type: framework-update tags: [portfolio-construction, conviction-tier, operating-margin, sizing, quality-divergence, equal-weight] confidence: medium created: 2026-04-07 source: PORTFOLIO portfolio-review 2026-04 persona: wpr provenance: legacy source_analysis_path: null source_paragraph_quote: null source_transcript_span: null source_loss_log_path: null

Equal-Weight Conviction Tiers Mask Within-Tier Quality Divergence

When multiple positions share the same top allocation tier (e.g., all at 10%), identical sizing implicitly treats them as equivalent quality bets. In practice, operating margin profiles within a conviction tier can diverge by 60-70 percentage points, making equal-weight an unreliable quality signal. The correct check: within each allocation tier, compare operating margins and SBC rates — if divergence exceeds ~30pp OM, the equal-weight assumption deserves scrutiny.

Evidence

• muji portfolio Apr 2026: APP (10%) and AXON (10%) both in the "highest conviction" tier. APP: 66% GAAP net margin, SBC <5% of revenue, FCF yield exceptional. AXON: -6.3% operating margin, 23% SBC/revenue, FY25 GAAP net income $125M on $2.8B TTM revenue (4.5%). Identical sizing obscures a ~70pp margin divergence.
• wpr assessment: "AXON at 10% gets the same allocation as APP (66% growth, 77% OM) and CRDO (201% growth, 37% OM). The numbers don't support that parity."
• Recommended trim: AXON from 10% to 6%, reallocating to RDDT and BLLN where profitability ratios are substantially better.

Implication

In portfolio reviews, audit quality homogeneity within each allocation tier — not just overall concentration levels. A simple OM spread check (max minus min operating margin within the top-5 positions) flags when equal-weight is implicitly equating very different risk profiles. If within-tier OM spread exceeds 30pp, investigate whether the lower-quality position warrants trimming to a lower tier. This is especially relevant in growth portfolios where conviction is commonly proxied by position size rather than quality-adjusted size.