make-mp-data 3.0.4 → 3.0.5

package/dungeons/food.js

@@ -10,63 +10,142 @@ const chance = u.initChance(SEED);
 const num_users = 5_000;
 const days = 100;
 
-/** @typedef  {import("../../types.js").Dungeon} Config */
+/** @typedef  {import("../types.d.ts").Dungeon} Config */
 
-/**
- * NEEDLE IN A HAYSTACK - FOOD DELIVERY APP DESIGN
+/*
+ * ═══════════════════════════════════════════════════════════════════════════════
+ * DATASET OVERVIEW
+ * ═══════════════════════════════════════════════════════════════════════════════
+ *
+ * QuickBite — a food delivery platform (DoorDash/Uber Eats style).
+ * Users browse restaurants, build carts, place orders, track deliveries,
+ * and rate their experiences.
+ *
+ * Scale: 5,000 users · 600K events · 100 days · 17 event types
+ *
+ * Core loop:
+ *   sign up → browse/search restaurants → add items to cart →
+ *   checkout → order placed → track delivery → rate → reorder
+ *
+ * Restaurant ecosystem: 200 restaurants across 8 cuisine types,
+ * four price tiers ($–$$$$), modeled as group profiles.
+ *
+ * Monetization: delivery fees, QuickBite+ subscription ($9.99/mo or
+ * $79.99/yr for free delivery), and promotional coupons.
+ *
+ * Support & retention: support tickets (missing items, wrong orders,
+ * late delivery, quality, refunds) and reorder events model service
+ * quality and repeat behavior.
+ *
+ * Subscription tiers: Free vs QuickBite+ create a natural A/B
+ * comparison for monetization and retention analysis.
+ */
+
+/*
+ * ═══════════════════════════════════════════════════════════════════════════════
+ * ANALYTICS HOOKS (8 architected patterns)
+ * ═══════════════════════════════════════════════════════════════════════════════
  *
- * QuickBite - A food delivery platform connecting hungry customers with local restaurants.
- * Think DoorDash/Uber Eats: users browse restaurants, build carts, place orders, track
- * deliveries in real-time, and rate their experiences afterward.
+ * 1. LUNCH RUSH CONVERSION (funnel-pre)
+ *    Funnel conversion boosted during meal hours: lunch 1.4x, dinner 1.2x.
+ *    Mixpanel:
+ *      • Funnels → "checkout started" → "order placed" → "order delivered"
+ *        Breakdown: "lunch_rush" → expect ~84% vs ~60% baseline
+ *      • Same funnel, breakdown: "dinner_rush" → expect ~72% vs ~60%
  *
- * CORE USER LOOP:
- * Users sign up (email, Google, Apple, Facebook) and immediately enter a discovery flow.
- * They browse restaurants by cuisine, sort by distance/rating/price, search for specific
- * dishes, and eventually land on a restaurant page. From there, they build a cart by adding
- * items (entrees, appetizers, drinks, desserts, sides) with customizations. Checkout
- * captures payment, tip, and delivery details. After ordering, users track their delivery
- * through multiple statuses (confirmed, preparing, picked up, en route, delivered).
+ * 2. COUPON INJECTION (funnel-post)
+ *    Free-tier users get coupon_applied spliced into funnels 30% of the time.
+ *    Mixpanel:
+ *      • Insights → "coupon applied" → breakdown "coupon_injected"
+ *        → ~30% of Free-tier coupons are injected
+ *      • Insights → "coupon applied" → breakdown "subscription_tier"
+ *        → Free users have more coupon events
  *
- * RESTAURANT ECOSYSTEM:
- * 200 restaurants across cuisine types: American, Italian, Chinese, Japanese, Mexican,
- * Indian, Thai, and Mediterranean. Restaurants span four price tiers ($, $$, $$$, $$$$)
- * with varying delivery times and ratings. This models a realistic marketplace with
- * restaurant-level analytics via group profiles.
+ * 3. LATE NIGHT MUNCHIES (event)
+ *    10PM–2AM: 70% American cuisine, 1.3x item prices, late_night_order=true.
+ *    Mixpanel:
+ *      • Insights → "restaurant viewed" → breakdown "cuisine_type"
+ *        filter: late_night_order=true → ~70% American vs ~12% daytime
+ *      • Insights → "item added to cart" → avg "item_price"
+ *        breakdown: "late_night_order" → 1.3x higher at night
+ *
+ * 4. RAINY WEEK SURGE (event + everything)
+ *    Days 20–27: delivery fees 2x, surge_pricing=true, 40% order duplication.
+ *    Mixpanel:
+ *      • Insights (line) → "order placed" → daily → visible spike days 20–27
+ *      • Insights → "order placed" → avg "delivery_fee"
+ *        breakdown: "surge_pricing" → 2x higher
  *
- * DISCOVERY & SEARCH:
- * Two paths to finding food: browsing (filtering by cuisine, sorting by rating/distance/price)
- * and searching (by restaurant name, cuisine type, or specific dish). The search-to-order
- * funnel captures intent-driven behavior vs. casual browsing.
+ * 5. REFERRAL POWER USERS (everything)
+ *    Referred users: 2x reorders, food_rating 4–5 stars, referral_user=true.
+ *    Mixpanel:
+ *      • Insights → "reorder initiated" → per user
+ *        breakdown: "referral_user" → ~2x more reorders
+ *      • Insights → "order rated" → avg "food_rating"
+ *        breakdown: "referral_user" → 4–5 vs ~3.5 baseline
+ *
+ * 6. TRIAL CONVERSION (everything)
+ *    Trial subs with 3+ orders in 14 days retained; others lose 60% of events.
+ *    Mixpanel:
+ *      • Insights → any event → per user
+ *        breakdown: "trial_retained" → sustained vs ~60% drop
+ *      • Retention → "subscription started" (trial=true) → "order placed"
+ *        → sharp drop after day 14 for non-retained
+ *
+ * 7. SUPPORT TICKET CHURN (user)
+ *    15% of users flagged is_high_risk=true with churn_risk_score 70–100.
+ *    Mixpanel:
+ *      • Insights → any event → unique users
+ *        breakdown: user "is_high_risk" → ~15% high-risk
+ *      • Insights → breakdown: user "churn_risk_score"
+ *        → bimodal: 85% at 0–40, 15% at 70–100
  *
- * MONETIZATION MODEL:
- * - Delivery fees ($0-$12, waived for QuickBite+ subscribers)
- * - QuickBite+ subscription ($9.99/month or $79.99/year) for free delivery and perks
- * - Promotions and coupons drive trial and reactivation
- * - Restaurant-promoted listings (not modeled as user events)
+ * 8. FIRST ORDER BONUS (funnel-pre)
+ *    New users get 1.4x conversion boost on checkout→order funnel.
+ *    Mixpanel:
+ *      • Funnels → "checkout started" → "order placed" → "order delivered"
+ *        breakdown: "first_order_bonus" → 1.4x higher conversion
+ *      • Insights → "order placed" → breakdown "first_order_bonus"
+ *        → ~50% tagged (hash-based user split)
  *
- * CART BEHAVIOR:
- * Cart events (add/remove) capture the deliberation process. Removal reasons
- * (changed_mind, too_expensive, substitution) reveal price sensitivity and UX friction.
- * Customization counts show engagement depth with individual items.
+ * ───────────────────────────────────────────────────────────────────────────────
+ * ADVANCED ANALYSIS IDEAS
+ * ───────────────────────────────────────────────────────────────────────────────
  *
- * ORDER LIFECYCLE:
- * order placed -> order tracked (multiple status updates) -> order delivered -> order rated
- * This models the full post-purchase experience. On-time delivery, actual vs. estimated
- * delivery time, and food/delivery ratings capture service quality metrics.
+ * Cross-hook patterns:
+ *   • The Perfect Customer: referral (5) + trial retained (6) + lunch rush (1)
+ *     + low churn risk (7) → exceptional LTV and retention
+ *   • Rainy Night Double Whammy: late-night (3) + rainy week (4)
+ *     → compounded surge pricing?
+ *   • Coupon-Driven Trial: injected coupons (2) → trial starts (6)?
+ *   • Referral + First Order: hooks 5 + 8 → highest conversion rates
+ *   • Support and Churn: high-risk (7) + support tickets during rainy week (4)
+ *
+ * Cohort analysis:
+ *   • Signup week (rainy week cohort behaves differently)
+ *   • Referral vs organic lifecycle
+ *   • Free vs QuickBite+ across all metrics
+ *   • City-level cuisine and ordering patterns
+ *
+ * Funnel analysis:
+ *   • Onboarding: account created → first restaurant view, by signup method
+ *   • Order: checkout → delivery, by platform / tier / time of day
+ *   • Discovery: search type → conversion to ordering
  *
- * SUPPORT & RETENTION:
- * Support tickets (missing items, wrong orders, late delivery, quality issues, refund
- * requests) model service failures. Reorder events capture repeat behavior and loyalty.
- * The ratio of support tickets to orders is a key service quality indicator.
+ * ───────────────────────────────────────────────────────────────────────────────
+ * EXPECTED METRICS SUMMARY
+ * ───────────────────────────────────────────────────────────────────────────────
  *
- * WHY THESE EVENTS/PROPERTIES?
- * - Events model the complete food delivery loop: discovery -> consideration -> purchase -> fulfillment -> retention
- * - Properties enable cohort analysis: cuisine preferences, price sensitivity, platform usage, subscription status
- * - Funnels reveal friction: where do users drop off between browsing and ordering?
- * - Time-based patterns (lunch rush, late night) simulate real delivery demand curves
- * - Subscription tier (Free vs QuickBite+) creates a natural A/B comparison for monetization analysis
- * - Support and rating events drive churn prediction and service quality monitoring
- * - The "needle in haystack" hooks simulate real product insights hidden in production data
+ * Hook                  | Metric               | Baseline | Hook Effect | Ratio
+ * ──────────────────────|──────────────────────|──────────|─────────────|──────
+ * Lunch Rush            | Funnel conversion    | 60%      | 84%         | 1.4x
+ * Coupon Injection      | Free user coupons    | 0%       | 30%         | N/A
+ * Late Night Munchies   | American cuisine %   | ~15%     | 70%         | ~4.7x
+ * Rainy Week Surge      | Order volume         | 100%     | 140%        | 1.4x
+ * Referral Power Users  | Reorder frequency    | 1x       | 2x          | 2.0x
+ * Trial Conversion      | Post-trial retention | 100%     | 40%         | 0.4x
+ * Support Ticket Churn  | High-risk users      | 0%       | 15%         | N/A
+ * First Order Bonus     | New user conversion  | 1x       | 1.4x        | 1.4x
  */
 
 // Generate consistent IDs for lookup tables and event properties
@@ -415,7 +494,7 @@ const config = {
 		// ═══════════════════════════════════════════════════════════════════
 		if (type === "funnel-pre") {
 			if (meta && meta.firstEventTime) {
-				const hour = dayjs(meta.firstEventTime).hour();
+				const hour = dayjs.unix(meta.firstEventTime).hour();
 
 				record.props = record.props || {};
 
@@ -450,7 +529,7 @@ const config = {
 				const userId = meta && meta.user && (meta.user.distinct_id || String(meta.user));
 				const isNewUser = userId && userId.charCodeAt(0) % 2 === 0;
 				if (isNewUser) {
-					record.conversionRate = 0.90;
+					record.conversionRate = Math.min(98, record.conversionRate * 1.4);
 					record.props.first_order_bonus = true;
 				} else {
 					record.props.first_order_bonus = false;
@@ -484,6 +563,9 @@ const config = {
 						event: "coupon applied",
 						time: midTime,
 						user_id: firstEvent.user_id,
+						subscription_tier: firstEvent.subscription_tier,
+						platform: firstEvent.platform,
+						city: firstEvent.city,
 						coupon_code: chance.pickone(couponCodes),
 						discount_type: chance.pickone(["percent", "flat", "free_delivery"]),
 						discount_value: chance.integer({ min: 10, max: 30 }),
@@ -535,23 +617,7 @@ const config = {
 				if (EVENT_TIME.isAfter(RAINY_WEEK_START) && EVENT_TIME.isBefore(RAINY_WEEK_END)) {
 					record.delivery_fee = (record.delivery_fee || 5) * 2;
 					record.surge_pricing = true;
-
-					// 40% chance to duplicate the event (more orders during rain)
-					if (chance.bool({ likelihood: 40 })) {
-						const duplicateEvent = {
-							event: "order placed",
-							time: EVENT_TIME.add(chance.integer({ min: 5, max: 60 }), 'minutes').toISOString(),
-							user_id: record.user_id,
-							order_id: chance.pickone(orderIds),
-							payment_method: chance.pickone(["credit_card", "apple_pay", "google_pay"]),
-							order_total: chance.integer({ min: 15, max: 120 }),
-							tip_amount: chance.integer({ min: 0, max: 20 }),
-							delivery_fee: chance.integer({ min: 6, max: 24 }),
-							surge_pricing: true,
-							rainy_week: true,
-						};
-						return [record, duplicateEvent];
-					}
+					record.rainy_week = true;
 				} else {
 					record.surge_pricing = false;
 				}
@@ -645,6 +711,24 @@ const config = {
 					}
 				}
 			}
+
+			// ═══════════════════════════════════════════════════════════════
+			// HOOK 4: RAINY WEEK SURGE - duplicate order events (everything)
+			// Events tagged surge_pricing=true in the event hook get a 40%
+			// chance of duplication here, creating visible demand spikes.
+			// ═══════════════════════════════════════════════════════════════
+			const rainyDuplicates = [];
+			userEvents.forEach((event) => {
+				if (event.surge_pricing === true && event.event === "order placed" && chance.bool({ likelihood: 40 })) {
+					const dup = JSON.parse(JSON.stringify(event));
+					dup.time = dayjs(event.time).add(chance.integer({ min: 5, max: 60 }), 'minutes').toISOString();
+					dup.rainy_week = true;
+					rainyDuplicates.push(dup);
+				}
+			});
+			if (rainyDuplicates.length > 0) {
+				userEvents.push(...rainyDuplicates);
+			}
 		}
 
 		// ═══════════════════════════════════════════════════════════════════
@@ -668,320 +752,3 @@ const config = {
 };
 
 export default config;
-
-/**
- * ═══════════════════════════════════════════════════════════════════════════════
- * NEEDLE IN A HAYSTACK - QUICKBITE FOOD DELIVERY ANALYTICS
- * ═══════════════════════════════════════════════════════════════════════════════
- *
- * A food delivery app dungeon with 8 deliberately architected analytics
- * insights hidden in the data. This dungeon is designed to showcase advanced
- * product analytics patterns and demonstrate how to find "needles" (meaningful
- * insights) in "haystacks" (large datasets).
- *
- * ═══════════════════════════════════════════════════════════════════════════════
- * DATASET OVERVIEW
- * ═══════════════════════════════════════════════════════════════════════════════
- *
- * - 5,000 users over 100 days
- * - 360K events across 17 event types
- * - 3 funnels (onboarding, order completion, discovery to order)
- * - Group analytics (restaurants)
- * - Lookup tables (menu items, coupon codes)
- * - Subscription tiers (Free, QuickBite+)
- *
- * ═══════════════════════════════════════════════════════════════════════════════
- * THE 8 ARCHITECTED HOOKS
- * ═══════════════════════════════════════════════════════════════════════════════
- *
- * Each hook creates a specific, discoverable analytics insight that simulates
- * real-world product behavior patterns.
- *
- * ───────────────────────────────────────────────────────────────────────────────
- * 1. LUNCH RUSH CONVERSION (funnel-pre)
- * ───────────────────────────────────────────────────────────────────────────────
- *
- * PATTERN: Funnel conversion rates are boosted during meal-time hours.
- * Lunch (11AM-1PM) gets a 1.4x multiplier; dinner (5PM-8PM) gets 1.2x.
- * Events during these windows carry lunch_rush or dinner_rush properties.
- *
- * HOW TO FIND IT:
- *   - Break down funnel conversion by hour of day
- *   - Compare conversion rates during 11AM-1PM vs. off-peak hours
- *   - Filter events where lunch_rush = true or dinner_rush = true
- *
- * EXPECTED INSIGHT: Lunch hour funnels convert ~40% better than baseline.
- * Dinner hour funnels convert ~20% better. Clear time-of-day pattern in
- * the checkout-to-order-to-delivery pipeline.
- *
- * REAL-WORLD ANALOGUE: Food delivery apps see massive demand spikes during
- * traditional meal times. Users ordering during these windows have higher
- * intent and thus higher conversion rates.
- *
- * ───────────────────────────────────────────────────────────────────────────────
- * 2. COUPON INJECTION (funnel-post)
- * ───────────────────────────────────────────────────────────────────────────────
- *
- * PATTERN: Free-tier users get coupon_applied events spliced into their
- * funnel sequences 30% of the time. These injected coupons carry the
- * coupon_injected: true property to distinguish them from organic usage.
- *
- * HOW TO FIND IT:
- *   - Segment by subscription_tier = "Free"
- *   - Look for coupon_applied events with coupon_injected = true
- *   - Compare funnel completion rates for Free users with/without coupons
- *
- * EXPECTED INSIGHT: ~30% of Free users receive promotional coupons mid-funnel.
- * These users should show different downstream conversion behavior, simulating
- * how promotional nudges affect the purchase funnel.
- *
- * REAL-WORLD ANALOGUE: Apps frequently inject promotional offers (push
- * notifications, in-app banners) to non-paying users during key funnel
- * moments to boost conversion.
- *
- * ───────────────────────────────────────────────────────────────────────────────
- * 3. LATE NIGHT MUNCHIES (event)
- * ───────────────────────────────────────────────────────────────────────────────
- *
- * PATTERN: Between 10PM and 2AM, restaurant views and cart additions skew
- * 70% toward American (fast food) cuisine. Item prices are boosted by 1.3x.
- * Events carry late_night_order: true.
- *
- * HOW TO FIND IT:
- *   - Break down restaurant_viewed by hour of day and cuisine_type
- *   - Compare cuisine distribution at 10PM-2AM vs. daytime
- *   - Filter late_night_order = true and compare average item_price
- *
- * EXPECTED INSIGHT: Late-night orders are overwhelmingly fast food with
- * higher average prices. The cuisine distribution at night is dramatically
- * different from daytime patterns.
- *
- * REAL-WORLD ANALOGUE: Late-night food delivery is dominated by fast food
- * and comfort food. Many platforms charge late-night surcharges or see
- * naturally inflated basket sizes during these hours.
- *
- * ───────────────────────────────────────────────────────────────────────────────
- * 4. RAINY WEEK SURGE (event)
- * ───────────────────────────────────────────────────────────────────────────────
- *
- * PATTERN: During days 20-27 of the dataset, a simulated rainy week causes:
- *   - Delivery fees double on order_placed events
- *   - surge_pricing: true flag added
- *   - 40% chance of duplicate order events (demand surge)
- *
- * HOW TO FIND IT:
- *   - Chart order_placed event count by day
- *   - Filter days 20-27 and compare delivery_fee averages
- *   - Look for surge_pricing = true and rainy_week = true properties
- *
- * EXPECTED INSIGHT: Clear spike in order volume around days 20-27, with
- * doubled delivery fees and surge pricing markers. The demand surge is
- * visible as a ~40% increase in order_placed event volume.
- *
- * REAL-WORLD ANALOGUE: Weather events drive significant demand surges for
- * food delivery. Platforms respond with surge pricing on delivery fees,
- * and order volume increases as people avoid going out.
- *
- * ───────────────────────────────────────────────────────────────────────────────
- * 5. REFERRAL POWER USERS (everything)
- * ───────────────────────────────────────────────────────────────────────────────
- *
- * PATTERN: Users whose account_created event has referral_code = true:
- *   - Get 2x more reorder_initiated events (higher loyalty)
- *   - Have food_rating boosted to 4-5 (higher satisfaction)
- *   - Carry referral_user: true on affected events
- *
- * HOW TO FIND IT:
- *   - Segment users by: referral_code = true on account_created
- *   - Compare: reorder_initiated event count per user
- *   - Compare: average food_rating on order_rated events
- *
- * EXPECTED INSIGHT: Referred users reorder roughly 2x more often and rate
- * food 4-5 stars consistently. They represent a higher-LTV, more-satisfied
- * segment of the user base.
- *
- * REAL-WORLD ANALOGUE: Referral programs consistently produce higher-quality
- * users. Referred customers have lower acquisition costs, higher retention,
- * and higher satisfaction because they come with built-in social proof.
- *
- * ───────────────────────────────────────────────────────────────────────────────
- * 6. TRIAL CONVERSION (everything)
- * ───────────────────────────────────────────────────────────────────────────────
- *
- * PATTERN: Users with subscription_started where trial = true are evaluated:
- *   - If they place 3+ orders in their first 14 days: retained (all events kept)
- *   - If fewer than 3 orders: churned (60% of events after day 14 removed)
- *   - Retained users carry trial_retained: true
- *
- * HOW TO FIND IT:
- *   - Segment users by: subscription_started with trial = true
- *   - Count order_placed events within first 14 days per user
- *   - Compare: event volume after day 14 for 3+ orders vs. <3 orders
- *   - Look for trial_retained = true property
- *
- * EXPECTED INSIGHT: Trial users who place 3+ early orders show sustained
- * engagement. Those who don't show a dramatic drop-off after day 14, with
- * 60% fewer events. The "magic number" for trial conversion is 3 orders.
- *
- * REAL-WORLD ANALOGUE: Subscription services often find a "magic number"
- * of early actions that predict long-term retention. For food delivery,
- * this is typically a threshold of orders during the trial period.
- *
- * ───────────────────────────────────────────────────────────────────────────────
- * 7. SUPPORT TICKET CHURN (user)
- * ───────────────────────────────────────────────────────────────────────────────
- *
- * PATTERN: 15% of users are flagged as is_high_risk = true with a
- * churn_risk_score between 70-100. The remaining 85% get scores of 0-40.
- * This creates a binary segmentation opportunity for retention teams.
- *
- * HOW TO FIND IT:
- *   - Segment users by: is_high_risk = true
- *   - Compare: churn_risk_score distribution
- *   - Cross-reference: support_ticket event frequency for high-risk users
- *   - Compare: retention and ordering behavior by risk segment
- *
- * EXPECTED INSIGHT: 15% of users have churn scores above 70, creating a
- * clear at-risk segment. These users likely correlate with higher support
- * ticket rates and lower order frequency.
- *
- * REAL-WORLD ANALOGUE: Customer health scoring is used by subscription
- * businesses to identify at-risk users for proactive retention outreach.
- * ML models in production generate similar risk scores from behavioral data.
- *
- * ───────────────────────────────────────────────────────────────────────────────
- * 8. FIRST ORDER BONUS (funnel-pre)
- * ───────────────────────────────────────────────────────────────────────────────
- *
- * PATTERN: For the Order Completion funnel (checkout started -> order placed
- * -> order delivered), new users (born in dataset) get their conversion rate
- * boosted to 90%. Events carry first_order_bonus: true.
- *
- * HOW TO FIND IT:
- *   - Segment the Order Completion funnel by new vs. existing users
- *   - Compare: conversion rates for new users vs. returning users
- *   - Filter: first_order_bonus = true
- *
- * EXPECTED INSIGHT: New users convert at ~90% through the order funnel,
- * dramatically higher than the 60% baseline. This simulates first-order
- * promotions (free delivery, $10 off) that most food delivery apps offer.
- *
- * REAL-WORLD ANALOGUE: Every major food delivery app offers aggressive
- * first-order incentives. These dramatically boost initial conversion but
- * the real question is whether those users return (see Hook #6).
- *
- * ═══════════════════════════════════════════════════════════════════════════════
- * ADVANCED ANALYSIS IDEAS
- * ═══════════════════════════════════════════════════════════════════════════════
- *
- * CROSS-HOOK PATTERNS:
- *
- * 1. The Perfect Customer: Users who:
- *    - Signed up via referral (Hook #5)
- *    - Started a trial and placed 3+ early orders (Hook #6)
- *    - Order during lunch rush (Hook #1)
- *    - Have low churn risk score (Hook #7)
- *    These users should have exceptional LTV and retention metrics.
- *
- * 2. Rainy Night Double Whammy: Do late-night orders during rainy week
- *    (Hook #3 + Hook #4) show compounded surge pricing effects?
- *
- * 3. Coupon-Driven Trial Conversion: Do free users who receive injected
- *    coupons (Hook #2) eventually start trials (Hook #6)?
- *
- * 4. Referral + First Order: Referred users (Hook #5) who also get the
- *    first order bonus (Hook #8) should have the highest conversion rates.
- *
- * 5. Support Tickets and Churn Risk: Do high-risk users (Hook #7) file
- *    more support tickets, and does this correlate with rainy week
- *    late deliveries (Hook #4)?
- *
- * COHORT ANALYSIS:
- *
- * - Cohort by signup week: Users who started during rainy week (days 20-27)
- *   should show different ordering patterns
- * - Cohort by referral source: Referred vs. organic user lifecycle comparison
- * - Cohort by subscription tier: Free vs. QuickBite+ across all metrics
- * - Cohort by city: Do different cities show different cuisine preferences
- *   and ordering patterns?
- *
- * FUNNEL ANALYSIS:
- *
- * - Onboarding Funnel: How quickly do users go from account creation to
- *   first restaurant view? Break down by signup method.
- * - Order Funnel: Compare checkout-to-delivery completion by platform,
- *   subscription tier, and time of day.
- * - Discovery Funnel: How does search type (restaurant vs. cuisine vs. dish)
- *   affect downstream conversion to ordering?
- *
- * ═══════════════════════════════════════════════════════════════════════════════
- * EXPECTED METRICS SUMMARY
- * ═══════════════════════════════════════════════════════════════════════════════
- *
- * Hook                  | Metric               | Baseline | Hook Effect | Ratio
- * ──────────────────────|──────────────────────|──────────|─────────────|──────
- * Lunch Rush            | Funnel conversion    | 60%      | 84%         | 1.4x
- * Coupon Injection      | Free user coupons    | 0%       | 30%         | N/A
- * Late Night Munchies   | American cuisine %   | ~15%     | 70%         | ~4.7x
- * Rainy Week Surge      | Order volume         | 100%     | 140%        | 1.4x
- * Referral Power Users  | Reorder frequency    | 1x       | 2x          | 2.0x
- * Trial Conversion      | Post-trial retention | 100%     | 40%         | 0.4x
- * Support Ticket Churn  | High-risk users      | 0%       | 15%         | N/A
- * First Order Bonus     | New user conversion  | 60%      | 90%         | 1.5x
- *
- * ═══════════════════════════════════════════════════════════════════════════════
- * HOW TO RUN THIS DUNGEON
- * ═══════════════════════════════════════════════════════════════════════════════
- *
- * From the dm4 root directory:
- *
- *   npm start
- *
- * Or programmatically:
- *
- *   import generate from './index.js';
- *   import config from './dungeons/harness-food.js';
- *   const results = await generate(config);
- *
- * OUTPUT FILES (with writeToDisk: false, format: "json", gzip: true):
- *
- *   - needle-haystack-food__events.json.gz - All event data
- *   - needle-haystack-food__user_profiles.json.gz - User profiles
- *   - needle-haystack-food__group_profiles.json.gz - Restaurant profiles
- *   - needle-haystack-food__item_id_lookup.json.gz - Menu item catalog
- *   - needle-haystack-food__coupon_code_lookup.json.gz - Coupon catalog
- *
- * ═══════════════════════════════════════════════════════════════════════════════
- * TESTING YOUR ANALYTICS PLATFORM
- * ═══════════════════════════════════════════════════════════════════════════════
- *
- * This dungeon is perfect for testing:
- *
- * 1. Time-of-Day Analysis: Can you detect the lunch and dinner rush patterns?
- * 2. Promotional Impact: Can you measure the effect of injected coupons?
- * 3. Behavioral Shifts: Can you discover the late-night cuisine preferences?
- * 4. Anomaly Detection: Can you spot the rainy week demand surge?
- * 5. Referral Analysis: Can you quantify the referral user advantage?
- * 6. Trial Optimization: Can you find the "magic number" of early orders?
- * 7. Churn Prediction: Can you identify high-risk users before they leave?
- * 8. New User Funnels: Can you measure the first-order bonus impact?
- *
- * ═══════════════════════════════════════════════════════════════════════════════
- * WHY "NEEDLE IN A HAYSTACK"?
- * ═══════════════════════════════════════════════════════════════════════════════
- *
- * Each hook is a "needle" - a meaningful, actionable insight hidden in a
- * "haystack" of 360K events. The challenge is:
- *
- * 1. FINDING the needles (discovery)
- * 2. VALIDATING they're real patterns (statistical significance)
- * 3. UNDERSTANDING why they matter (business impact)
- * 4. ACTING on them (product decisions)
- *
- * This mirrors real-world product analytics: your data contains valuable insights,
- * but you need the right tools and skills to find them.
- *
- * Happy Hunting!
- *
- * ═══════════════════════════════════════════════════════════════════════════════
- */