time, Last 24h, Last 7d, Last 30d, Last 90d or Custom.	Last 7d
Interval	Sets the chart buckets. The choices come from the platform and change with the Period; the control is dimmed while they load.	Set by the platform
Token usage	Shows the token figures for all models, or for one model you pick.	All models
Insight tab	Sets which tab shows in Insights mode.	Artifact errors

Token usage is hidden in Insights mode.
Picking a model changes the token figures only. The message, conversation, artifact, report and presentation counts stay the same.
The Token usage list offers the models that appear in the figures, most used first. A nameless model, shown as Unknown, is never offered as a choice.
Custom accepts a range of up to one year. A longer range shows Date range too large, with the message "Custom date range cannot exceed 1 year."
All time covers about the last 90 days, roughly the same window as Last 90d. It is not the full history of the platform.
Mind Analytics remembers your mode, period, model and insight tab in this browser. It does not save them to your account.

Read the figures at the top

The figures change with the mode.

Mode	Token figures	Other figures
Mind	Total tokens	Unique users, Messages, Conversations, Artifacts, Reports, Presentations
User	Mind tokens and Oda tokens	Unique minds, Messages, Conversations, Artifacts, Reports, Presentations
Role	Total tokens	Active users, Messages, Conversations, Artifacts, Reports, Presentations

Oda tokens are tokens used through DataMind Oda. In User mode they are shown apart from Mind tokens. Oda work is not tied to a Mind, so with a single Mind focused the figure is 0.
When you pick a model in Token usage, each token label starts with that model's name.
Mind Analytics counts tokens. It does not show a cost figure.

Note

Active users (Σ roles) appears when the total is added up from the roles rather than counted once per person. Role figures sum the members shown, so a person who holds several roles is counted once for each role, and role totals can exceed the platform total.

To inspect a single AI request, open AI, socket and PII logs. The Minds you see here are described in Minds.

Find the Minds, users and roles that use the most tokens

Choose Mind, User or Role.
Search the side list. The search box reads Search minds..., Search users... or Search roles....
Click an item to focus the figures and cards on it. Click All Minds, All Users or All Roles to return to the whole list.

Each list is ranked by token use, highest first, and each row shows its share of the list's tokens.
In Role mode, click a role to see its active members, ranked by token use, each with its share of the role's total. A role with no active members reads No active members in this period.
Switching mode resets the search box and clears the selection you made in the other mode. While you search, the All Minds, All Users or All Roles row is hidden.
Each item has a card that charts its activity over the period. The card's tooltip lists the token usage, Messages, Conversations, Artifacts, Reports and Presentations.

If nothing matches your search, the list shows No items found. If there is no activity in the period, the cards show No data for this period.

Review feedback, failed artifacts and slow answers

Pick Insights at the top.
In the side panel, pick a tab: Artifact errors, Heavy chats, Likes, Dislikes or Slow responses.
Read the chart and the list. Where a row links to its chat or artifact, click the link to open it.

Tab	What it lists
Artifact errors	Failed tool calls for a SQL query, a report or a presentation, with the error details.
Heavy chats	Conversations with high context usage and total tokens.
Likes	Messages where users left positive (like) feedback.
Dislikes	Messages where users left dislike feedback, with the reason they picked (Bug, Instructions, Inaccurate, Incomplete, Inappropriate, Outdated or Other).
Slow responses	Assistant messages with a large gap between their start and completion times.

Artifact errors groups the failures by SQL query, Report and Presentation.
Usage that no longer matches a Mind, such as usage left by a Mind that was deleted, is grouped under Unknown mind.

If a tab has no rows for the period, the panel shows No insights for this period, with a line that says what would appear there.

Who can open Mind Analytics

Mind Analytics needs the View Mind Overview permission. It is part of the Mind Overview Viewer preset, described as "View Mind usage overview", and of the Observability Viewer and Observability Administrator presets.

Without this permission the Mind Analytics tab is hidden, and Observability opens on the first tab you may see. Your administrator grants the permission through a role. See Roles and permissions and the Permission reference. For the other tabs, start at Observability.

Limits and edge cases

Each list loads at most 10,000 rows.
A card that cannot load reads Could not load data; change the period or refresh.
When something else goes wrong, see Observability troubleshooting.

Questions, answered

Which Mind uses the most tokens?

In Mind mode, the side list ranks the Minds by token use, highest first. Each row shows its share of the total.

How do I see one user's usage?

Pick User, search for the person in the side list and click their name. The figures then show that person's Mind tokens and Oda tokens separately.

Why does Active users (Σ roles) look too high?

The figure is added up across roles. A person who holds several roles is counted once for each of them, so the total can be higher than the number of people, and role totals can exceed the platform total.

Where do I see answers people disliked?

Pick Insights, then Dislikes in the side panel. The tab lists the messages where users left dislike feedback, each with the reason the person picked.

What are Artifact errors?

They are the failed tool calls a Mind made when it tried to produce a SQL query, a report or a presentation. Each entry shows the error details. Artifact errors is the default insight tab.

Can I see token use for one model only?

Yes. Pick the model in Token usage. The token figures then show that model only, and their labels start with the model's name. The counts, such as Messages and Conversations, do not change with the model. Mind Analytics shows tokens, not cost.

What does All time cover?

About the last 90 days, roughly the same window as Last 90d. It is not the full history of the platform; for a specific window, use Custom with a range of up to one year.

Is a chat with many messages counted once?

Yes. Conversations counts each conversation once, however many messages it holds. Unique users counts the distinct people who created conversations with the Mind.

What is Unknown mind?

Usage that no longer matches a Mind, such as usage left by a Mind that was deleted, is grouped under Unknown mind in the insight lists.

Which permission opens Mind Analytics?

View Mind Overview. It is included in the Mind Overview Viewer, Observability Viewer and Observability Administrator presets. Without it, the tab does not appear. If you hold it and still see no tab, your installation may have turned the tab off. See Observability.

Does Mind Analytics remember my choices?

Yes, in the browser you are using. It keeps your mode, period, model and insight tab there, so a different browser or computer starts from the defaults.