1The Cross-Domain Uniformity Paradox
Cross-domain participation is universal, but cross-domain INTENSITY is demographically random across 14 cuts tested.
Confirmed Meta-finding
Per-sub spread (segment × $)
1.0%
Evidence: Per-sub lifetime ticket spend by segment ranges $4,514 (Business Small) to $4,539 (Consumer) — a 1.0% spread. By age band, by credit, by tenure, by acquisition channel: every cut produces spreads inside the statistical noise floor.
So what: This warehouse should not be demoed with persona-based marketing narratives. Lead with cross-domain VOLUME and event-based cuts (F4) instead.
2No Carrier ↔ Platform Parent Affinity
Comcast subs do NOT over-stream Peacock. Vertical integration produces no detectable streaming preference.
Null Finding Cross-Domain
Evidence: Comcast subscribers' streaming distribution by platform — Peacock (Comcast-owned) is their LOWEST at 19,138 sessions; Tubi (Fox) is highest at 19,676. Spread: 2.8%. Sqrt-counting noise floor: 0.7%.
So what: Killing the "bundle synergy" narrative for this dataset. Useful as a credibility-establishing null result in prospect demos.
35G Doesn't Drive 4K HDR Streaming
Identical quality-tier proportions across 5G, 4G LTE, and Fiber subscribers.
Null Finding
Evidence: 4K HDR share: 15.0% (5G) vs 15.0% (4G LTE) vs 14.9% (Fiber). 1080p: 54.7% / 54.7% / 54.6%. Streaming quality choice is statistically independent of network technology.
So what: The 5G-upsell-for-4K narrative has no support in this data. Use plan-type ARPU differentiation for the 5G ROI demo instead.
4COVID Q2 2020 Cross-Domain Substitution
Same population substituted streaming for live events. Cleanest cross-domain signal in the warehouse.
Confirmed Cross-Domain
Streaming Subs Q2'20
+41.8%
Tickets Q2'21 vs '19
+12.4%
Evidence: Distinct subscribers in each fact, by quarter 2019–2021. Q2 2020 ticket-buying subs collapsed from 9,263 (Q2'19) to 804. Streaming subs jumped from 25,292 to 35,874. View minutes rose from 1.46M to 2.23M. By Q2 2021 ticket buying recovered to 10,413 (12.4% above pre-COVID baseline); streaming returned to baseline trajectory.
So what: Lead with this story for prospect demos. It's the cleanest demonstration of cross-domain substitution capability — same SUBSCRIBER_KEY, two completely different behaviors at two points in time.
5Heavy-Tailed Spending, Demographically Random
Top 5% drives 16.7% of spend. But they look identical demographically.
Confirmed Counterintuitive
Evidence: Top ventile (4,404 subs) avg $15,129/sub vs bottom ventile $42 — 360× spread. Top 20% generates 48.1% of total spend. Yet super-spenders are 4.83–5.06% of every segment — perfect uniform distribution.
So what: "VIP cohort" exists but cannot be predicted from demographics. Build behavioral cohorts (RFM, recency, cumulative spend) — not demographic.
6Enterprise Plans Skew Non-Purchaser
The only demographic discriminator that survives statistical noise: 1.1pp difference between Enterprise and Prepaid 5G.
Weak Signal
Evidence: Enterprise Fiber subs are 47% more likely to never buy a ticket vs Prepaid 5G (3.45% vs 2.35%). Business Direct acquisition channel also under-indexes ($4,448/sub vs $4,591 Dealer). Consistent corporate-account / work-account signal.
So what: The one defensible "plan type matters for entertainment behavior" claim. Add an Enterprise vs Consumer plan profile to the Plan & Carrier Mix dashboard.
7Tenure Doesn't Predict Churn
Churned and not-churned subscribers have identical avg tenure (60.4 months).
Null Finding
Evidence: 55,064 subs with 2024–25 churn events: avg tenure 60.4 months. 39,556 active subs with no churn: avg tenure 60.4 months. Identical to the decimal place. Combined with near-uniform churn-reason distribution, the warehouse models churn as memoryless.
So what: Don't promise churn-prediction demos on this dataset. The data has no demographic predictive signal.