FURN | AI IMAGE ECONOMICS RECORD Guide: https://furncmo.com/blog/furniture-marketing-roi-ai-images Updated: October 21, 2026 1. APPROVED OUTPUT BRIEF Product / variant / source images / intended channel: Required views and output count / acceptance standard: Product fidelity / rights / channel rules / metadata checks: Reviewer / exact approved files / unresolved requirements: 2. PRODUCTION COMPARISON (REPEAT PER METHOD) Currency / period / cost basis (cash, allocated labor or both): Source preparation / production / tool usage: Review / rejected candidates / rework / export and delivery: Shared costs and overhead included or excluded consistently: Total listed cost / completed approved images: Cost per approved image = listed cost / approved images. If no approved images, ratio unavailable; retain incurred costs. Required outputs still missing / estimated cost to finish: Cash savings vs released staff capacity vs committed costs: 3. CREATIVE TEST PLAN Hypothesis / primary business outcome / guardrails: Comparison method / eligibility / allocation / budget limit: Product, offer, audience, destination and other controlled factors: Tracking validation / value definition / attribution settings: Duration / conversion delay / stopping rules / owner: Review uncertainty and primary outcome, not only best metric. No fixed duration guarantees a conclusive result. 4. CAMPAIGN LEDGER (REPEAT PER COHORT) Revenue basis: after discounts/refunds; tax treatment: Product / fulfillment / returns / payment costs: Media / allocated creative and management costs: ROAS = attributed revenue / media spend (if spend is positive). Contribution after listed costs = revenue minus all listed costs. Other overhead and tax excluded? State explicitly: No double-counting costs or mixing currencies without a stated basis. Attribution is not evidence of incremental sales caused by imagery. 5. DECISION Outcome / interval or uncertainty / data unavailable and why: Quality failures / cost differences / campaign confounders: Continue, revise, stop or inconclusive / rationale / owner: Next review trigger / original test and evidence location: Worked examples in the guide are fictional, not forecasts.