mike boone commited on
Commit
cfcdfcd
·
1 Parent(s): 8997cc3

feat: add ad monetization dataset scenario

Browse files
demoprep_app/dataset/generators/template_generator.py CHANGED
@@ -116,6 +116,37 @@ class TemplateDatasetGenerator(ScenarioDatasetGenerator):
116
  rows.update({"APPLICATIONS": conversions, "APPROVED_ACCOUNTS": approved, "ACCOUNTS_OPENED": approved, "APPROVAL_RATE_PCT": self._pct(approved, conversions)})
117
  return rows
118
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
119
  if profile == "pipeline":
120
  amount = round(rng.uniform(12_000, 420_000) * mult, 2)
121
  probability = round(rng.uniform(8, 82), 2)
@@ -1397,6 +1428,72 @@ def _dimension_values(
1397
  "Limited Service",
1398
  ],
1399
  "PRODUCT_CATEGORIES": ["Denim", "Outerwear", "Sportswear", "Dresses", "Footwear", "Accessories", "Kids", "Underwear"],
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1400
  "STORES": [f"{company} Flagship", f"{company} Core", f"{company} Outlet", f"{company} Digital", f"{company} Urban", f"{company} Suburban"],
1401
  "WAREHOUSES": warehouse_values,
1402
  "SUPPLIERS": supplier_values,
@@ -1901,6 +1998,11 @@ def _clean_values(values: list[Any]) -> list[str]:
1901
  def _dimension_base(name: str) -> str:
1902
  singulars = {
1903
  "PRODUCT_CATEGORIES": "PRODUCT_CATEGORY",
 
 
 
 
 
1904
  "CUSTOMER_SEGMENTS": "CUSTOMER_SEGMENT",
1905
  "EQUIPMENT_TYPES": "EQUIPMENT_TYPE",
1906
  "TERRITORIES": "TERRITORY",
@@ -1923,11 +2025,11 @@ def _dimension_base(name: str) -> str:
1923
  def _dimension_target_count(name: str) -> int:
1924
  if name == "SCENARIOS":
1925
  return 3
1926
- if name in {"REGIONS", "CHANNELS", "PRIORITIES", "SHIFTS", "CABIN_CLASSES", "SEVERITIES", "TERMS"}:
1927
  return 8
1928
  if name in {"PRODUCTS", "STORES", "ACCOUNTS", "CUSTOMERS", "CAMPAIGNS", "ROUTES", "EVENTS", "CLIENTS", "DEALERS", "PROPERTIES"}:
1929
  return 36
1930
- if name in {"SALES_REPS", "SUPPLIERS", "CARRIERS", "AIRPORTS", "AIRCRAFT", "WAREHOUSES", "TEAMS", "VENUES", "FAN_SEGMENTS", "CONSULTANTS", "VEHICLE_MODELS"}:
1931
  return 24
1932
  return 18
1933
 
 
116
  rows.update({"APPLICATIONS": conversions, "APPROVED_ACCOUNTS": approved, "ACCOUNTS_OPENED": approved, "APPROVAL_RATE_PCT": self._pct(approved, conversions)})
117
  return rows
118
 
119
+ if profile == "ad_monetization":
120
+ active_users = self._whole(rng, 25_000, 1_800_000, mult)
121
+ sessions = self._whole(rng, active_users, max(active_users + 1, active_users * 8), 1.0)
122
+ watch_hours = round(sessions * rng.uniform(0.35, 2.4), 2)
123
+ ad_requests = self._whole(rng, sessions * 2, max(sessions * 3, sessions * 14), 1.0)
124
+ filled_impressions = self._bounded_child(rng, ad_requests, 0.62, 0.94)
125
+ completed_views = self._bounded_child(rng, filled_impressions, 0.58, 0.91)
126
+ ecpm = round(rng.uniform(11.0, 54.0), 2)
127
+ ad_revenue = round(filled_impressions / 1000.0 * ecpm, 2)
128
+ yield_per_hour = self._div(ad_revenue, watch_hours)
129
+ arpu = self._div(ad_revenue, active_users)
130
+ unfilled = max(ad_requests - filled_impressions, 0)
131
+ revenue_leakage = round(unfilled / 1000.0 * ecpm * rng.uniform(0.35, 0.85), 2)
132
+ return {
133
+ "ACTIVE_USERS": active_users,
134
+ "SESSIONS": sessions,
135
+ "WATCH_HOURS": watch_hours,
136
+ "AD_REQUESTS": ad_requests,
137
+ "FILLED_IMPRESSIONS": filled_impressions,
138
+ "COMPLETED_VIEWS": completed_views,
139
+ "FILL_RATE_PCT": self._pct(filled_impressions, ad_requests),
140
+ "COMPLETION_RATE_PCT": self._pct(completed_views, filled_impressions),
141
+ "ECPM_USD": ecpm,
142
+ "AD_REVENUE_USD": ad_revenue,
143
+ "ARPU_USD": arpu,
144
+ "REVENUE_PER_SESSION_USD": self._div(ad_revenue, sessions),
145
+ "AD_YIELD_PER_WATCH_HOUR_USD": yield_per_hour,
146
+ "UNFILLED_REQUESTS": unfilled,
147
+ "REVENUE_LEAKAGE_USD": revenue_leakage,
148
+ }
149
+
150
  if profile == "pipeline":
151
  amount = round(rng.uniform(12_000, 420_000) * mult, 2)
152
  probability = round(rng.uniform(8, 82), 2)
 
1428
  "Limited Service",
1429
  ],
1430
  "PRODUCT_CATEGORIES": ["Denim", "Outerwear", "Sportswear", "Dresses", "Footwear", "Accessories", "Kids", "Underwear"],
1431
+ "AD_PLACEMENTS": [
1432
+ "SmartCast Home Hero",
1433
+ "SmartCast Home Rail",
1434
+ "FAST Channel Pre-Roll",
1435
+ "FAST Channel Mid-Roll",
1436
+ "WatchFree Plus Pause Ad",
1437
+ "Search Sponsored Tile",
1438
+ "App Launch Takeover",
1439
+ "Content Detail Native",
1440
+ "Screensaver Sponsorship",
1441
+ "Recommendation Rail",
1442
+ "Voice Search Sponsored Result",
1443
+ "Live Guide Sponsored Row",
1444
+ ],
1445
+ "CONTENT_CATEGORIES": [
1446
+ "Movies",
1447
+ "Sports",
1448
+ "News",
1449
+ "Lifestyle",
1450
+ "Kids and Family",
1451
+ "Reality",
1452
+ "Documentary",
1453
+ "Comedy",
1454
+ "Drama",
1455
+ "Home and Design",
1456
+ "Food",
1457
+ "Music",
1458
+ ],
1459
+ "AUDIENCE_SEGMENTS": [
1460
+ "High Engagement Streamers",
1461
+ "FAST Channel Loyalists",
1462
+ "Sports Heavy Households",
1463
+ "Movie Night Families",
1464
+ "Cord Cutter Households",
1465
+ "New SmartCast Users",
1466
+ "Lapsed Viewers",
1467
+ "Premium App Users",
1468
+ "Value Seekers",
1469
+ "Weekend Bingers",
1470
+ "News Daily Viewers",
1471
+ "Low Ad Load Cohort",
1472
+ ],
1473
+ "DEVICE_TYPES": [
1474
+ "OLED Smart TV",
1475
+ "Quantum Pro TV",
1476
+ "M-Series TV",
1477
+ "V-Series TV",
1478
+ "Bedroom TV",
1479
+ "Living Room TV",
1480
+ "Large Format TV",
1481
+ "Legacy SmartCast TV",
1482
+ ],
1483
+ "AD_PRODUCTS": [
1484
+ "Programmatic Video",
1485
+ "Direct Sold Video",
1486
+ "Sponsored Home Screen",
1487
+ "FAST Channel Sponsorship",
1488
+ "Native Recommendation",
1489
+ "Pause Ad",
1490
+ "Interactive Overlay",
1491
+ "Shoppable TV",
1492
+ "Frequency Extension",
1493
+ "Audience Retargeting",
1494
+ "Contextual Package",
1495
+ "Seasonal Takeover",
1496
+ ],
1497
  "STORES": [f"{company} Flagship", f"{company} Core", f"{company} Outlet", f"{company} Digital", f"{company} Urban", f"{company} Suburban"],
1498
  "WAREHOUSES": warehouse_values,
1499
  "SUPPLIERS": supplier_values,
 
1998
  def _dimension_base(name: str) -> str:
1999
  singulars = {
2000
  "PRODUCT_CATEGORIES": "PRODUCT_CATEGORY",
2001
+ "AD_PLACEMENTS": "AD_PLACEMENT",
2002
+ "CONTENT_CATEGORIES": "CONTENT_CATEGORY",
2003
+ "AUDIENCE_SEGMENTS": "AUDIENCE_SEGMENT",
2004
+ "DEVICE_TYPES": "DEVICE_TYPE",
2005
+ "AD_PRODUCTS": "AD_PRODUCT",
2006
  "CUSTOMER_SEGMENTS": "CUSTOMER_SEGMENT",
2007
  "EQUIPMENT_TYPES": "EQUIPMENT_TYPE",
2008
  "TERRITORIES": "TERRITORY",
 
2025
  def _dimension_target_count(name: str) -> int:
2026
  if name == "SCENARIOS":
2027
  return 3
2028
+ if name in {"REGIONS", "CHANNELS", "PRIORITIES", "SHIFTS", "CABIN_CLASSES", "SEVERITIES", "TERMS", "DEVICE_TYPES"}:
2029
  return 8
2030
  if name in {"PRODUCTS", "STORES", "ACCOUNTS", "CUSTOMERS", "CAMPAIGNS", "ROUTES", "EVENTS", "CLIENTS", "DEALERS", "PROPERTIES"}:
2031
  return 36
2032
+ if name in {"SALES_REPS", "SUPPLIERS", "CARRIERS", "AIRPORTS", "AIRCRAFT", "WAREHOUSES", "TEAMS", "VENUES", "FAN_SEGMENTS", "CONSULTANTS", "VEHICLE_MODELS", "AD_PLACEMENTS", "CONTENT_CATEGORIES", "AUDIENCE_SEGMENTS", "AD_PRODUCTS"}:
2033
  return 24
2034
  return 18
2035
 
demoprep_app/pipeline/dataset_first.py CHANGED
@@ -134,6 +134,7 @@ def build_dataset_first_demo(
134
  "workforce_hr",
135
  "legal_matter_management",
136
  "marketing_funnel",
 
137
  "ecommerce_marketing",
138
  "finance_unit_economics",
139
  "subscription_revenue",
 
134
  "workforce_hr",
135
  "legal_matter_management",
136
  "marketing_funnel",
137
+ "ad_monetization",
138
  "ecommerce_marketing",
139
  "finance_unit_economics",
140
  "subscription_revenue",
demoprep_app/scenario/extractor.py CHANGED
@@ -224,6 +224,8 @@ def _infer_from_text(text: str) -> str | None:
224
  return "professional_services_engagements"
225
  if _has_any(low, ("arena", "venue", "stadium", "ticket", "attendance", "fan", "season ticket", "suite", "sponsorship", "basketball", "hockey", "concert", "theater", "theatre")):
226
  return "sports_venue_fan_engagement"
 
 
227
  return None
228
 
229
 
 
224
  return "professional_services_engagements"
225
  if _has_any(low, ("arena", "venue", "stadium", "ticket", "attendance", "fan", "season ticket", "suite", "sponsorship", "basketball", "hockey", "concert", "theater", "theatre")):
226
  return "sports_venue_fan_engagement"
227
+ if _has_any(low, ("ad yield", "ad monetization", "advertising monetization", "arpu", "ecpm", "fill rate", "ctv", "connected tv", "smart tv", "smartcast", "fast channel", "ad requests", "completion rate")):
228
+ return "ad_monetization"
229
  return None
230
 
231
 
demoprep_app/scenario/families.py CHANGED
@@ -68,6 +68,20 @@ SCENARIO_FAMILIES: dict[str, ScenarioFamilyTemplate] = {
68
  "digital_funnel",
69
  ("Where is the biggest funnel drop-off?", "Which campaigns have the best conversion?", "How is CPA trending?"),
70
  ),
 
 
 
 
 
 
 
 
 
 
 
 
 
 
71
  "apparel_marketing": ScenarioFamilyTemplate(
72
  "apparel_marketing",
73
  "APPAREL_MARKETING",
 
68
  "digital_funnel",
69
  ("Where is the biggest funnel drop-off?", "Which campaigns have the best conversion?", "How is CPA trending?"),
70
  ),
71
+ "ad_monetization": ScenarioFamilyTemplate(
72
+ "ad_monetization",
73
+ "AD_MONETIZATION",
74
+ "placement-content-audience-device-region-month",
75
+ "AD_REVENUE_USD",
76
+ ("AD_PLACEMENTS", "CONTENT_CATEGORIES", "AUDIENCE_SEGMENTS", "DEVICE_TYPES", "REGIONS", "AD_PRODUCTS"),
77
+ ("CTV Homescreen", "FAST Channel", "Video Pre-Roll", "Video Mid-Roll", "Native Placement", "Sponsored Content"),
78
+ "ad_monetization",
79
+ (
80
+ "Which placements, content categories, and audience cohorts drive ad yield?",
81
+ "Where are fill rate, eCPM, or completion issues suppressing revenue?",
82
+ "Which Smart TV user segments have the highest ARPU and revenue expansion opportunity?",
83
+ ),
84
+ ),
85
  "apparel_marketing": ScenarioFamilyTemplate(
86
  "apparel_marketing",
87
  "APPAREL_MARKETING",
demoprep_app/scenario/selector.py CHANGED
@@ -25,6 +25,7 @@ def infer_scenario_family(use_case: str, vertical: str | None = None, function:
25
  ("hotel_finance", ("hotel", "hotels", "hospitality", "property", "room", "rooms", "occupancy", "revpar", "adr"), ("finance", "margin", "budget", "forecast")),
26
  ("apparel_marketing", ("apparel", "fashion", "garment", "footwear", "seasonal drop", "markdown"), ("marketing", "campaign", "loyalty", "channel", "conversion")),
27
  ("sports_venue_fan_engagement", ("sports", "arena", "venue", "stadium", "ticket", "tickets", "ticketing", "attendance", "fan", "fans", "season ticket", "premium seating", "suite", "suites", "sponsorship", "sponsor", "team", "league", "basketball", "hockey", "soccer", "baseball", "football", "theater", "theatre", "concert"), ("marketing", "email", "demographics", "tiers", "events", "live entertainment")),
 
28
  ("airline_route_operations", ("air transport", "airline", "airport", "aircraft", "flight", "passenger", "baggage", "route"), ("transportation", "logistics", "operations", "sales")),
29
  ("restaurant_store_operations", ("restaurant", "coffee", "cafe", "store operations", "daypart", "drive-thru", "mobile order", "delivery"), ("store", "transactions", "labor", "wait time", "customer satisfaction")),
30
  ("retail_sales", ("retail", "store", "merchandising", "apparel", "fashion", "electronics"), ("sales", "revenue", "margin")),
 
25
  ("hotel_finance", ("hotel", "hotels", "hospitality", "property", "room", "rooms", "occupancy", "revpar", "adr"), ("finance", "margin", "budget", "forecast")),
26
  ("apparel_marketing", ("apparel", "fashion", "garment", "footwear", "seasonal drop", "markdown"), ("marketing", "campaign", "loyalty", "channel", "conversion")),
27
  ("sports_venue_fan_engagement", ("sports", "arena", "venue", "stadium", "ticket", "tickets", "ticketing", "attendance", "fan", "fans", "season ticket", "premium seating", "suite", "suites", "sponsorship", "sponsor", "team", "league", "basketball", "hockey", "soccer", "baseball", "football", "theater", "theatre", "concert"), ("marketing", "email", "demographics", "tiers", "events", "live entertainment")),
28
+ ("ad_monetization", ("ad yield", "ad monetization", "advertising monetization", "arpu", "ecpm", "fill rate", "ctv", "smart tv", "smartcast", "fast channel", "connected tv"), ("placement", "impressions", "ad requests", "completion rate", "watch time", "audience", "inventory")),
29
  ("airline_route_operations", ("air transport", "airline", "airport", "aircraft", "flight", "passenger", "baggage", "route"), ("transportation", "logistics", "operations", "sales")),
30
  ("restaurant_store_operations", ("restaurant", "coffee", "cafe", "store operations", "daypart", "drive-thru", "mobile order", "delivery"), ("store", "transactions", "labor", "wait time", "customer satisfaction")),
31
  ("retail_sales", ("retail", "store", "merchandising", "apparel", "fashion", "electronics"), ("sales", "revenue", "margin")),
tests/test_dataset_first_builders.py CHANGED
@@ -447,6 +447,37 @@ def test_trucking_sales_does_not_route_to_trucking_finance():
447
  assert "TRUCKING_FINANCIALS" not in build.dataset.table_map()
448
 
449
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
450
  def test_grocery_finance_routes_to_cpg_financials_not_sales_fact():
451
  build = build_dataset_first_demo(
452
  company_name="General Mills",
 
447
  assert "TRUCKING_FINANCIALS" not in build.dataset.table_map()
448
 
449
 
450
+ def test_ad_yield_arpu_routes_to_ad_monetization_not_sales_pipeline():
451
+ build = build_dataset_first_demo(
452
+ company_name="Vizio",
453
+ company_url="https://www.vizio.com/",
454
+ use_case=(
455
+ "Focus on Ad yield and ARPU, optimizing revenue through SmartCast users "
456
+ "across CTV placements, FAST channels, ad requests, eCPM, fill rate, "
457
+ "completion rate, watch time, and audience cohorts."
458
+ ),
459
+ vertical="Media & Entertainment",
460
+ function="Revenue",
461
+ row_count_guidance=100,
462
+ )
463
+
464
+ assert build is not None
465
+ assert build.scenario.scenario_type == "ad_monetization"
466
+ table_map = build.dataset.table_map()
467
+ assert "AD_MONETIZATION" in table_map
468
+ assert "SALES_PIPELINE" not in table_map
469
+ fact_columns = {column.name for column in table_map["AD_MONETIZATION"].columns}
470
+ assert {
471
+ "AD_REVENUE_USD",
472
+ "ARPU_USD",
473
+ "ECPM_USD",
474
+ "FILL_RATE_PCT",
475
+ "AD_YIELD_PER_WATCH_HOUR_USD",
476
+ "REVENUE_LEAKAGE_USD",
477
+ }.issubset(fact_columns)
478
+ assert validate_dataset_quality(build.dataset).ok
479
+
480
+
481
  def test_grocery_finance_routes_to_cpg_financials_not_sales_fact():
482
  build = build_dataset_first_demo(
483
  company_name="General Mills",