Skip to content
Beta Multi-model support (OpenAI, Gemini, and more) launching soon.
RBA RBA Skills

metrics-collector

Installable Ships in case-study-writer — or copy the raw skill below.
Open Claude.ai

Paste into Claude Code, Claude.ai, the desktop app, or the browser extension — ⌘/Ctrl+V.

ExperimentalNo security review

Metrics Collector

Structure raw project data from exports, reports, and team interviews into a clean, reusable metrics register. Designed to run before case-study-draft, award-submission-writer, or partner-evidence-package when the engagement team has not yet formalised project outcomes.

When this skill applies

  • Project has closed or is in hypercare but metrics are scattered across spreadsheets, retro notes, and team memories
  • A case study, award submission, or partner reference is needed and the metrics input is missing
  • An Azure DevOps, Jira, Lighthouse, or analytics export is available as a local file
  • The PM or delivery lead needs a structured metrics register for their own records

Skip when the project brief already contains a complete, quantified results section.

Inputs

Parameter Required Default Notes
project-name-or-folder Yes — Folder path or project name; skill will find relevant files
--source No manual Data source(s) for automated extraction
--baseline-date No — Month/year when baseline measurements were taken
--go-live-date No — Month/year of go-live; used to frame "days post-launch" metrics

Supported data sources

Source flag What to provide What is extracted
ado Azure DevOps work-item export (CSV/JSON) Sprint velocity, defect counts, cycle time, deployment frequency
jira Jira issue export (CSV) Sprint velocity, defect counts, resolution time
lighthouse Lighthouse JSON reports (before + after) LCP, CLS, INP, Lighthouse scores by category
ga4 GA4 export (CSV or JSON) Sessions, bounce rate, engagement rate, conversion events
adobe-analytics Adobe Analytics Workspace export (CSV) Page views, bounce, conversion, search clicks
manual No file — skill runs an interview Guided question sequence (see Step 3)

Workflow

Step 1 — File discovery

If a folder path is provided, search for relevant files:

find <folder> -name "*.csv" -o -name "*.json" -o -name "*.xlsx" -o -name "lighthouse*.json"

List discovered files and confirm with the user which to process. If no files are found, proceed to manual interview (Step 3).

Step 2 — Automated extraction (when source files are present)

Azure DevOps / Jira (sprint data)

Parse work-item export. Extract:

  • Sprint velocity — story points completed per sprint; average and trend
  • Defect escape rate — bugs raised in UAT and production as a % of total stories delivered
  • Cycle time — average days from "In Progress" to "Done" for stories
  • Deployment frequency — releases per sprint or per month (if pipeline data is present)
  • Planned vs. actual scope — story points planned at sprint 1 vs. delivered at go-live

Flag: "Story point inflation" patterns where estimates increase consistently — note this for the "what was hard" section of the win story, not for client-facing metrics.

Lighthouse (performance)

Parse before and after Lighthouse JSON reports. Extract per-page for homepage and 2–3 key pages:

Page Metric Before After Delta
Homepage LCP (ms)
Homepage CLS
Homepage INP (ms)
Homepage Lighthouse Performance
Homepage Lighthouse Accessibility

Note mobile vs. desktop and the Lighthouse version / run date.

Google Analytics 4

Parse GA4 export (7–14 days before go-live vs. equivalent period after). Extract:

Metric Baseline Post-launch Delta
Sessions
Bounce rate
Average engagement time
Key conversion event rate
Organic search sessions

Label any GA4 metrics with the event date range and note if the comparison windows include seasonal effects (e.g. December/January comparison is unreliable for most sectors).

Adobe Analytics

Parse Adobe Analytics Workspace export. Extract the same dimensions as GA4 above. Note: Adobe Analytics calculates "bounce" differently from GA4 — do not compare the two in the same table without a conversion note.

Step 3 — Manual interview (when no files are available)

Ask the following questions sequentially. Accept rough estimates — the goal is to establish a structured register with confidence levels, not perfect data.

Delivery metrics

  1. "What was the estimated duration at project kick-off, and what was the actual go-live date?"
  2. "How many story points or user stories were scoped at the start vs. delivered at go-live?"
  3. "How many bugs were raised during UAT? How many were P1 or P2?"
  4. "How many content items (pages, documents, media assets) were migrated? What was the target and what was the actual?"
  5. "Were any deadlines missed? What was the impact?"

Performance metrics 6. "Do you have Lighthouse scores or PageSpeed Insights data from before the project started?" 7. "Do you have a post-launch Lighthouse report? When was it run?" 8. "What were the Core Web Vitals results at launch?"

Business metrics (confirm these with the client before including in case studies) 9. "Has the client shared any data on conversion rates, leads, or revenue since go-live?" 10. "Has the client shared any data on content author satisfaction, publishing speed, or support tickets?" 11. "Is there a customer satisfaction score (CSAT or NPS) from the project?"

Editorial / operational metrics 12. "How many content authors were trained? Over what period?" 13. "How long does it take to publish a new page now vs. before the project?" 14. "Are there any managed services SLA metrics available (uptime, response times)?"

Relationship signals 15. "Has the client agreed to be a reference? Has anyone asked them yet?" 16. "Is there an ongoing retainer or expansion conversation?"

Step 4 — Build the metrics register

Produce a structured register as a Markdown table grouped by category. Apply confidence levels:

  • Confirmed — supported by a report, export, or written client communication
  • Team estimate — based on team recall; flag for verification before client-facing use
  • Client-provided — shared verbally or in email by the client; flag for written confirmation
## Metrics Register — [Project Name]

Generated: [date] | Baseline: [date] | Go-live: [date]
Sources: [list of data sources used]

### Delivery quality

| Metric | Value | Confidence | Source | Notes |
|---|---|---|---|---|
| Estimated duration | [n] months | Confirmed | SOW | |
| Actual duration | [n] months | Confirmed | PM log | |
| Variance | [+/- n days] | Confirmed | | |
| Defect escape rate (UAT) | [n] bugs / [n] stories | Team estimate | Retro | |
| Content items migrated | [n] / [n] target | Confirmed | Migration log | |

### Performance

| Metric | Baseline | Post-launch | Delta | Confidence | Source |
|---|---|---|---|---|---|
| LCP (mobile, homepage) | [n]ms | [n]ms | [%] | Confirmed | Lighthouse [date] |
| CLS (mobile, homepage) | [n] | [n] | [delta] | Confirmed | Lighthouse [date] |
| INP (mobile, homepage) | [n]ms | [n]ms | [%] | Confirmed | Lighthouse [date] |

### Business outcomes (client-provided — require written confirmation before publishing)

| Metric | Value | Timeframe | Confidence | Notes |
|---|---|---|---|---|
| Conversion rate delta | [%] | 90 days post-launch | Client-provided | Verbal; request written confirmation |

### Editorial / operational

| Metric | Value | Confidence | Source |
|---|---|---|---|
| Authors trained (week 1) | [n] | Confirmed | Training register |
| Time-to-publish (before) | [n] days | Team estimate | Retro |
| Time-to-publish (after) | [n] hours | Team estimate | Retro |

Step 5 — Gap report and next steps

After producing the register, output a gap report:

## Metrics Gap Report

### Missing metrics that would strengthen a case study
- [ ] Post-launch business conversion data — request from client [contact name]
- [ ] Before/after editorial timing — team estimate only; validate with content lead
- [ ] Infrastructure cost comparison — requires Azure cost analysis from cloud team

### Items requiring client confirmation before external use
- [ ] [Metric A] — currently "Client-provided, verbal"
- [ ] [Metric B] — client name attached; confirm approval to publish

### Recommended next steps
1. Share this register with [PM/delivery lead] for accuracy review
2. Request written confirmation from [client contact] for business metrics
3. Run case-study-draft once gaps are resolved
Technical details
Plugin
case-study-writer
Author
Dakota Smith
Skill name
metrics-collector
Maturity
experimental
Security review
No security review
Owners
  • business-development
  • strategy
Argument hint
<project-name-or-folder> [--source ado|jira|lighthouse|ga4|adobe-analytics|manual] [--baseline-date <YYYY-MM>] [--go-live-date <YYYY-MM>]
Allowed tools
  • Read
  • Write
  • Edit
  • Bash(find *)
  • Bash(ls *)
  • Bash(grep *)
  • Bash(cat *)
Source
plugins/case-study-writer/skills/metrics-collector/SKILL.md

Edit this skill on GitHub — opens GitHub's web editor. No git knowledge needed.