> For the complete documentation index, see [llms.txt](https://shoppad.gitbook.io/yedric/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://shoppad.gitbook.io/yedric/going-further/dashboard.md). # Dashboard The **Dashboard** is the first tab on every assistant and the screen you will spend the most time on. It answers two questions: how is this assistant performing, and what should I change next. When an assistant is new it shows a setup checklist instead, so you can track your way through the first-run steps. Once conversations start arriving, it switches to the reporting view described here.
Screenshot of the assistant Dashboard, focused on the headline metric cards above the conversation volume and sentiment panels.

The Dashboard is where performance and recommendations meet.

Pick a date range at the top. Every figure below, including the comparison arrows, is scoped to it. Each headline card shows its change against the previous period of the same length, so a range of 30 days compares against the 30 days before it. *** ## The headline numbers Five cards run across the top. **Conversations** How many conversations happened in the range. **Engaged %** The share of people who saw the widget and actually chatted. The line underneath tells you which denominator it used, either identified visitors or raw widget loads, because it prefers identified visitors when it has them. It reads as a dash until there is at least one widget impression, so a new install with no traffic shows nothing rather than zero. **Time saved** Total time your users spent in conversations with the assistant, expressed as hours and minutes. Read it as the support time the assistant absorbed rather than a precise saving. **Positive** The share of conversations the analyzer judged positive, calculated as everything left after negative and frustrated conversations are removed. Negative and Frustrated are deliberately not broken out as their own cards, because this single number already accounts for both. **Feature requests** How many conversations contained a feature request. This depends on **Detect feature requests** being on for the assistant. Turn that off in [Settings](/yedric/going-further/settings.md) and this reads zero. *** ## Additional metrics A quieter bar underneath carries three supporting figures. **Multi-session users** The share of identified visitors who came back for more than one session, with the underlying counts beside it. It needs identified visitors, so it stays a dash unless you are passing a username. See [Secure Mode](/yedric/developers/secure-mode.md). **Avg tokens** Average tokens per conversation, excluding simulation runs. **Avg time spent** Average conversation length. *** ## Recommended actions This is the part of the Dashboard that tells you what to do rather than what happened. It is one ranked list, highest severity first, mixing three kinds of recommendation. Each row carries its own type and impact badge, so you work down the list rather than across three separate cards. **Prompt improvement suggestions** Proposed edits to your assistant's instructions, drawn from patterns across recent conversations. Each one shows the pattern it observed, why the change would help, and the proposed prompt text. You can apply a suggestion straight to your Instructions tab without leaving the Dashboard. **Documentation gaps** Places where users got stuck because your knowledge base did not cover something. Each gap summarizes what was missing, lists the actual questions users asked, and proposes a section to write. **Opportunities to improve** Where the assistant could serve customers better, with a count of how often it came up. Every row links to sample conversations, so you can read the evidence before acting on it. Rows are tagged **high**, **medium**, or **low** impact. Once you have dealt with a row, mark it **Resolved** or **Dismissed**. That state persists, so a recommendation you have already handled does not reappear the next time the analysis runs. Each kind also has its own refresh control if you want to look for new findings immediately. *** ## Charts and lists Below the recommendations, the Dashboard breaks the same data down over time and by theme: * **Conversations by day** and **Conversations by time** show volume across the range and across the hours of the day. * **Actions taken** shows which of your assistant's actions actually ran. * **Top starting pages** ranks the pages on your site that conversations begin from, with a count and share for each. It counts only the page a visitor was on when they started, not every page they visited afterwards, and it treats one page with different query strings as one row, so it answers where the widget earns its place and where opening it by default might help. Only conversations that come through the website widget record a page. * Topic, feature request, and pain point lists surface what people are talking about, each one clickable through to the conversations behind it. The Dashboard reports on AI conversations only. If your organization has [live chat](/yedric/going-further/live-chat.md), its numbers live under **Live chat → Analytics**, which covers every assistant at once. *** ## Related * [Settings](/yedric/going-further/settings.md): the Conversation tagging controls that decide what the Positive and Feature requests figures are built from. * [Conversations](/yedric/going-further/conversations.md): the individual conversations behind every number here. * [Usage](/yedric/going-further/usage.md): token usage, rather than volume and sentiment.