Every manufacturer you bring into the Product Master costs one Mint Namespace — a compiler, normalization rules, a canonical source. Skip one and every product they make stays hand-keyed forever. This ranks all 464 feeds by what building one actually saves: products off the manual-entry pile first, then spend and reach.
Ranked by manual-entry burden — the distinct products you'd otherwise hand-key. Flip the axis to spend or reach; the meter follows whichever you pick.
Two owner families collapse to a single feed — one compiler covers the whole shelf, so they rank as one build, not a scatter of siblings. AMX stays separate. Numbers are deduped at the feed level (products and projects), so these aren't sums.
AMX is kept as its own feed (— products, —) — it reaches the Master through a separate Harman channel, so it earns its own compiler. Legrand AV folds in every owned brand in your data; WattBox is deliberately left out — it's Snap One, not Legrand.
Products is the rank that matters most — it's the manual entry you erase. But spend, reach, and raw volume each pull differently, and where they disagree is the call.
line.product_id → product), kind = physical, owner-furnished excluded, hidden kept. Project phases exclude TEMPLATE and TRASH.line.cost → line.price → product.cost → product.price → mapp → msrp, times quantity. Avoids multiplying by a missing price.