BHI 3.1·Q2 2026 Bulletin
Published Jun 4 2026·Spearman ρ = 0.77·Validation: preliminary
Today's bulletin →
Movers · Q2 2026 · Jun 4 2026
NVIDIAB 8.29▼ 0.10VisaB 8.82▲ 0.09ChinaB 8.64▲ 0.08WeChatB 17.03▼ 0.08MastercardB 8.14▼ 0.07EpicB 11.91▲ 0.07Synopsys/CadenceB 12.97▼ 0.06OracleB 8.43▼ 0.06ASMLB 9.62▼ 0.06NationalB 11.27▼ 0.06FAAB 12.20▲ 0.05PJMB 12.20▼ 0.05NVIDIAB 8.29▼ 0.10VisaB 8.82▲ 0.09ChinaB 8.64▲ 0.08WeChatB 17.03▼ 0.08MastercardB 8.14▼ 0.07EpicB 11.91▲ 0.07Synopsys/CadenceB 12.97▼ 0.06OracleB 8.43▼ 0.06ASMLB 9.62▼ 0.06NationalB 11.27▼ 0.06FAAB 12.20▲ 0.05PJMB 12.20▼ 0.05
§ 06 · VALIDATION REPORT · V3.1

Validation

Preliminary empirical validation of B-index scores against observable annual retention rates for 14 platforms with directly comparable metrics.

Spearman ρ
0.77
B-index vs. annual retention rate · n = 14 · p < 0.001
ICC(3,k)
Pending
Seeking independent evaluators · target > 0.75
Scoring
Single
Single evaluator with anchored rubrics
Status
Preliminary
Pre-empirical · validation roadmap in progress

BHI score vs. annual retention

B-index vs. annual retention rate (%) for 14 platforms with directly comparable, publicly documented retention metrics (SEC filings, earnings calls, CIRP, Antenna). Dashed line is the OLS fit.

051060%70%80%90%100%B-INDEX (V3.1)ANNUAL RETENTION RATE (%)TSMC · B=10.91, retention=99%Visa · B=8.82, retention=99.9%Amazon · B=8.40, retention=95%SWIFT · B=6.78, retention=99.9%Nubank · B=5.21, retention=94%Apple · B=4.88, retention=89%Salesforce · B=4.54, retention=92%Spotify · B=2.31, retention=96.5%Netflix · B=1.00, retention=78%Zoom · B=0.90, retention=92%ρ = 0.77 (p < 0.001) · n = 14

Key insight

BHI predicts a retention floor, not a ceiling. No high-BHI platform shows weak retention; several low-BHI platforms show strong retention driven by product quality rather than structural lock-in.

Informative outliers: Spotify (B = 2.31, retention 96.5%) — retention exceeds structural lock-in prediction, driven by product quality. Disney+ (B = 0.71, retention 61%) — lowest retention in cohort confirms low structural lock-in. SWIFT (B = 6.78, retention 99.9%) — cooperative infrastructure creating deeper lock-in than corporate platforms.

Limitations of this validation:
Small sample (n = 14) · Heterogeneous retention metrics · Source quality varies · Single evaluator · Correlation does not establish causation

Reference cohort · n = 14

Platforms with directly comparable, publicly documented annual retention rates.

PlatformSectorBHI · V3.1Retention (%)Source
TSMCtech10.9199.0TSMC annual report
Visafintech8.8299.9Visa 10-K filing
Amazontech8.4095.0CIRP consumer survey
SWIFTbanks6.7899.9SWIFT annual review
Nubankfintech5.2194.0Nubank earnings
Appletech4.8889.0Antenna churn data
Salesforcesaas4.5492.0Salesforce 10-K
Spotifygaming2.3196.5Spotify earnings
Netflixgaming1.0078.0Antenna churn data
Zoomsaas0.9092.0Zoom earnings

Validation roadmap

Q2 26
Expanded reference set. Lift cohort from 14 to 30+ platforms with documented switching cost or retention data.
Q3 26
Independent evaluators. Recruit 3-5 independent scorers. Compute ICC and Krippendorff's alpha on 15-20 platforms.
Q3 26
Public scoring rubric audit. Anchor definitions opened for 30-day public comment.
Q4 26
Pre-registered V4 study. OSF-registered protocol filed before scoring begins.
Q1 27
External replication package. Full scoring spreadsheets and code released under CC BY 4.0.

Open challenges

  1. Single evaluator. All 184 platforms were scored by a single evaluator. This is the most critical methodological limitation. Multi-evaluator scoring is the top priority.
  2. Sector heterogeneity. The B-scale is calibrated globally; sector-relative percentiles may be more useful.
  3. Time resolution. Quarterly cadence is coarse for fast-moving AI-platform dynamics.
  4. Causality. BHI is descriptive, not causal. Correlation does not establish causation.