Conclusion — OPEN

No conclusion is drawn. The study's decision rule (method.md) is cross-µarch by construction, and only one µarch (apple-m2-mba, NEON) is registered. Drawing a general bitmap-vs-selvec recommendation from a single ISA would violate both the pre-registered rule (≥3 ISAs, ≥5 µarchs) and Charter T2 (no over-claimed data).

What can be stated now (one µarch only)

On apple-m2 / NEON, bitmap and selvec K1 production are both flat in selectivity, and bitmap is the cheaper producer by a flat ~9 % (no producer-side crossover). This is an existence result for H1 and one point of H2 — not a recommendation.

What closing the study requires (deferred, R-06)

1. Registered ≥3 ISAs across ≥5 µarchs (x86 AVX2/AVX-512 + Graviton-class ARM added to apple-m2-mba).
2. qle:-referenced ledger entries for compare (both representations) per µarch.
3. The bench_pipeline producer+consumer measurement per µarch (the crossover is downstream).
4. Evaluate the pre-registered decision rule; then this file states the recommendation.

The DaMoN/ADMS workshop-paper draft (Charter §9.1) is tracked outside the repo; its link is recorded in the M9 gate when it exists. It, too, waits on the multi-µarch data.

Traceability: gate M9; risk R-06 / REQ-LEDGER-012 (owned by M10).