docs(TSP-1427): add eval metrics to analytics breakdown tables

[claude]

Summary

Updates enterprise/analytics.mdx to document the new evaluation metrics features added in PR #17207:

• Workforce Breakdown: Added Eval score, Alarm status, and Live evals columns; updated Actions & Credits description to note it includes live eval monitoring spend
• Agent Breakdown: Same three new columns plus the Actions & Credits live eval spend note
• Cost breakdown modal: New accordion documenting how to click Credits/Actions values to open the cost breakdown modal, which separates task execution spend from live eval monitoring spend (and is distinct from the eval run breakdown on individual eval results)
• Empty states: Documented 'Set up evals →' CTA and 'Insufficient data' states within the Eval score column descriptions
• Feature flag callout: Added <Info> callout to Workforce and Agent Breakdown sections noting that eval columns are gated behind account-level Evals access, using the same rollout language as the Evals page
• Cross-references: Linked to /build/agents/build-your-agent/evals#monitor when referencing Monitor dashboards and eval alarms

Linear issue: https://linear.app/relevance/issue/TSP-1427/

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
relevanceai	🟢 Ready	View Preview	Sep 1, 2026, 4:37 AM

💡 Tip: Enable Workflows to automatically generate PRs for you.

[linear]

TSP-1427

[github-actions]

🎯 Vibe check

Reviewed: 1 file (1 with issues, 0 clean)

Scores

	Dimension	Score	What's holding it back
🔴	Consistency	4/10	Pervasive heading/title case errors across all four heading levels and every Card, Accordion, and Step component. Multiple "agent"/"agents"/"workforce"/"workforces" lowercase misses for product terms. Bold labels inside callouts in three places.
🟡	Technical clarity	8/10	Content is accurate, internal anchor links verified (#monitor, #cost-and-billing, #monitoring-concurrency-usage all resolve). UI element names are specific and actionable throughout.
🟡	Non-technical clarity	7/10	Good plain-English explanations with concrete examples. The bold-label callout fragments (**Pro tip:**, **Filter and Focus**:) read as mini-headers trapped in the wrong container, which is mildly disorienting.
🟡	Structure	6/10	Two accordions ("Task and Action Trends", "Credits") use bullet lists inside them, violating the CLAUDE.md rule on accordion content. Bold labels inside callouts break the same rules. Page-level flow is otherwise logical.

✨ Overall vibe: The content here is genuinely useful — good use of tables for metric definitions, the cost breakdown modal explanation is clear, and the concurrency note correctly cross-references the plans page. The main problem is mechanical: nearly every title-like element is in title case instead of sentence case, and there are several callout/accordion component violations that need fixing before this is merge-ready.

🔧 Issues (29)

Heading sentence case — only proper nouns and product names get capitals:

• enterprise/analytics.mdx:19 — ## Getting Started → ## Getting started
• enterprise/analytics.mdx:37 — ## Key Metrics at a Glance → ## Key metrics at a glance
• enterprise/analytics.mdx:63 — ## Dashboard Sections → ## Dashboard sections
• enterprise/analytics.mdx:161 — ## Why Use Analytics? → ## Why use Analytics?

Step titles:

• enterprise/analytics.mdx:25 — <Step title="Set Your Filters"> → "Set your filters"
• enterprise/analytics.mdx:28 — <Step title="Explore Your Data"> → "Explore your data"

Card titles:

• enterprise/analytics.mdx:45 — <Card title="Tasks to Review"> → "Tasks to review"
• enterprise/analytics.mdx:48 — <Card title="Actions Used"> → "Actions used"
• enterprise/analytics.mdx:51 — <Card title="Credits Used"> → "Credits used"
• enterprise/analytics.mdx:164 — <Card title="Cost Control"> → "Cost control"
• enterprise/analytics.mdx:168 — <Card title="Performance Optimization"> → "Performance optimization"
• enterprise/analytics.mdx:172 — <Card title="Error Monitoring"> → "Error monitoring"
• enterprise/analytics.mdx:176 — <Card title="Capacity Planning"> → "Capacity planning"
• enterprise/analytics.mdx:180 — <Card title="ROI Visibility"> → "ROI visibility"

Accordion titles:

• enterprise/analytics.mdx:65 — <Accordion title="Task and Action Trends"> → "Task and action trends"
• enterprise/analytics.mdx:87 — <Accordion title="Workforce Breakdown"> → "Workforce breakdown"
• enterprise/analytics.mdx:108 — <Accordion title="Agent Breakdown"> → "Agent breakdown"
• enterprise/analytics.mdx:135 — <Accordion title="Action Breakdown"> → "Action breakdown"

Product term capitalization — Agent, Workforce are Relevance AI proper nouns:

• enterprise/analytics.mdx:9 — "how your AI workforce is performing" → "Workforce"
• enterprise/analytics.mdx:58 — "view individual agent performance" (inside the Note) → "Agent"
• enterprise/analytics.mdx:87 (accordion body) — "each of your AI workforces" → "Workforces"
• enterprise/analytics.mdx:113 — "individual agent performance" → "Agent"
• enterprise/analytics.mdx:153 — "not data specific to a single agent" → "Agent"
• enterprise/analytics.mdx:169 — "Sort agents by Credits/Task" → "Agents"
• enterprise/analytics.mdx:173 — "which agents or actions are failing" → "Agents"
• enterprise/analytics.mdx:185 — "agents with abnormally high" → "Agents"

Bold labels inside callouts — CLAUDE.md: callouts must be a single short paragraph with no bold labels:

• enterprise/analytics.mdx:33 — <Tip>**Pro tip:** — remove the bold label; the <Tip> component already signals the register. Rewrite as plain prose: "Scroll to the bottom of any breakdown table to see summary stats…"
• enterprise/analytics.mdx:58 — <Note>**Filter and Focus**: — remove the bold label; rewrite as a plain sentence.
• enterprise/analytics.mdx:89 — <Info>**Evals feature flag**: inside the Workforce Breakdown accordion (same pattern repeated at line 110 in the Agent Breakdown accordion) — remove bold labels; the <Info> container is sufficient framing.

🧩 Component suggestions (3)

• enterprise/analytics.mdx:65–73 — "Task and Action Trends" accordion uses **Quickly identify:** + bullet list + **How to use it:**. CLAUDE.md says accordion content should use flowing sentences, not bullet lists. Rewrite as prose: "Hover over any bar to see exact numbers for that day. The tallest bars show peak demand; unexpected spikes or drops may signal issues. Use the date range picker to zoom into specific weeks, and compare task volume against the Credits chart to understand cost patterns."

• enterprise/analytics.mdx:77–84 — "Credits" accordion has the same structure (bold subheads + bullets). Same fix: convert to flowing sentences that cover expensive days, cost anomalies, and how to compare with the Tasks chart.

• enterprise/analytics.mdx:41–53 — "Key Metrics at a Glance" uses a four-card <CardGroup> without href links. Per CLAUDE.md, non-navigable descriptors work better as a table or prose. These four metrics (Tasks, Tasks to Review, Actions Used, Credits Used) each have a consistent two-field structure (name + what it shows), making a two-column markdown table a natural fit and easier to scan than cards.

⚠️ Contradictions (0)

No contradictions found. The concurrency chart description at line 153 ("organization-wide concurrency data — not data specific to a single agent") matches the plans page wording exactly. The Evals rollout language in the Info banners (lines 89–90, 110–111) is consistent with the rollout note on evals.mdx:8. All internal anchors resolve.

🔋 Credit usage

Item	Count
Files reviewed	1
Context pages read	3
Total lines processed	~1,296

Files read: enterprise/analytics.mdx (187 lines), build/agents/build-your-agent/evals.mdx (447 lines), admin/subscriptions/plans.mdx (203 lines), enterprise/quick-start-guide.mdx (459 lines)