make-mp-data 2.1.11 → 3.0.1

package/lib/templates/funnels-instructions.txt

@@ -0,0 +1,272 @@
+You are an AI assistant that generates funnel definitions for an existing Dungeon schema. Funnels represent user journeys through a product - the sequences of events users take as they move toward goals like purchasing, subscribing, or completing onboarding.
+
+## What are Funnels?
+
+A funnel is a sequence of events that represents a user journey. For example:
+- **Signup funnel**: signup → onboarding_start → onboarding_complete → first_purchase
+- **Checkout funnel**: add_to_cart → view_cart → begin_checkout → complete_purchase
+- **Feature adoption funnel**: discover_feature → try_feature → adopt_feature
+
+Funnels have conversion rates (what % of users complete the full journey) and can include properties that carry across ALL steps - this is powerful for analytics because you can segment funnels by these properties.
+
+## The Current Schema
+
+The user has already generated a dungeon schema with these events and properties:
+
+--------------
+
+<CURRENT_SCHEMA>
+
+--------------
+
+## CRITICAL - USE ONLY EXISTING EVENTS:
+
+You MUST use the EXACT event names from the schema above. These are CASE-SENSITIVE and must match EXACTLY.
+
+DO NOT:
+- Invent new event names that don't exist in the schema
+- Change the case of any names (e.g., "sign up" ≠ "Sign Up" ≠ "signup")
+- Use similar but different names (e.g., "purchase" ≠ "checkout" ≠ "buy")
+
+DO:
+- Copy event names exactly as they appear in the events array
+- Build sequences from the available events only
+- Create meaningful user journeys that make sense for the business
+
+--------------
+
+## Funnel Structure
+
+Each funnel is a JavaScript object with these properties:
+
+```javascript
+{
+    sequence: ["event1", "event2", "event3"],  // REQUIRED: Array of event names (must exist in schema)
+    conversionRate: 15,                         // REQUIRED: Integer 0-100 representing percentage (e.g., 15 = 15%)
+    props: {                                    // OPTIONAL: Properties that apply to ALL events in this funnel
+        experiment_variant: ["control", "variant_a", "variant_b"],
+        signup_source: ["organic", "paid", "referral"]
+    },
+    conditions: { plan: "free" },               // OPTIONAL: Only users matching this condition run this funnel
+    order: "sequential",                        // OPTIONAL: "sequential" (default), "random", "first-fixed"
+    timeToConvert: 24                           // OPTIONAL: Hours to complete funnel (default: 1)
+}
+```
+
+--------------
+
+## IMPORTANT: Funnel-Specific Properties (props)
+
+Funnel props are POWERFUL because they carry across ALL steps of the funnel. This means you can:
+- Segment funnel analytics by these properties
+- Track experiments across the entire journey
+- Analyze cohort behavior through conversion paths
+
+### Categories of Funnel Props to Consider:
+
+**1. Experiment/Variant Properties** (for A/B testing and feature flags)
+
+NOTE: For Mixpanel experiments, the standard format is:
+- Event: `$experiment_started`
+- Properties: `Experiment name` and `Variant name`
+
+For funnel props, you can track experiment variants that carry across all funnel steps:
+```javascript
+props: {
+    "Experiment name": ["Checkout Flow Test", "Onboarding Test", "Pricing Test"],
+    "Variant name": ["control", "variant_a", "variant_b"],
+    test_group: ["holdout", "treatment", "pilot"],
+    feature_flag: ["enabled", "disabled"],
+    onboarding_flow: ["guided", "self_serve", "video_intro"]
+}
+```
+
+**2. Cohort/Segment Properties** (for user segmentation)
+```javascript
+props: {
+    acquisition_channel: ["organic", "google_ads", "facebook", "referral", "influencer"],
+    signup_cohort: ["q1_2024", "q2_2024", "q3_2024", "q4_2024"],
+    user_segment: ["power_user", "casual", "new", "returning"],
+    geographic_region: ["north_america", "europe", "asia", "latam"]
+}
+```
+
+**3. Flow Tracking Properties** (for journey analysis)
+```javascript
+props: {
+    entry_point: ["homepage", "product_page", "search", "deep_link", "email"],
+    device_at_start: ["mobile", "desktop", "tablet"],
+    session_type: ["first_session", "returning", "reactivated"],
+    traffic_source: ["direct", "seo", "sem", "social", "email"]
+}
+```
+
+**4. Business-Specific Properties** (for domain analysis)
+```javascript
+// E-commerce
+props: {
+    cart_value_tier: ["low", "medium", "high", "premium"],
+    coupon_type: ["none", "percentage", "fixed", "free_shipping"],
+    product_category: ["electronics", "clothing", "home", "sports"]
+}
+
+// SaaS
+props: {
+    trial_type: ["7_day", "14_day", "30_day", "enterprise"],
+    plan_intent: ["starter", "pro", "enterprise"],
+    team_size_bucket: ["solo", "small", "medium", "large"]
+}
+
+// Media/Content
+props: {
+    content_source: ["recommended", "search", "trending", "bookmarked"],
+    subscription_status: ["free", "trial", "premium"],
+    engagement_level: ["passive", "active", "power_viewer"]
+}
+```
+
+--------------
+
+## Conditional Funnels
+
+Use conditions to create segment-specific funnels. This models how different user types have different journeys:
+
+```javascript
+// Free users have a trial-to-upgrade journey
+{
+    sequence: ["view_premium_feature", "start_trial", "use_trial", "upgrade"],
+    conversionRate: 8,
+    conditions: { plan: "free" },
+    props: {
+        upgrade_trigger: ["paywall", "feature_limit", "storage_limit", "proactive"],
+        trial_source: ["in_app", "email", "push"]
+    }
+}
+
+// Premium users have a feature adoption journey
+{
+    sequence: ["discover_feature", "try_feature", "use_feature_regularly"],
+    conversionRate: 45,
+    conditions: { plan: "premium" },
+    props: {
+        feature_discovery: ["tooltip", "release_notes", "support", "organic"]
+    }
+}
+```
+
+--------------
+
+## Conversion Rate Guidelines
+
+Realistic conversion rates vary by funnel type:
+
+- **High-intent funnels** (user already committed): 40-70%
+  - Login → Use feature: 65%
+  - Add to cart → Checkout: 50%
+
+- **Medium-intent funnels** (user exploring): 15-40%
+  - Browse → Add to cart: 25%
+  - Start onboarding → Complete: 35%
+
+- **Low-intent funnels** (user just arrived): 5-20%
+  - Visit → Signup: 10%
+  - Signup → First purchase: 12%
+
+- **Long/complex funnels** (many steps): 5-15%
+  - Full e-commerce journey: 8%
+  - Multi-step onboarding: 12%
+
+--------------
+
+## Example Funnels
+
+**E-commerce Purchase Journey**
+```javascript
+{
+    sequence: ["page view", "view item", "add to cart", "checkout"],
+    conversionRate: 12,
+    props: {
+        experiment_variant: ["control", "new_checkout", "express_checkout"],
+        traffic_source: ["organic", "paid", "referral", "email"],
+        device_type: ["mobile", "desktop"]
+    }
+}
+```
+
+**SaaS Onboarding Journey**
+```javascript
+{
+    sequence: ["sign up", "complete_profile", "create_first_project", "invite_team"],
+    conversionRate: 25,
+    props: {
+        onboarding_flow: ["guided_tour", "video_intro", "self_directed"],
+        signup_source: ["google", "github", "email", "sso"],
+        company_size: ["solo", "small_team", "enterprise"]
+    },
+    conditions: { plan: "free" }
+}
+```
+
+**Content Engagement Journey**
+```javascript
+{
+    sequence: ["page view", "watch video", "save item", "share item"],
+    conversionRate: 8,
+    props: {
+        content_source: ["recommended", "search", "trending", "followed"],
+        user_segment: ["casual", "regular", "power_user"]
+    }
+}
+```
+
+**Feature Adoption with Experiment**
+```javascript
+{
+    sequence: ["view_feature", "enable_feature", "use_feature", "recommend_feature"],
+    conversionRate: 20,
+    props: {
+        experiment_variant: ["control", "prominent_cta", "video_demo", "social_proof"],
+        user_tier: ["free", "pro", "enterprise"]
+    }
+}
+```
+
+--------------
+
+## Output Format
+
+Return a JavaScript object with a funnels array:
+
+```javascript
+{
+    funnels: [
+        {
+            sequence: ["event1", "event2", "event3"],
+            conversionRate: 15,
+            props: {
+                // Funnel-specific properties
+            }
+        },
+        {
+            sequence: ["event1", "event2"],
+            conversionRate: 30,
+            conditions: { plan: "premium" }
+        }
+    ]
+}
+```
+
+--------------
+
+## Final Rules
+
+1. **ONLY use event names that exist in the current schema** - check the events array carefully
+2. Each funnel should tell a story (e.g., "signup journey", "purchase flow", "feature adoption")
+3. Include funnel props when they add analytical value - especially for experiments and cohorts
+4. Use conditions for segment-specific behaviors (free vs premium, new vs returning)
+5. Conversion rates are integers 0-100 representing percentage - rarely above 50 unless high-intent
+6. Create 3-6 funnels that represent different user journeys
+7. Do NOT include explanations - only output the JavaScript object
+8. Start with { and end with }
+
+Generate funnels based on the user's description of the user journeys and conversion paths they want to track.

package/lib/templates/hook-examples.json

@@ -0,0 +1,187 @@
+{
+  "examples": [
+    {
+      "id": "time-based-search-degradation",
+      "prompt": "Search quality degraded 21 days ago. Since then, 50% of searches return zero results and 18% of users churn immediately after searching.",
+      "hookTypes": ["event"],
+      "patterns": ["time-based", "quality degradation", "churn"],
+      "response": "// Constants - Edit these values to customize the trend\nconst DAYS_AGO_SEARCH_BROKE = 21;\nconst LIKELIHOOD_ZERO_RESULTS = 50;\nconst LIKELIHOOD_CHURN_AFTER_SEARCH = 18;\n\nif (type === \"event\") {\n    const NOW = dayjs();\n    const TIME_WHEN_SEARCH_GOT_BAD = NOW.subtract(DAYS_AGO_SEARCH_BROKE, 'days');\n    const EVENT_TIME = dayjs(record.time);\n\n    // After search quality degraded\n    if (EVENT_TIME.isAfter(TIME_WHEN_SEARCH_GOT_BAD)) {\n        // Search returns no results more often\n        if (record.event === \"search\") {\n            if (chance.bool({ likelihood: LIKELIHOOD_ZERO_RESULTS })) {\n                record[\"results count\"] = 0;\n            }\n        }\n\n        // Users churn (remove event)\n        if (chance.bool({ likelihood: LIKELIHOOD_CHURN_AFTER_SEARCH })) {\n            return {};\n        }\n    }\n}\nreturn record;",
+      "useCase": "Model product quality issues and their impact on user behavior",
+      "complexity": "simple"
+    },
+    {
+      "id": "ab-test-watch-time-subscription",
+      "prompt": "Users in variant A for the video recommendation experiment watch 70% longer and are 75% more likely to subscribe compared to users in variant B",
+      "hookTypes": ["everything"],
+      "patterns": ["A/B testing", "property correlation", "event injection", "conversion engineering"],
+      "response": "// Constants - Edit these to customize experiment effects\nconst DAYS_AGO_EXPERIMENT_LAUNCHED = 14;\nconst VARIANT_A_WATCH_TIME_BOOST = 1.7; // 70% increase\nconst VARIANT_A_SUBSCRIBE_LIKELIHOOD = 75;\nconst VARIANT_B_WATCH_TIME_REDUCTION = 0.5; // 50% decrease\nconst VARIANT_B_UNSUBSCRIBE_LIKELIHOOD = 75;\n\nif (type === \"everything\") {\n    const NOW = dayjs();\n    const TIME_WE_EXPERIMENTED = NOW.subtract(DAYS_AGO_EXPERIMENT_LAUNCHED, 'days');\n\n    // Find which variant the user was in\n    const hadVariantA = record.some(event =>\n        event.event === \"$experiment_started\" &&\n        event[\"Variant name\"] === \"A\"\n    );\n\n    const hadVariantB = record.some(event =>\n        event.event === \"$experiment_started\" &&\n        event[\"Variant name\"] === \"B\"\n    );\n\n    // Modify events based on variant\n    record.forEach((event, idx) => {\n        const EVENT_TIME = dayjs(event.time);\n\n        // Only apply changes after experiment launch\n        if (EVENT_TIME.isAfter(TIME_WE_EXPERIMENTED)) {\n            if (hadVariantA) {\n                // Variant A: Better outcomes\n                if (event.event === \"watch video\" && chance.bool({ likelihood: VARIANT_A_SUBSCRIBE_LIKELIHOOD })) {\n                    // Increase watch time\n                    event[\"watch time\"] = v.round(event[\"watch time\"] * VARIANT_A_WATCH_TIME_BOOST);\n\n                    // Add subscribe event after watching\n                    const subscribeEvent = {\n                        event: \"subscribe\",\n                        time: dayjs(event.time).add(1, 'minute').toISOString(),\n                        user_id: event.user_id\n                    };\n                    record.splice(idx + 1, 0, subscribeEvent);\n                }\n            } else if (hadVariantB) {\n                // Variant B: Worse outcomes\n                if (event.event === \"subscribe\" && chance.bool({ likelihood: VARIANT_B_UNSUBSCRIBE_LIKELIHOOD })) {\n                    // Remove subscribe events\n                    record.splice(idx, 1);\n                }\n\n                if (event.event === \"watch video\") {\n                    // Decrease watch time\n                    event[\"watch time\"] = v.round(event[\"watch time\"] * VARIANT_B_WATCH_TIME_REDUCTION);\n                }\n            }\n        }\n    });\n}\nreturn record;",
+      "useCase": "Create realistic A/B test results with clear winner and loser variants",
+      "complexity": "advanced"
+    },
+    {
+      "id": "revenue-boost-after-improvement",
+      "prompt": "After the improvement 15 days ago, checkout amounts increased 50% and video watch times increased 50%, but before that date, overall volume was down 33%",
+      "hookTypes": ["event"],
+      "patterns": ["time-based", "revenue engineering", "property manipulation"],
+      "response": "// Constants - Edit these to customize the trend\nconst DAYS_AGO_IMPROVEMENT = 15;\nconst REVENUE_BOOST_AFTER = 1.5; // 50% increase\nconst WATCH_TIME_BOOST_AFTER = 1.5; // 50% increase\nconst VOLUME_DROP_BEFORE = 33; // % of events to drop\n\nif (type === \"event\") {\n    const NOW = dayjs();\n    const IMPROVEMENT_DATE = NOW.subtract(DAYS_AGO_IMPROVEMENT, 'day');\n    const EVENT_TIME = dayjs(record.time);\n\n    // After improvement: Better metrics\n    if (EVENT_TIME.isAfter(IMPROVEMENT_DATE)) {\n        if (record.event === \"checkout\") {\n            record.amount = Math.round(record.amount * REVENUE_BOOST_AFTER);\n        }\n        if (record.event === \"watch video\") {\n            record.watchTimeSec = Math.round(record.watchTimeSec * WATCH_TIME_BOOST_AFTER);\n        }\n    }\n\n    // Before improvement: Lower volume\n    if (EVENT_TIME.isBefore(IMPROVEMENT_DATE)) {\n        if (chance.bool({ likelihood: VOLUME_DROP_BEFORE })) {\n            return {}; // Remove event\n        }\n    }\n}\nreturn record;",
+      "useCase": "Model product improvements and their impact on key metrics",
+      "complexity": "simple"
+    },
+    {
+      "id": "premium-users-higher-engagement",
+      "prompt": "Users with custom theme have 3x more checkout events compared to users with light or dark themes",
+      "hookTypes": ["everything"],
+      "patterns": ["user segmentation", "property correlation", "event duplication"],
+      "response": "// Constants - Edit these to customize user segmentation\nconst PREMIUM_THEME_VALUE = \"custom\";\nconst PREMIUM_CHECKOUT_MULTIPLIER = 3;\nconst TIME_VARIANCE_HOURS = 48;\n\nif (type === \"everything\") {\n    // Count events by theme\n    const numCustomMode = record.filter(a => a.theme === PREMIUM_THEME_VALUE).length;\n    const numLightMode = record.filter(a => a.theme === 'light').length;\n    const numDarkMode = record.filter(a => a.theme === 'dark').length;\n\n    // Premium theme users get more checkout events\n    if (numCustomMode > numLightMode || numCustomMode > numDarkMode) {\n        const checkoutEvents = record.filter(a => a.event === 'checkout');\n\n        // Create additional checkouts\n        const newCheckouts = checkoutEvents.map(a => {\n            const randomInt = u.integer(-TIME_VARIANCE_HOURS, TIME_VARIANCE_HOURS);\n            return {\n                ...a,\n                time: dayjs(a.time).add(randomInt, 'hour').toISOString(),\n                event: \"checkout\",\n                amount: a.amount * 2,\n                coupon: \"50%OFF\"\n            };\n        });\n\n        // Add new checkout events\n        record.push(...newCheckouts);\n    }\n}\nreturn record;",
+      "useCase": "Create correlations between user properties and behavior",
+      "complexity": "intermediate"
+    },
+    {
+      "id": "low-quality-users-churn",
+      "prompt": "Users who watch low quality videos (480p, 360p, 240p) more than high quality videos have a 50% chance of churning halfway through their lifecycle",
+      "hookTypes": ["everything"],
+      "patterns": ["churn modeling", "quality correlation", "event deletion"],
+      "response": "// Constants - Edit these to customize churn behavior\nconst LOW_QUALITY_FORMATS = [\"480p\", \"360p\", \"240p\"];\nconst CHURN_LIKELIHOOD = 50;\n\nif (type === \"everything\") {\n    // Count video quality\n    const lowQualityWatches = record.filter(a =>\n        a.event === 'watch video' &&\n        LOW_QUALITY_FORMATS.includes(a.quality)\n    );\n\n    const highQualityWatches = record.filter(a =>\n        a.event === 'watch video' &&\n        !LOW_QUALITY_FORMATS.includes(a.quality)\n    );\n\n    // Users who watch more low quality content churn\n    if (lowQualityWatches.length > highQualityWatches.length) {\n        if (chance.bool({ likelihood: CHURN_LIKELIHOOD })) {\n            // Remove second half of events (churn)\n            const midpoint = Math.floor(record.length / 2);\n            record = record.slice(0, midpoint);\n        }\n    }\n}\nreturn record;",
+      "useCase": "Model churn based on product quality or engagement patterns",
+      "complexity": "intermediate"
+    },
+    {
+      "id": "infrastructure-outage",
+      "prompt": "AWS outage from October 19 to October 22 caused 88% of events to be dropped during that time window",
+      "hookTypes": ["event"],
+      "patterns": ["time-based", "infrastructure issues", "event deletion"],
+      "response": "// Constants - Edit these to customize the outage\nconst OUTAGE_START = \"2025-10-19\";\nconst OUTAGE_END = \"2025-10-22\";\nconst EVENT_DROP_RATE = 88;\n\nif (type === \"event\") {\n    const AWS_OUTAGE_START = dayjs(OUTAGE_START);\n    const AWS_OUTAGE_END = dayjs(OUTAGE_END);\n    const EVENT_TIME = dayjs(record.time);\n\n    // During outage, drop most events\n    if (EVENT_TIME.isAfter(AWS_OUTAGE_START) && EVENT_TIME.isBefore(AWS_OUTAGE_END)) {\n        if (chance.bool({ likelihood: EVENT_DROP_RATE })) {\n            return {}; // Remove event\n        }\n    }\n}\nreturn record;",
+      "useCase": "Simulate infrastructure failures and their data impact",
+      "complexity": "simple"
+    },
+    {
+      "id": "product-launch-adoption",
+      "prompt": "After the BYND stock squeeze on October 22, 10% of all trades are for that symbol",
+      "hookTypes": ["event"],
+      "patterns": ["time-based", "product launch", "property substitution"],
+      "response": "// Constants - Edit these to customize product launch\nconst LAUNCH_DATE = \"2025-10-22\";\nconst NEW_PRODUCT_NAME = \"BYND\";\nconst ADOPTION_RATE = 10;\nconst PROPERTY_NAME = \"asset_symbol\";\n\nif (type === \"event\") {\n    const PRODUCT_LAUNCH = dayjs(LAUNCH_DATE);\n    const EVENT_TIME = dayjs(record.time);\n\n    // After launch, some users adopt new product\n    if (EVENT_TIME.isAfter(PRODUCT_LAUNCH)) {\n        if (record.hasOwnProperty(PROPERTY_NAME)) {\n            if (chance.bool({ likelihood: ADOPTION_RATE })) {\n                record[PROPERTY_NAME] = NEW_PRODUCT_NAME;\n            }\n        }\n    }\n}\nreturn record;",
+      "useCase": "Model new product launches and adoption curves",
+      "complexity": "simple"
+    },
+    {
+      "id": "email-campaign-lift",
+      "prompt": "Users who read marketing emails increase their purchase amounts by 10% per email read, up to a maximum of 50% total increase",
+      "hookTypes": ["everything"],
+      "patterns": ["marketing attribution", "cumulative effects", "property correlation"],
+      "response": "// Constants - Edit these to customize campaign effect\nconst EMAIL_UTM_MEDIUM = \"email (marketing)\";\nconst BOOST_PER_EMAIL = 0.10; // 10% per email\nconst MAX_BOOST = 0.50; // Max 50% boost\nconst PROPERTY_TO_BOOST = \"amount_usd\";\n\nif (type === \"everything\") {\n    // Count email reads\n    const emailReads = record.filter(e => {\n        if (e.hasOwnProperty('utm_medium')) {\n            if (e.utm_medium === EMAIL_UTM_MEDIUM) {\n                return true;\n            }\n        }\n        return false;\n    });\n\n    const numEmailReads = emailReads.length;\n\n    // Apply cumulative boost based on email engagement\n    if (numEmailReads > 0) {\n        record.forEach(e => {\n            if (e.hasOwnProperty(PROPERTY_TO_BOOST)) {\n                // Calculate boost, capped at max\n                const boostFactor = 1 + Math.min(numEmailReads * BOOST_PER_EMAIL, MAX_BOOST);\n                e[PROPERTY_TO_BOOST] = Math.round(e[PROPERTY_TO_BOOST] * boostFactor);\n            }\n        });\n    }\n}\nreturn record;",
+      "useCase": "Model cumulative marketing campaign effects on user behavior",
+      "complexity": "intermediate"
+    },
+    {
+      "id": "loan-approval-rate-change",
+      "prompt": "Before the rate change 17 days ago, student loans were declined 60% of the time. After the change, student loans convert to auto loans 60% of the time, and auto loans get approved 60% of the time but take 1-20 minutes longer",
+      "hookTypes": ["funnel-post"],
+      "patterns": ["time-based", "funnel manipulation", "approval engineering", "timing manipulation"],
+      "response": "// Constants - Edit these to customize approval rates\nconst DAYS_AGO_RATE_CHANGE = 17;\nconst BEFORE_DECLINE_RATE = 60;\nconst AFTER_CONVERSION_RATE = 60;\nconst AFTER_APPROVAL_RATE = 60;\nconst MIN_DELAY_MINUTES = 1;\nconst MAX_DELAY_MINUTES = 20;\n\nif (type === \"funnel-post\") {\n    const RATE_CHANGE_DATE = dayjs().subtract(DAYS_AGO_RATE_CHANGE, 'd');\n\n    // Check if this is a loan application funnel\n    if (record[0]?.event === \"loan app start\") {\n        const funnelStartTime = dayjs(record[0].time);\n\n        // Before rate change: Student loans declined\n        if (funnelStartTime.isBefore(RATE_CHANGE_DATE)) {\n            if (record[0][\"loan type\"] === \"student loan\") {\n                if (chance.bool({ likelihood: BEFORE_DECLINE_RATE })) {\n                    record.forEach((r) => {\n                        r[\"application status\"] = \"declined\";\n                    });\n                }\n            }\n        }\n\n        // After rate change: Better outcomes\n        if (funnelStartTime.isAfter(RATE_CHANGE_DATE)) {\n            // Convert student loans to auto loans\n            if (record[0][\"loan type\"] === \"student loan\") {\n                if (chance.bool({ likelihood: AFTER_CONVERSION_RATE })) {\n                    record.forEach((r) => {\n                        r[\"loan type\"] = \"auto loan\";\n                    });\n                }\n            }\n\n            // Auto loans get approved\n            if (record[0][\"loan type\"] === \"auto loan\") {\n                if (chance.bool({ likelihood: AFTER_APPROVAL_RATE })) {\n                    record.forEach((r) => {\n                        r[\"application status\"] = \"approved\";\n                    });\n                }\n\n                // But they take longer\n                record.forEach((ev, index) => {\n                    if (index > 0) {\n                        const minutes = u.integer(MIN_DELAY_MINUTES, MAX_DELAY_MINUTES);\n                        ev.time = dayjs(ev.time).add(minutes, \"minute\").toISOString();\n                    }\n                });\n            }\n        }\n    }\n}\nreturn record;",
+      "useCase": "Engineer complex funnel behavior changes over time",
+      "complexity": "advanced"
+    },
+    {
+      "id": "ai-model-product-adoption",
+      "prompt": "Users exposed to AI model Q use more products. Users exposed to model BERT use fewer products. Product count directly affects event volume: users with 3+ products have 40% more events, users with <3 products have 40% fewer events",
+      "hookTypes": ["everything"],
+      "patterns": ["AI impact", "product adoption", "event volume correlation", "retention"],
+      "response": "// Constants - Edit these to customize AI impact\nconst HIGH_PERFORMING_MODEL = \"Q\";\nconst LOW_PERFORMING_MODEL = \"BERT\";\nconst PRODUCT_THRESHOLD = 3;\nconst HIGH_PRODUCT_EVENT_BOOST_LIKELIHOOD = 40;\nconst LOW_PRODUCT_EVENT_REDUCTION_LIKELIHOOD = 40;\nconst EVENT_DUPLICATION_RATE = 20;\nconst EVENT_DELETION_RATE = 20;\nconst MIN_EVENT_DELAY_MINUTES = 1;\nconst MAX_EVENT_DELAY_MINUTES = 90;\n\nif (type === \"everything\") {\n    // Count AI model exposures\n    const AIModelExposure = record.map(a => a[\"AI model\"]).filter(Boolean);\n    const numHighPerformingExposure = AIModelExposure.filter(a => a === HIGH_PERFORMING_MODEL).length;\n    const numLowPerformingExposure = AIModelExposure.filter(a => a === LOW_PERFORMING_MODEL).length;\n\n    // AI exposure affects product count\n    if (chance.bool({ likelihood: 50 })) {\n        const ratio = Math.floor(numHighPerformingExposure / numLowPerformingExposure) || 1;\n        // Note: This assumes a listOfProducts() function exists in meta.defaults\n        // Adjust based on actual schema\n    }\n\n    const { profile } = meta;\n\n    // High product users: More events and retention\n    if (profile[\"products used\"] && profile[\"products used\"].length > PRODUCT_THRESHOLD) {\n        if (chance.bool({ likelihood: HIGH_PRODUCT_EVENT_BOOST_LIKELIHOOD })) {\n            // Duplicate events randomly\n            record.forEach((ev, index) => {\n                if (index > 0) {\n                    if (chance.bool({ likelihood: EVENT_DUPLICATION_RATE })) {\n                        const clone = u.clone(ev);\n                        const delay = u.integer(MIN_EVENT_DELAY_MINUTES, MAX_EVENT_DELAY_MINUTES);\n                        clone.time = dayjs(ev.time).add(delay, \"minute\").toISOString();\n                        record.push(clone);\n                    }\n                }\n            });\n        }\n    }\n\n    // Low product users: Fewer events and worse retention\n    if (profile[\"products used\"] && profile[\"products used\"].length < PRODUCT_THRESHOLD) {\n        if (chance.bool({ likelihood: LOW_PRODUCT_EVENT_REDUCTION_LIKELIHOOD })) {\n            // Delete events randomly\n            record = record.filter((ev, index) => {\n                return !(index > 0 && chance.bool({ likelihood: EVENT_DELETION_RATE }));\n            });\n        }\n    }\n}\nreturn record;",
+      "useCase": "Model AI/ML feature impact on product adoption and engagement",
+      "complexity": "advanced"
+    },
+    {
+      "id": "department-role-correlation",
+      "prompt": "Users in the engineering department are admin 50% of the time, while users in sales have no access 50% of the time",
+      "hookTypes": ["scd-pre"],
+      "patterns": ["user segmentation", "role-based access", "SCD manipulation"],
+      "response": "// Constants - Edit these to customize role assignments\nconst ENGINEERING_DEPT = \"engineering\";\nconst SALES_DEPT = \"sales\";\nconst ENGINEERING_ADMIN_LIKELIHOOD = 50;\nconst SALES_NO_ACCESS_LIKELIHOOD = 50;\nconst ADMIN_ROLE = \"admin\";\nconst NO_ACCESS_ROLE = \"no access\";\n\nif (type === \"scd-pre\") {\n    const { type: scdType, profile, scd } = meta;\n    const prop = Object.keys(scd)[0];\n\n    // Only apply to user SCDs for role property\n    if (scdType === \"user\" && prop === \"role\") {\n        // Engineering users are admins more often\n        if (profile.department === ENGINEERING_DEPT) {\n            for (const entry of record) {\n                if (chance.bool({ likelihood: ENGINEERING_ADMIN_LIKELIHOOD })) {\n                    entry.role = ADMIN_ROLE;\n                }\n            }\n        }\n\n        // Sales users have no access more often\n        if (profile.department === SALES_DEPT) {\n            for (const entry of record) {\n                if (chance.bool({ likelihood: SALES_NO_ACCESS_LIKELIHOOD })) {\n                    entry.role = NO_ACCESS_ROLE;\n                }\n            }\n        }\n    }\n}\nreturn record;",
+      "useCase": "Create realistic role-based access patterns based on user attributes",
+      "complexity": "intermediate"
+    },
+    {
+      "id": "nps-improvement-over-time",
+      "prompt": "Users with 3 or more preferred features see NPS scores improve by 10% month over month",
+      "hookTypes": ["scd-pre"],
+      "patterns": ["progressive improvement", "feature adoption impact", "SCD manipulation"],
+      "response": "// Constants - Edit these to customize NPS improvement\nconst FEATURE_THRESHOLD = 3;\nconst MONTHLY_NPS_IMPROVEMENT = 0.10; // 10% improvement\nconst MAX_NPS_SCORE = 10;\n\nif (type === \"scd-pre\") {\n    const { type: scdType, profile, scd } = meta;\n    const prop = Object.keys(scd)[0];\n\n    // Only apply to user NPS property\n    if (scdType === \"user\" && prop === \"NPS\") {\n        // Check if user has enough preferred features\n        const hasEnoughFeatures = profile[\"preferred features\"] &&\n            profile[\"preferred features\"].length >= FEATURE_THRESHOLD;\n\n        if (hasEnoughFeatures) {\n            // Apply progressive improvement over time\n            let lastNPS = record[0].NPS;\n\n            for (const [index, entry] of record.entries()) {\n                if (index === 0) continue; // Skip first entry\n\n                // Increase NPS by improvement factor\n                entry.NPS = Math.floor(lastNPS + (lastNPS * MONTHLY_NPS_IMPROVEMENT));\n\n                // Cap at max score\n                if (entry.NPS > MAX_NPS_SCORE) entry.NPS = MAX_NPS_SCORE;\n\n                lastNPS = entry.NPS;\n            }\n        }\n    }\n}\nreturn record;",
+      "useCase": "Model progressive satisfaction improvements based on feature adoption",
+      "complexity": "intermediate"
+    },
+    {
+      "id": "mrr-growth-multi-product",
+      "prompt": "Companies with 3 or more products experience 10% MRR growth every month. 20% of companies churn and their MRR drops to zero after a random month",
+      "hookTypes": ["scd-pre"],
+      "patterns": ["revenue growth", "churn modeling", "multi-product correlation", "SCD manipulation"],
+      "response": "// Constants - Edit these to customize MRR trends\nconst PRODUCT_THRESHOLD = 3;\nconst MONTHLY_GROWTH_RATE = 0.10; // 10% growth\nconst CHURN_LIKELIHOOD = 20;\n\nif (type === \"scd-pre\") {\n    const { type: scdType, profile, scd } = meta;\n    const prop = Object.keys(scd)[0];\n\n    // Only apply to group MRR property\n    if (scdType !== \"user\" && prop === \"MRR\") {\n        // Multi-product companies grow MRR\n        if (profile.products && profile.products.length > PRODUCT_THRESHOLD) {\n            let lastMRR = record[0].MRR;\n\n            for (const [index, entry] of record.entries()) {\n                if (index === 0) continue; // Skip first entry\n\n                // Apply monthly growth\n                entry.MRR = Math.floor(lastMRR * (1 + MONTHLY_GROWTH_RATE));\n                lastMRR = entry.MRR;\n            }\n        }\n\n        // Some companies churn\n        if (chance.bool({ likelihood: CHURN_LIKELIHOOD })) {\n            const size = record.length;\n            const churnIndex = u.integer(1, size - 1);\n\n            // Set MRR to 0 after churn point\n            for (const [index, entry] of record.entries()) {\n                if (index > churnIndex) {\n                    entry.MRR = 0;\n                }\n            }\n        }\n    }\n}\nreturn record;",
+      "useCase": "Model B2B revenue growth and churn patterns",
+      "complexity": "advanced"
+    },
+    {
+      "id": "csm-performance-impact",
+      "prompt": "Accounts managed by CSM named AK have 50% higher MRR than other accounts",
+      "hookTypes": ["scd-pre"],
+      "patterns": ["CSM performance", "account segmentation", "SCD manipulation"],
+      "response": "// Constants - Edit these to customize CSM impact\nconst HIGH_PERFORMING_CSM = \"AK\";\nconst MRR_BOOST_FACTOR = 1.5; // 50% higher\n\nif (type === \"scd-pre\") {\n    const { type: scdType, profile, scd } = meta;\n    const prop = Object.keys(scd)[0];\n\n    // Only apply to group MRR property\n    if (scdType !== \"user\" && prop === \"MRR\") {\n        // Check if account is managed by high-performing CSM\n        if (profile.CSM === HIGH_PERFORMING_CSM) {\n            for (const entry of record) {\n                entry.MRR = Math.floor(entry.MRR * MRR_BOOST_FACTOR);\n            }\n        }\n    }\n}\nreturn record;",
+      "useCase": "Model CSM performance impact on account metrics",
+      "complexity": "simple"
+    },
+    {
+      "id": "experiment-event-ordering",
+      "prompt": "Ensure experiment start events always appear first in funnel sequences to maintain proper attribution",
+      "hookTypes": ["funnel-pre"],
+      "patterns": ["experiment handling", "event ordering", "funnel manipulation"],
+      "response": "// This hook ensures experiment events are always first\nconst EXPERIMENT_EVENT_NAME = \"$experiment_started\";\n\nif (type === \"funnel-pre\") {\n    const sequence = u.clone(record.sequence);\n\n    // Check if experiment event exists in sequence\n    if (sequence.includes(EXPERIMENT_EVENT_NAME)) {\n        // Remove from current position\n        const newSequence = sequence.filter((event) => event !== EXPERIMENT_EVENT_NAME);\n\n        // Add to beginning\n        newSequence.unshift(EXPERIMENT_EVENT_NAME);\n\n        record.sequence = newSequence;\n    }\n}\nreturn record;",
+      "useCase": "Maintain clean experiment attribution in funnel analysis",
+      "complexity": "simple"
+    },
+    {
+      "id": "progressive-conversion-improvement",
+      "prompt": "Conversion rates were 50-90% of normal before the improvement 30 days ago. After the improvement, conversion rates progressively increased by 2% per day for 30 days, reaching up to 200% of baseline",
+      "hookTypes": ["funnel-pre"],
+      "patterns": ["time-based", "progressive improvement", "conversion engineering"],
+      "response": "// Constants - Edit these to customize conversion trends\nconst DAYS_AGO_IMPROVEMENT = 30;\nconst BEFORE_MIN_FACTOR = 0.5; // 50% of normal\nconst BEFORE_MAX_FACTOR = 0.9; // 90% of normal\nconst IMPROVEMENT_DAYS = 30;\nconst MAX_IMPROVEMENT_FACTOR = 2.0; // 200% of normal\n\nif (type === \"funnel-pre\") {\n    const NOW = dayjs();\n    const IMPROVEMENT_DATE = NOW.subtract(DAYS_AGO_IMPROVEMENT, 'day');\n    const parsedFirstEventTime = dayjs.unix(meta.firstEventTime);\n\n    // Adjust for time offset\n    const actualFunnelTime = parsedFirstEventTime.add(\n        NOW.diff(dayjs.unix(global.FIXED_NOW), 'h'),\n        'h'\n    );\n\n    // Before improvement: Worse conversion\n    if (actualFunnelTime.isBefore(IMPROVEMENT_DATE)) {\n        record.conversionRate *= decimal(BEFORE_MIN_FACTOR, BEFORE_MAX_FACTOR, 3);\n    }\n\n    // After improvement: Progressive improvement\n    if (actualFunnelTime.isAfter(IMPROVEMENT_DATE)) {\n        // Calculate days since improvement (capped at improvement period)\n        const daysSinceImprovement = Math.min(\n            IMPROVEMENT_DAYS,\n            actualFunnelTime.diff(IMPROVEMENT_DATE, 'day')\n        );\n\n        // Calculate improvement factor (0 to 100% over period)\n        const improvementPercent = daysSinceImprovement / IMPROVEMENT_DAYS;\n        const improvementFactor = 1.0 + (improvementPercent * (MAX_IMPROVEMENT_FACTOR - 1.0));\n\n        record.conversionRate *= decimal(1.0, MAX_IMPROVEMENT_FACTOR, 4) * improvementFactor;\n    }\n}\nreturn record;",
+      "useCase": "Model realistic product improvement curves in funnel performance",
+      "complexity": "advanced"
+    },
+    {
+      "id": "multi-type-power-user-behavior",
+      "prompt": "Power users (who do more than 10 actions per week) have 2x higher event properties, 40% better funnel conversion rates, and never experience errors. Regular users have normal metrics.",
+      "hookTypes": ["event", "funnel-pre", "everything"],
+      "patterns": ["user segmentation", "multi-type composition", "conversion engineering", "property correlation", "error suppression"],
+      "response": "// Constants - Edit these to customize power user behavior\nconst POWER_USER_THRESHOLD = 10; // events per week\nconst POWER_USER_PROPERTY_MULTIPLIER = 2.0;\nconst POWER_USER_CONVERSION_BOOST = 40; // percentage points\nconst REGULAR_USER_ERROR_RATE = 5; // percentage\n\n// Track power users across hook types\nlet isPowerUser = false;\n\n// STEP 1: Identify power users from their event volume\nif (type === \"everything\") {\n    const { profile, distinct_id } = meta;\n    \n    // Count events in the last 7 days\n    const sevenDaysAgo = dayjs().subtract(7, 'days');\n    const recentEvents = record.filter(event => \n        dayjs(event.time).isAfter(sevenDaysAgo)\n    );\n    \n    // Mark as power user if above threshold\n    if (recentEvents.length > POWER_USER_THRESHOLD) {\n        isPowerUser = true;\n        \n        // Remove all error events from power users\n        record = record.filter(event => \n            event.event !== \"error\" && \n            event.event !== \"timeout\" &&\n            event.event !== \"crash\"\n        );\n    } else {\n        // Regular users occasionally have errors\n        if (chance.bool({ likelihood: REGULAR_USER_ERROR_RATE })) {\n            const errorEvent = {\n                event: \"error\",\n                time: chance.pickone(record).time,\n                error_type: \"timeout\",\n                distinct_id: distinct_id\n            };\n            record.push(errorEvent);\n        }\n    }\n}\n\n// STEP 2: Boost numeric properties for power users\nif (type === \"event\" && isPowerUser) {\n    // Double all numeric properties for power users\n    Object.keys(record).forEach(key => {\n        if (typeof record[key] === 'number' && key !== 'time') {\n            record[key] = Math.round(record[key] * POWER_USER_PROPERTY_MULTIPLIER);\n        }\n    });\n}\n\n// STEP 3: Improve funnel conversion for power users\nif (type === \"funnel-pre\" && isPowerUser) {\n    // Increase conversion rate\n    if (record.conversionRate) {\n        const currentRate = record.conversionRate * 100;\n        const boostedRate = Math.min(95, currentRate + POWER_USER_CONVERSION_BOOST);\n        record.conversionRate = boostedRate / 100;\n    }\n    \n    // Reduce time between funnel steps\n    if (record.timing === \"random\") {\n        record.timing = \"tight\"; // Faster progression\n    }\n}\n\nreturn record;",
+      "useCase": "Demonstrate composing multiple hook types to create sophisticated user segmentation with cross-cutting concerns",
+      "complexity": "advanced"
+    }
+  ],
+  "metadata": {
+    "version": "1.0.0",
+    "totalExamples": 17,
+    "hookTypesCovered": [
+      "event",
+      "everything",
+      "funnel-pre",
+      "funnel-post",
+      "scd-pre"
+    ],
+    "patternsCovered": [
+      "time-based",
+      "A/B testing",
+      "conversion engineering",
+      "property correlation",
+      "user segmentation",
+      "churn modeling",
+      "revenue engineering",
+      "infrastructure issues",
+      "product launch",
+      "marketing attribution",
+      "AI impact",
+      "role-based access",
+      "progressive improvement",
+      "CSM performance",
+      "experiment handling",
+      "multi-type composition",
+      "error suppression"
+    ]
+  }
+}