Forward-integration niche selection · BetterStudio

★ First bet 4 / 5 council seats

Sporting goods

CAD-native bikes · fitness · outdoor hard goods

The only finalist in the winning quadrant: structural fidelity × highest asset-readiness (9/10). The variant explosion (one model = ~60 colour×size variants) is the automation tailwind — touch-per-image falls with volume.

Co-lead / fast-follow score leader · 72

Furniture & homeware

Biggest, best-proven demand — and the #1 services-trap

Wins on TAM and quantified ROI, but every lens flagged the same failure mode: buyers may be paying for merchandising taste, not accuracy. "Revenue that looks good and margins that never arrive." Test it; don't lead with it.

The one move

Run a single blind first-pass-usable bake-off on one CAD-native brand's real catalog — hard-weighted to chrome/carbon — wrapped in a $2–5k paid pilot, and start with whichever brand says yes first.

It checks product, asset-premise, and demand in one motion. Sporting goods should win the race on asset-readiness.

Three findings for the room

▸ 1 · The scorecard leader is not the first bet tap

Furniture has the biggest, best-proven demand — and is the niche most likely to become an agency. Lead with the one hardest to turn into a studio (sporting goods), not the one with the biggest TAM. Council #1 mandated this: "the winner is NOT the biggest TAM."

▸ 2 · Footwear — the old favourite — is now last tap

The most useful finding. ~70% of footwear returns are sizing/fit, not visual → the pixel-fidelity wedge owns a minority of the pain; the rest leaks to AR/fit-confidence where Fibbl already wins.

And brands don't hold production 3D — Fibbl scans physical shoes precisely because usable GLB doesn't exist, so "consume the brand's 3D" breaks. Discipline: size the addressable pain, not the total pain.

▸ 3 · "Buy a cheap working business" is a mirage tap

Every niche scored acquisition 3–4/10. No clean sub-$1M productized targets on Acquire/Flippa/Empire Flippers/BizBuySell — only fragmented founder-labor studios where the founder is the product. A buy is an acqui-hire (relationships/pipeline), not a validation shortcut. Validate-fast, not buy-fast — revisit acquisition only after the gate is proven.

Honest caveat · Demand here is revealed-preference, not search volume. Ahrefs was hard-blocked this cycle (workspace over its monthly unit limit; 0 credits burned — calls rejected, not billed; resets 2026-07-12). Every finalist is "investigate," not "go."

The kill-funnel · 12 → 5 → 1

7 killed cheaply, 5 deep-dived, 1 first bet

A council-designed funnel so weak niches die in minutes, before any deep diligence or scarce Ahrefs credits.

12

verticals screened

5

passed to deep-dive

1

first bet

Two axes decided everything

1 · Structural vs specular fidelity

Can an automated gate score it? Structural (shape/colour/material → yes): furniture, sporting, bags, footwear. Specular/refractive (no): jewelry, eyewear, lighting, cosmetics, chrome fixtures.

2 · 3D-asset readiness

Do buyers already have usable 3D? High: sporting (9), kitchen&bath (8), furniture (7). Weak: bags (4), footwear (4) — modelling becomes the real business.

The winning quadrant — structural × high-asset — holds Sporting goods alone. Furniture is structural but services-trap-heavy; bags & footwear are structural but asset-poor.

The 7 killed — and why

The kill pattern: most deaths trace to demand≠edge (the niche's fidelity is specular, where the gate is weakest) or wedge-doesn't-apply (the real pain is fitment data, 2D texture-warp, or fit/try-on — not 3D-object fidelity).

5 finalists · the 8-dimension rubric

Shortlist (≥75) is empty

No slam-dunk — every finalist is "investigate," gated on a cheap test. Acquisition scored 3-4/10 for all five. Tap any niche for the full breakdown + evidence.

Read the scores this way · these rank raw fit. The councils deliberately overrode them for the first bet — so Sporting goods (66) leads despite Furniture's 72, because size isn't the tie-breaker: asset-readiness + lowest services-trap risk are.

Rubric weights: Demand 20 · Fidelity-pain 20 (hard-kill <12) · Validation 15 · Automation 12 (hard-kill <6) · Asset 10 · Acquisition 10 · Runflow-edge 8 · Flywheel 5.

How it was decided · 2 councils

Plan it, then red-team it

Each council = a 4-persona Codex panel (growth · product · CFO · risk) + one Claude cross-check seat on a different engine, to surface disagreement.

Council #1 · plan

Designed the kill-funnel, the validation playbook, the 100-pt rubric, and 10 false-positive traps. 5/5 CAUTION: run a staged funnel, not a broad market scan. The #1 fear: the services-trap.

Council #2 · decide

Evaluated the findings and overrode the scorecard. 4/5 seats: Sporting goods first. All five agreed on the footwear demotion, the acquisition mirage, and the same cheapest test.

The one disagreement — and why it's small

The Claude cross-check seat dissented. That split is the most valuable output: it's about sequencing, not direction.

Sporting goods first · 4 seats

Weights operational/margin safety + asset-readiness. Least likely to become a services business; objective pilot signal (geometry/colour/finish → publishable yes/no).

Furniture first · 1 seat (Claude PM)

Weights demand certainty + wedge-fit. The only finalist where demand, fidelity, structural wedge and asset supply all clear at once — lead with the strongest proven market.

Both are structural-fidelity niches that pass the same test, so a result transfers. Both name the other as fast-follow. Resolution: a two-way door — run the same bake-off and let buyer-pull + asset-readiness break the tie. Whichever CAD-native brand signs a paid pilot first goes first.

▸ What changed vs the prior thesis (research/040) tap

The prior favourite footwear (Fibbl ACV proof) is now last — the ACV is real but the addressable pain and the asset premise both fail.
The real first bet is sporting goods — a niche not even on the original candidate list, surfaced by forcing discovery beyond furniture/footwear/eyewear/jewelry.
The "acquire a cheap business" half of the strategy is not viable as a validation shortcut.

The decisive test · run before committing runway

One bake-off, three risks retired

One CAD-native brand. One model / catalog slice (5–15 SKUs), deliberately weighted to the hard cases. Generate via the engine + Sentinel gate. The brand's own merchandiser blind-grades "PDP-publishable, yes/no" vs their current output — no retouch — wrapped in a $2–5k paid pilot sourced within ~10 business days.

Checks product

Does the gate clear ~42% first-pass (≥25 pts over the ~17% baseline) without retouch?

Checks asset

Does usable production CAD actually exist?

Checks demand

Will a real brand pay for it?

Pre-registered kill criteria · decide before you see the images

✕ First-pass-usable on the hard finishes doesn't clear ~42% (≥25 pts over the ~17% incumbent baseline), or
✕ "Publishable" requires manual art direction on most variants (→ services trap), or
✕ No brand will provide usable CAD and paid intent within 10 business days (→ demand/asset signal is thin).

Pre-registering the kill criteria stops a "pretty but needs-retouch" result from rescuing a dead niche. The pilot also captures willingness to switch, not just the usability percentage.