Audit my analytics setup for [type of business]. Platform: [GA4/Mixpanel/etc.]. Currently tracking: [what's configured]. Tell me what's missing, what's misconfigured, and what to prioritize fixing.

---
name: analytics-setup
description: Configure analytics foundations. Account structure, goals, data quality, verification. Use when setting up new analytics, auditing existing setup, or migrating platforms.
---

# Analytics Setup

Configure analytics foundations using platform-agnostic principles. This skill covers account structure, goal and conversion setup, data quality controls, and verification processes that apply whether you use GA4, Mixpanel, Amplitude, Plausible, or any other analytics platform.

## When to Use This Skill

- Setting up analytics for a new site or product
- Auditing an existing analytics configuration
- Migrating between analytics platforms
- Verifying that tracking matches reality
- Troubleshooting data quality issues
- Preparing analytics foundation before adding event tracking

## Account Structure Principles

### Environment Separation

Never mix data from different environments:

| Environment | Purpose | Data Handling |
|-------------|---------|---------------|
| Production | Real user data | Primary analytics property |
| Staging | Pre-release testing | Separate property, exclude from reports |
| Development | Local development | Separate property or filtered out |

**Why it matters:** Test traffic pollutes real data. A developer refreshing a page 500 times during testing looks like a traffic spike. Internal testing of checkout flows creates fake conversions.

### Property/Account Naming

Use consistent, descriptive naming:

```
[Company] - [Product/Site] - [Environment]
```

**Examples:**
- Acme Corp - Main Site - Production
- Acme Corp - Main Site - Staging
- Acme Corp - Mobile App - Production

### Access Control

| Role | Permissions | Who |
|------|-------------|-----|
| Admin | Full configuration access | Analytics owner, engineering lead |
| Editor | Can modify goals, filters | Marketing leads |
| Analyst | Full read access | Team members who analyze |
| Viewer | Read access to reports | Stakeholders, executives |

**Principle:** Minimum necessary access. Most people need to read data, not modify configuration.

### Data Retention

| Consideration | Recommendation |
|---------------|----------------|
| Minimum retention | At least 14 months (year-over-year comparison) |
| Ideal retention | 26 months (2+ years of trends) |
| Legal constraints | GDPR and privacy laws may limit retention |
| Storage costs | Some platforms charge for longer retention |

Set retention before collecting data. Extending later doesn't recover deleted data.

## Goal and Conversion Setup

### Conversion Types

| Type | Definition | Examples |
|------|------------|----------|
| Macro conversions | Primary business outcomes | Purchase, paid signup, qualified lead |
| Micro conversions | Steps toward macro conversions | Add to cart, form start, demo request |

Track both. Macro conversions tell you the outcome. Micro conversions help diagnose where funnels break.

### Goals by Business Type

| Business | Macro Conversions | Micro Conversions |
|----------|-------------------|-------------------|
| SaaS | Paid subscription, trial start | Signup, activation, feature adoption |
| Ecommerce | Purchase completed | Add to cart, checkout start, account creation |
| Lead generation | Qualified lead submitted | Content download, form start, demo request |
| Content/Media | Subscription, membership | Newsletter signup, article completion, share |
| Marketplace | Transaction completed | Listing view, contact seller, account creation |

### Goal Value Assignment

Assign monetary values to conversions when possible:

| Conversion | Value Approach |
|------------|----------------|
| Purchase | Actual transaction value |
| Subscription | Lifetime value or first-year value |
| Lead | Average lead value (leads × close rate × deal size) |
| Trial | Trial × conversion rate × subscription value |
| Content action | Estimated value or indexed value |

Even rough estimates help prioritize. A $500 lead submission matters more than a newsletter signup.

### Goal Configuration Checklist

- [ ] All macro conversions defined and tracking
- [ ] Key micro conversions defined and tracking
- [ ] Values assigned where possible
- [ ] Goal names are clear and consistent
- [ ] Goals tested and verified working
- [ ] Documentation of what each goal measures

## Data Quality Controls

### Bot Filtering

| Control | Implementation |
|---------|---------------|
| Known bots | Enable platform's bot filtering (most have this) |
| Suspicious patterns | Filter unusually high pageviews per session |
| Data center traffic | Some platforms filter known data center IPs |
| Custom filters | Filter based on known bot user agents |

**Verification:** Check traffic for suspicious spikes, unusually low bounce rates from certain sources, or unrealistic session durations.

### Internal Traffic Exclusion

| Method | Best For |
|--------|----------|
| IP filtering | Office networks with static IPs |
| Cookie/flag based | Remote teams, dynamic IPs |
| User ID exclusion | Logged-in internal users |
| UTM parameter | Testing traffic (utm_source=internal_test) |

**Verification:** Visit your site from internal network, verify traffic doesn't appear in reports.

### Referral Exclusions

Exclude domains that shouldn't appear as referral sources:

| Exclude | Why |
|---------|-----|
| Your own domains | Cross-subdomain traffic isn't referral |
| Payment processors | Stripe, PayPal redirects aren't referral |
| Auth providers | OAuth redirects aren't referral |
| CDNs | If your CDN appears as referrer |

**Symptom of missing exclusions:** High bounce rate from payment processor "referrals" that are actually returning customers.

### Cross-Domain Tracking

If users cross domains (e.g., shop.example.com to checkout.example.com):

| Check | Requirement |
|-------|-------------|
| Linker configured | Client ID passes between domains |
| Referral exclusion | Domains excluded from referral |
| Testing | Verify session continues across domains |

**Verification:** Start session on domain A, cross to domain B, confirm same session in reports.

### Time and Currency

| Setting | Recommendation |
|---------|----------------|
| Timezone | Match your business timezone for reporting clarity |
| Currency | Single currency for all revenue reporting |
| Consistency | Same settings across all properties/views |

## Verification Process

### Real-Time Testing

| Step | What to Check |
|------|---------------|
| 1 | Visit site in new incognito window |
| 2 | Perform tracked actions (pageview, button click, form submit) |
| 3 | Check real-time reports immediately |
| 4 | Verify event appears with correct properties |

### Conversion Verification

| Comparison | What to Check |
|------------|---------------|
| Analytics conversions vs. system records | Do reported purchases match order system? |
| Analytics signups vs. database | Do reported signups match user database? |
| Analytics leads vs. CRM | Do reported leads match CRM entries? |

**Acceptable variance:** Within 5% is normal (timing, filtering differences). Greater than 10% indicates a problem.

### Traffic Sanity Checks

| Check | What to Look For |
|-------|------------------|
| Traffic trends | Major unexplained spikes or drops |
| Bounce rates | Unusually high (>90%) or low (<20%) rates |
| Session duration | Average durations that seem impossible |
| Geographic data | Unexpected traffic from countries you don't serve |

### Regular Audit Schedule

| Cadence | What to Audit |
|---------|---------------|
| Weekly | Check for tracking anomalies, verify key goals firing |
| Monthly | Compare analytics to source systems, review data quality |
| Quarterly | Full configuration review, filter effectiveness |
| Annually | Major cleanup, deprecate unused goals, review access |

## Common Configuration Issues

| Issue | Symptom | Fix |
|-------|---------|-----|
| Test traffic in production | Spikes during development, fake conversions | Implement environment separation |
| Missing bot filtering | Inflated traffic, low conversion rates | Enable bot filtering |
| Broken goal tracking | Goals show zero or incorrect counts | Re-verify goal configuration |
| Cross-domain session breaks | High "referral" traffic from own domains | Configure cross-domain tracking |
| Internal traffic included | Higher traffic than expected, test conversions | Implement IP or flag filtering |
| Time zone mismatch | Reports don't match business day | Align time zone settings |

## Connecting to Claude

To use these skills with your actual analytics data, Claude needs access to it.

### Connection Methods

| Method | How It Works | Setup Required |
|--------|--------------|----------------|
| Manual export | Export data as CSV, upload to Claude Project | None |
| Screenshots | Capture dashboards/reports, upload as images | None |
| MCP connector | Direct platform connection via Model Context Protocol | MCP setup |

### Manual Export Approach

Works with any platform, no setup required:

1. **Export relevant data** as CSV (traffic, conversions, events)
2. **Upload to Claude Project** as context
3. **Reference in conversation** ("analyze the analytics export I uploaded")

**What to export:**
- Traffic summary by source/medium
- Goal/conversion data
- Event counts and properties
- User/session metrics

### MCP Connectors

Model Context Protocol enables direct platform connections. When available, Claude can query your analytics directly without manual exports.

**Available connectors vary.** Check for MCP connectors for:
- Google Analytics / GA4
- Your data warehouse (BigQuery, Snowflake, etc.)
- BI tools (Looker, Tableau, etc.)

**Benefits of MCP:**
- No manual export step
- Fresh data on every query
- Can pull specific metrics on demand
- Enables more interactive analysis

**If no MCP connector exists** for your platform, manual export works fine. The frameworks in this skill apply regardless of how data reaches Claude.

## Red Flags

- [ ] No separation between production and test data
- [ ] Goals/conversions not configured (only pageviews tracked)
- [ ] Bot traffic significantly inflating numbers
- [ ] Internal traffic mixed with real traffic
- [ ] Analytics numbers don't match source systems
- [ ] Data retention too short for year-over-year analysis
- [ ] Too many people with admin access
- [ ] No documentation of configuration
- [ ] Filters not tested after implementation
- [ ] Cross-domain tracking broken

## Output Format

When auditing or configuring analytics:

```
## Analytics Setup: [Property/Account Name]

### Current State
- Platform: [GA4 / Mixpanel / Amplitude / etc.]
- Property/account: [Name and ID]
- Data collection status: [Active / Issues]

### Environment Structure
| Environment | Property | Status |
|-------------|----------|--------|
| Production | [Name] | [OK / Issues] |
| Staging | [Name] | [OK / Missing] |
| Development | [Name] | [OK / Missing] |

### Goal Configuration

**Macro Conversions:**
| Goal | Status | Value | Notes |
|------|--------|-------|-------|
| [Goal] | [OK/Missing/Broken] | $[Value] | [Notes] |

**Micro Conversions:**
| Goal | Status | Value | Notes |
|------|--------|-------|-------|
| [Goal] | [OK/Missing/Broken] | $[Value] | [Notes] |

### Data Quality Controls

| Control | Status | Notes |
|---------|--------|-------|
| Bot filtering | [Enabled/Disabled] | |
| Internal traffic exclusion | [Configured/Missing] | Method: [IP/Cookie/etc.] |
| Referral exclusions | [Configured/Missing] | Domains: [list] |
| Cross-domain tracking | [Configured/N/A] | |

### Verification Results

| Check | Result | Notes |
|-------|--------|-------|
| Real-time testing | [Pass/Fail] | |
| Conversion accuracy | [X% variance] | Compared to [source] |
| Traffic sanity | [Pass/Issues] | |

### Issues Found

| Priority | Issue | Impact | Fix |
|----------|-------|--------|-----|
| High | [Issue] | [Impact] | [Action needed] |
| Medium | [Issue] | [Impact] | [Action needed] |
| Low | [Issue] | [Impact] | [Action needed] |

### Recommendations
1. [Most important fix]
2. [Second priority]
3. [Third priority]

### Access Review
| User/Team | Current Access | Appropriate? |
|-----------|---------------|--------------|
| [User] | [Level] | [Yes/Reduce] |
```

## Chaining to Other Skills

Analytics setup is the foundation for other measurement skills:

- **Foundation ready, need event tracking** → Chain to `event-tracking`
- **Events tracked, need dashboard** → Chain to `dashboard-design`
- **Need to run experiments** → Chain to `experiment-design`
- **Need attribution analysis** → Chain to `conversion-attribution`

When chaining, pass along: platform being used, goals configured, any known data quality issues.

Design a weekly performance dashboard for my [type of business]. Here's what I care about: [goals]. Here's what I currently track: [metrics]. Give me a dashboard layout with the right metrics and how to organize them.

---
name: dashboard-design
description: Build actionable dashboards. Metric selection, layout principles, avoiding vanity metrics. Use when creating dashboards, simplifying cluttered ones, or setting up reporting.
---

# Dashboard Design

Build dashboards that surface insights and drive decisions. This skill covers metric selection, layout principles, avoiding vanity metrics, and establishing reporting cadence. The goal is dashboards people actually use.

## When to Use This Skill

- Creating a new dashboard from scratch
- Redesigning a cluttered existing dashboard
- Deciding what metrics actually matter
- Setting up regular reporting cadence
- Simplifying reports that have grown unwieldy
- Designing dashboards for different audiences (exec vs. operator)

## The Dashboard Problem

Most dashboards fail because they show too much. Someone asked "what should we track?" and the answer was "everything, just in case." The result: 47 metrics, nobody looks at it, and when something breaks, nobody notices.

Good dashboards answer specific questions quickly. Bad dashboards require 20 minutes of investigation to figure out if anything is wrong.

## Metric Selection

### The Metric Hierarchy

Not all metrics deserve dashboard space. Prioritize:

| Level | Purpose | Dashboard Presence |
|-------|---------|-------------------|
| North Star | The one metric that best captures value | Always visible, top of dashboard |
| Health metrics | 3-5 indicators something is wrong | Always visible, alert if anomalous |
| Input metrics | Levers you can pull | Operational dashboards, not exec |
| Diagnostic metrics | Drill-down for investigation | Available but not on main view |

### Finding Your North Star

| Business Type | Possible North Stars |
|---------------|---------------------|
| SaaS | Weekly Active Users, Monthly Recurring Revenue |
| Ecommerce | Revenue per Visitor, Orders |
| Marketplace | Gross Merchandise Value, Transactions |
| Content | Engaged Time, Subscribers |
| Lead gen | Marketing Qualified Leads, Pipeline Value |

**Test:** If this one number goes up, is the business healthier? If it's not clear, that's not your North Star.

### Health Metrics Selection

Health metrics answer: "Is anything broken right now?"

| Type | Examples | Alert When |
|------|----------|------------|
| Conversion rates | Signup rate, checkout rate | Drop >20% from baseline |
| Error rates | 500 errors, payment failures | Spike above threshold |
| Engagement rates | Email open rate, feature adoption | Significant decline |
| Churn/cancellation | Daily cancellations, refund rate | Spike above normal |

**Rule:** 3-5 health metrics maximum. More than that and you'll ignore them.

### Vanity Metrics vs. Actionable Metrics

| Vanity (Feels Good) | Actionable (Drives Decisions) |
|---------------------|-------------------------------|
| Total pageviews | Conversion rate by page |
| Total registered users | Activated users this period |
| Email list size | Email click-through rate |
| Social followers | Social traffic that converts |
| Time on site (average) | Completion rate of key flows |
| Total revenue (all-time) | Revenue this period vs. last |
| Downloads (all-time) | Active installations |

**The test:** Does this metric tell me what to do differently? If not, it's vanity.

### Metrics by Audience

| Audience | What They Need | Metric Focus |
|----------|---------------|--------------|
| Executive | Is the business healthy? | North Star, revenue, growth rate |
| Manager | Is the team effective? | Team KPIs, goal progress |
| Operator | What needs attention today? | Health metrics, anomalies, queues |
| Analyst | What's driving results? | Full metrics, segmentation, trends |

Don't put everything on one dashboard. Different audiences need different views.

## Layout Principles

### Visual Hierarchy

| Principle | Implementation |
|-----------|---------------|
| Most important top-left | Eye naturally starts there |
| Related metrics grouped | Conversion funnel stages together |
| Consistent sizing | Similar importance = similar size |
| White space matters | Don't pack everything together |

### Dashboard Anatomy

```
┌─────────────────────────────────────────────────┐
│  NORTH STAR METRIC         │  Time Period      │
│  [Big number with trend]   │  [Selector]       │
├─────────────────────────────────────────────────┤
│  HEALTH METRICS (3-5 cards)                     │
│  [Card] [Card] [Card] [Card]                    │
├─────────────────────────────────────────────────┤
│  PRIMARY CHART             │  SECONDARY CHART  │
│  [Main trend/funnel]       │  [Breakdown]      │
├─────────────────────────────────────────────────┤
│  DATA TABLE (if needed)                         │
│  [Detailed breakdown for drilling down]         │
└─────────────────────────────────────────────────┘
```

### Context Is Required

Every metric needs comparison context:

| Context Type | When to Use |
|--------------|-------------|
| vs. Previous period | Daily/weekly performance |
| vs. Same period last year | Seasonal businesses |
| vs. Goal | When targets are set |
| vs. Benchmark | Industry comparisons |

**Never show a number alone.** "Conversion rate: 3.2%" means nothing without context. "Conversion rate: 3.2% (↑12% vs. last week)" tells a story.

### Chart Selection

| Data Type | Best Chart |
|-----------|------------|
| Trend over time | Line chart |
| Part of whole | Pie/donut (sparingly) or stacked bar |
| Comparison across categories | Bar chart |
| Exact values needed | Table |
| Single KPI | Big number card |
| Funnel progression | Funnel chart or horizontal bar |
| Geographic data | Map |

**Default to simplicity.** Tables are often more useful than charts when precision matters.

### What to Avoid

| Anti-pattern | Problem | Alternative |
|--------------|---------|-------------|
| 3D charts | Distorts proportions | Use 2D |
| Dual Y-axes | Confusing, manipulable | Separate charts |
| Pie charts with 10+ slices | Unreadable | Bar chart or table |
| Truncated Y-axes | Exaggerates changes | Full axis or note it |
| Too many colors | Visual noise | 5-7 colors max |
| Animation | Distracting | Static |

## Dashboard Types

### Health Dashboard

**Purpose:** Is anything broken right now?
**Audience:** Operators, on-call
**Update frequency:** Real-time or near-real-time
**Key elements:**
- Status indicators (green/yellow/red)
- Anomaly alerts
- Key conversion rates vs. thresholds
- Error rates
- System performance

### Performance Dashboard

**Purpose:** How are we doing vs. goals?
**Audience:** Team, managers
**Update frequency:** Daily
**Key elements:**
- North Star with trend
- Goal progress (actual vs. target)
- Period-over-period comparison
- Key drivers broken down

### Funnel Dashboard

**Purpose:** Where are we losing people?
**Audience:** Product, growth team
**Update frequency:** Daily or weekly
**Key elements:**
- Stage-by-stage conversion rates
- Drop-off between stages
- Segmentation (by source, device, cohort)
- Comparison vs. previous period

### Channel Dashboard

**Purpose:** What's driving results?
**Audience:** Marketing, growth
**Update frequency:** Weekly
**Key elements:**
- Traffic by channel
- Conversion by channel
- Cost per acquisition by channel
- Channel trends over time

### Executive Dashboard

**Purpose:** Is the business healthy?
**Audience:** Leadership, board
**Update frequency:** Weekly or monthly
**Key elements:**
- Revenue and growth
- Key ratios (LTV/CAC, margins)
- Progress vs. quarterly/annual goals
- Major leading indicators

## Reporting Cadence

### Cadence Framework

| Frequency | What to Include | Audience | Distribution |
|-----------|-----------------|----------|--------------|
| Real-time | Anomaly alerts | Operators | Slack/PagerDuty |
| Daily | Health check, key metrics | Team | Email/Slack summary |
| Weekly | Performance summary, trends | Team + managers | Email + meeting |
| Monthly | Strategic metrics, goals | Leadership | Presentation |
| Quarterly | Business review | Exec/board | Document + meeting |

### Automation Requirements

| Requirement | Why |
|-------------|-----|
| Auto-refresh | Manual updates get skipped |
| Scheduled delivery | Ensures regular review |
| Alert thresholds | Catches problems before check-in |
| Self-service drill-down | Reduces ad-hoc requests |

## Building a New Dashboard

### Step 1: Define the Audience

| Question | Answer for this dashboard |
|----------|--------------------------|
| Who will use this? | |
| What decisions will they make? | |
| How often will they check it? | |
| What questions should it answer? | |

### Step 2: Select Metrics

| Metric | Why It Matters | Target/Threshold |
|--------|---------------|------------------|
| [Metric 1] | | |
| [Metric 2] | | |
| [Metric 3] | | |

**Constraint:** Start with 5-7 metrics maximum. Add more only if necessary.

### Step 3: Design Layout

Sketch before building:
- What goes top-left (most important)?
- What groups together?
- What needs comparison context?
- What chart type for each metric?

### Step 4: Add Context and Alerts

- Set comparison periods
- Define alert thresholds
- Add goal lines where applicable
- Include last-updated timestamp

### Step 5: Test with Users

- Does it answer their questions?
- What's confusing?
- What's missing?
- What's unnecessary?

### Step 6: Document

- What does each metric measure?
- Where does the data come from?
- What do alert thresholds mean?
- Who owns this dashboard?

## Connecting Dashboards to Claude

To analyze dashboard data with Claude, you need to get that data into the conversation.

### Connection Methods

| Method | How It Works | Best For |
|--------|--------------|----------|
| Screenshot upload | Capture dashboard, upload as image | Quick questions, visual context |
| Data export | Export underlying data as CSV | Deeper analysis, calculations |
| MCP connector | Direct connection to BI tool or data source | Ongoing analysis, live data |

### Screenshot Approach

For quick analysis or design feedback:

1. **Capture the dashboard** as image
2. **Upload to Claude**
3. **Ask specific questions** ("What stands out?" "What's missing?" "How would you redesign this?")

Good for: Dashboard reviews, layout feedback, identifying issues

### Data Export Approach

For deeper analysis:

1. **Export underlying metrics** as CSV
2. **Upload to Claude Project**
3. **Reference in analysis** ("Analyze trends in the dashboard data I uploaded")

Good for: Trend analysis, anomaly detection, recommendations

### MCP Connectors

Model Context Protocol enables direct connections to data sources. When available, Claude can pull fresh dashboard data without manual exports.

**Relevant MCP connectors:**
- Data warehouses (BigQuery, Snowflake, Redshift)
- BI tools (Looker, Tableau, Metabase)
- Analytics platforms (GA4, Mixpanel)

**Benefits:**
- No manual export step
- Always current data
- Can query specific metrics on demand
- Interactive analysis across multiple data points

**If no connector exists:** Screenshots and CSV exports work fine. The frameworks in this skill apply regardless of how data reaches Claude.

## Red Flags

- [ ] Dashboard has 20+ metrics
- [ ] No clear hierarchy (everything equally emphasized)
- [ ] Numbers shown without comparison context
- [ ] Nobody actually looks at it regularly
- [ ] Updated manually (not automated)
- [ ] Different dashboards show conflicting numbers
- [ ] Vanity metrics prominently featured
- [ ] No date range controls or filters
- [ ] No documentation of what metrics mean
- [ ] Metric definitions changed without notice
- [ ] Alerts that fire so often they're ignored
- [ ] Dashboards created and then abandoned

## Output Format

When designing a dashboard:

```
## Dashboard Design: [Dashboard Name]

### Audience and Purpose
- **Primary audience:** [Who]
- **Purpose:** [What questions it answers]
- **Check frequency:** [How often reviewed]
- **Update frequency:** [How often data refreshes]

### Metrics

**North Star:**
| Metric | Definition | Target | Comparison |
|--------|------------|--------|------------|
| [Metric] | [What it measures] | [Goal] | [vs. what] |

**Health Metrics:**
| Metric | Definition | Threshold | Alert When |
|--------|------------|-----------|------------|
| [Metric] | [Definition] | [Normal range] | [Condition] |

**Supporting Metrics:**
| Metric | Definition | Why Included |
|--------|------------|--------------|
| [Metric] | [Definition] | [Purpose] |

### Layout

[ASCII sketch or description of layout]

```
┌────────────────────────────────┐
│ [Section 1]    │ [Section 2]   │
├────────────────────────────────┤
│ [Section 3]                    │
└────────────────────────────────┘
```

### Visualization Specifications

| Metric/Section | Chart Type | Notes |
|----------------|------------|-------|
| [Metric] | [Chart type] | [Special requirements] |

### Filters and Controls

| Filter | Options | Default |
|--------|---------|---------|
| Time period | [Options] | [Default] |
| Segment | [Options] | [Default] |

### Alerts

| Alert | Condition | Notify | Channel |
|-------|-----------|--------|---------|
| [Name] | [Trigger] | [Who] | [How] |

### Data Sources

| Metric | Source | Refresh |
|--------|--------|---------|
| [Metric] | [Where data comes from] | [How often] |

### Documentation
[Link to detailed metric definitions]

### Owner
[Who maintains this dashboard]
```

## Chaining to Other Skills

Dashboard design connects to other measurement skills:

- **Need analytics foundation** → Chain to `analytics-setup`
- **Need event tracking for metrics** → Chain to `event-tracking`
- **Dashboard shows test results** → Chain to `experiment-design`
- **Need channel breakdown** → Chain to `conversion-attribution`

When chaining, pass along: metrics needed, audience requirements, and data sources available.

Design an event tracking plan for my [type of business]. Here's my main funnel: [steps]. What events should I track, what properties should each have, and how should I name them consistently?

Design an A/B test for [page/element]. Current conversion rate: [X%]. I want to detect at least a [Y%] improvement. Traffic: [Z] visitors per week. Give me the test design, expected duration, and how to interpret results.

Help me understand attribution for my [type of business]. Here are my channels: [list]. Here's my current attribution data: [summary]. Average time from first touch to conversion: [X days]. How should I think about what's actually driving conversions?"