Funnel Reporter

Funnel Reporter

Purpose

Pull live analytics data from the Humblytics API to generate comprehensive funnel reports. Covers the full journey from traffic acquisition through conversion and retention. Designed for SaaS businesses that need regular reporting on marketing and product metrics.

When to Use

• Generating weekly or monthly marketing performance reports
• Analyzing traffic trends and source attribution
• Reporting on signup and trial-to-paid conversion rates
• Identifying changes in funnel performance over time
• Preparing board or stakeholder updates with marketing metrics
• Comparing period-over-period performance

Connection

Live data comes from the Humblytics MCP (server humblytics). The skill calls mcp__humblytics__* tools — the MCP handles auth, base URL, and property resolution. Connect it once per the repo README; never paste API keys into chat. The API key lives in the MCP connection headers, set at connect time.

Property: for a single-property key the MCP auto-resolves the property. For a multi-property key, call list_properties and pass propertyId to each tool.

Before You Start

1. Confirm the property — Which Humblytics property to report on (call list_properties if the key covers more than one)
2. Define the time period — This week, last 30 days, month-over-month, quarter, custom range
3. Identify the audience — Is this for the team, leadership, investors? This shapes detail level and framing.
4. Check for comparison period — Most useful reports compare current vs previous period
5. Understand the funnel steps — Confirm the key conversion events tracked in Humblytics
6. Look for context — Check project docs for business model, pricing, and target metrics

Core Workflow

Step 1: Pull Traffic Data

Retrieve top-of-funnel metrics via the Humblytics MCP traffic tools. They take start, end (ISO 8601 datetimes) and timezone (IANA). Optional: granularity (hour/day/month) for time-series. NOTE: granularity=week is currently bugged on get_traffic_trends and returns all-zero buckets — use day and aggregate to weeks client-side instead.

MCP tools:

• get_traffic_summary — Aggregate metrics (pageviews, sessions, bounce rate, avg session duration)
• get_traffic_trends — Timeseries pageviews & unique visitors (use granularity=day; avoid week — it returns all-zero buckets)
• get_pages_breakdown — Page-level performance: views, visitors, scroll depth, bounce rate
• get_traffic_breakdown — UTM source/medium/campaign + device + location dimensions, all from the same tool
• get_entry_exit_pages — Top entry and exit pages

Key traffic metrics to pull:

• Total sessions and unique visitors
• Page views and pages per session
• Average session duration
• Bounce rate
• New vs returning visitors ratio

Step 2: Analyze Traffic Sources

Break down where traffic is coming from:

Flag any source with:

• Significant volume change (>20% vs previous period)
• Unusually high or low conversion rate
• High bounce rate (>70%) suggesting poor targeting or landing page mismatch

Step 3: Page Performance

Identify top-performing and underperforming pages:

Top pages by traffic — Which pages attract the most visitors? Top pages by conversion — Which pages drive the most signups/purchases? Highest bounce rate pages — Where are people leaving immediately? Lowest engagement pages — Short time-on-page, low scroll depth

For key landing pages, report:

• Sessions, bounce rate, avg time on page
• Conversion rate and total conversions
• Period-over-period change

Step 4: Funnel Step Analysis

Map the full conversion funnel with data:

Copy code
Visitors → Signups → Activated → Trial → Paid → Retained

  10,000 → 500 → 300 → 200 → 80 → 65
         5.0%  60.0%  66.7%  40.0%  81.3%

For each transition, report:

• Volume: How many users at each step
• Conversion rate: Percentage moving to next step
• Period comparison: How this compares to the previous period
• Trend: Improving, declining, or stable

Step 5: Conversion Events

There's no generic events tool — pull conversion data from the dedicated form and click tools instead:

• get_forms_breakdown — All form submissions across pages
• get_forms_details (pass page: "/path") — Conversion rates for a specific form/page
• get_clicks_breakdown — Click data with top targets across all pages
• get_clicks_details (pass page: "/path") — Clicks on a specific page with UTM attribution

Report on:

• Signup completions (get_forms_breakdown filtered to signup pages)
• CTA clicks (get_clicks_details by page and CTA target)
• Form submissions (get_forms_breakdown)
• Pricing page views (get_page_details with page: "/pricing")
• Trial starts (get_forms_breakdown filtered to the trial-start form)
• Upgrade/purchase events (get_forms_breakdown filtered to checkout/purchase)

Step 5.5: Attach Paid Attribution (when reporting paid channels, plus revenue/trial activations only when a revenue connector is attached — else 0)

If the report needs to surface paid-channel performance, ROAS, or revenue-by-campaign, enrich the funnel with the get_ads_attribution tool (pass startDate and endDate as YYYY-MM-DD):

• get_ads_attribution — per-campaign attribution rows

Returns per-campaign rows with spend, impressions, clicks, sessions (always populated) plus revenue, revenue_conversions, roas, trial_count (these are 0 unless a revenue connector such as Stripe/ChartMogul is linked — do NOT report ROAS or revenue from this tool alone), and unmatched_ad_campaigns (UTM hygiene gaps) / unmatched_utm_campaigns (organic/email traffic). Pair with get_traffic_breakdown source data to build a paid-vs-organic split (spend/sessions only).

For raw connector metadata (ad accounts, daily insights, ad creative), use the meta/google MCP tools — list_meta_connections, get_meta_daily_insights, list_google_ads_connections, and related — see revenue-attributor for the full workflow.

Step 6: Period-over-Period Comparison

Always compare against the previous period (week-over-week or month-over-month):

Highlight:

• Metrics that improved significantly (celebrate wins)
• Metrics that declined (flag for investigation)
• Metrics that are flat but should be growing (stagnation risk)

Step 7: Insights and Recommendations

Do not just present data — interpret it:

The Report Summary Structure:

1. Executive Summary (3 sentences max)

   • Overall funnel health: healthy / needs attention / critical
   • Biggest win this period
   • Biggest concern this period

2. Key Metrics Table — The 5-8 most important numbers

3. Traffic Analysis — Sources, trends, notable changes

4. Funnel Performance — Step-by-step with conversion rates

5. Page Performance — Top and bottom performers

6. Insights (3-5 bullets)

   • What changed and why
   • What is working well
   • What needs attention

7. Recommended Actions (3 bullets)

   • One quick win
   • One strategic initiative
   • One thing to investigate further

Report Templates

Weekly Report (concise)

• Executive summary (3 lines)
• Key metrics table (5 numbers)
• Top 3 insights
• Top 3 actions

Monthly Report (comprehensive)

• Executive summary
• Full traffic analysis with source breakdown
• Complete funnel with step-by-step conversion rates
• Page-level performance (top 10 pages)
• Channel-by-channel breakdown
• Month-over-month trends
• Insights and strategic recommendations

Board/Stakeholder Report (high-level)

• 3 headline metrics (traffic, signups, revenue)
• Trend arrows and period comparison
• One paragraph narrative
• Strategic outlook

Metric Benchmarks (SaaS)

Use these as reference points when analyzing data:

Related Skills

• cro-optimizer — Take report findings and turn them into optimization actions
• ab-test-generator — Create tests based on underperforming pages or steps
• marketing-strategist — Use report data to inform strategic planning
• page-cro — Deep-dive into specific underperforming pages

Shared Frameworks (REQUIRED reading)

Two primitives are load-bearing for this skill. Without them, funnel reports fall back to flagging the highest-percentage drop and quoting raw decimals — both of which mislead.

• _shared/frameworks/largest-leak-first.md — rank funnel steps by absolute people lost, not percentage drop. A 10% drop on 10,000 visitors (1,000 lost) outranks a 50% drop on 100 visitors (50 lost) by 20×. Compute absolute_loss_per_step = step_in - step_out first; only sort by percentage as a secondary view. This often inverts the priority order.

• _shared/frameworks/percentile-framing.md — replace "your trial-to-paid CVR is low" with "your trial-to-paid CVR is 11%, at p25 in the OpenView B2B SaaS distribution (p50=14%, p75=22%) — meaningful headroom to median." Use the bands in _shared/benchmarks/baselines.json:

  • All sites desktop: p50 = 3.82% CVR
  • All sites mobile: p50 = 1.32% CVR (~3× desktop gap)
  • B2B SaaS pricing page: p50 = 3.8% CVR (well-optimized: p75 = 8–12%)
  • Ecommerce checkout fields: p50 = 11.3 fields (Baymard optimum = 8)

• _shared/frameworks/preflight-checklist.md — confirm time range and traffic volume before the report. Funnel diagnostics on <500 sessions are noise, not signal — flag rather than pretend the percentages are meaningful.

• _shared/benchmarks/baselines.json — full vertical baselines library. Cite the source on every comparison ("Aggregated Statsig/Eppo/FirstMark 2023-2024 reports").

The existing benchmark table in this SKILL.md (Poor/OK/Good/Excellent) is a coarse heuristic. Prefer percentile-band framing from the baselines library when the vertical-stage-metric combination is present there.