/kai-budget — Spend Allocated Against Unit Economics, Not Vibes

Objective

A marketing budget with defensible arithmetic behind every line: unit economics modeled or explicitly estimated, channel-by-channel allocation with expected CAC and volume, three ROI scenarios, a ramp schedule, and the kill/scale criteria that decide what happens next quarter. The number the user leaves with is one they could defend to a board.

Broken economics are the first finding, not a footnote. An LTV:CAC below 1:1 or a payback beyond 18 months changes the whole recommendation.

Done when

Work type strategy-plan — floor E3/C3/O1 (harness/eco-floors.yaml).

• E3 — a named human approved the exact budget document. Approving the plan is not authorizing the spend; spend authorization is a separate decision.
• C3 — banned_word_check clean, every projection carries a confidence level, every assumption that needs validation is flagged as such, and someone other than the author read it end to end.
• O1 — every allocation line names the metric that proves it worked (CAC, volume, payback), its baseline, its threshold, and its owner, recorded before the first dollar moves. A budget with no predeclared thresholds cannot be graded later.

Constraints

• Read MARKETING.md from the project root first. It carries product, ICP, value prop, monetization, voice, current channels, and competitive landscape. If it does not exist, build it from the codebase — CLAUDE.md, README.md, PROJECT.md, package.json, landing pages, email/ad/analytics config — using the template from /kai-email-system, and confirm the draft. Do not ask the user what the product is.
• Eight things must be known before allocating anything: business model (SaaS, ecommerce, services, marketplace); current MRR/ARR; growth stage (pre-launch, early, growth, scale); current spend total and by channel; the target growth rate or revenue goal; sales cycle length and average deal size; current CAC and LTV if known; and team composition (in-house, agency, contractors).
• Where metrics are missing, use industry benchmark ranges and label them as estimates. An estimated input never gets presented as a measured one.
• Never recommend spend the user cannot sustain for 6 months.
• Always reserve 10–20% for testing.
• CAC targets are validated against LTV. Never recommend acquisition that loses money at the target volume.
• Every projection carries a confidence level — High, Medium, or Low.
• Stage-appropriate only. No enterprise tactics for a pre-launch company.
• Round to practical amounts. $4,700 or $5,000, not $4,731.28.
• This document does not authorize spend. It recommends it.

Context

Need	Load
Budget models, benchmarks, forecasting method	knowledge/playbooks/marketing-budget-forecasting.md
Unit economics, benchmarks by business model	knowledge/playbooks/saas-metrics-guide.md
Persona alignment for channel choice	knowledge/personas/_persona-index.md
Revenue, stage, current channels	MARKETING.md (project root)

Unit economics to compute or estimate before allocating: CAC (total spend / new customers) · LTV (ARPU × average lifespan) · LTV:CAC, target 3:1 or better · payback period in months · gross margin, which sets spend capacity. Benchmark each against the business model. Flag broken economics explicitly: LTV:CAC below 1:1, or payback beyond 18 months.

Per active channel, assess: current spend, volume (leads, signups, purchases), channel CAC, trend direction (improving, stable, degrading), and saturation risk at current spend.

Deliverables:

• Allocation model — total recommended monthly/quarterly budget, the stage-appropriate percentage-of-revenue benchmark, and a per-channel table: Channel / Monthly Spend / Expected CAC / Expected Volume / Confidence.
• Three ROI scenarios — Conservative (80% of target performance), Base case (expected), Aggressive (120% with increased spend). Each states spend, leads, customers, revenue, ROI, and payback period.
• Spend ramp plan — month-by-month schedule rather than a single dump, the testing allocation, kill criteria (when to stop spending on a channel), and scale criteria (when to increase).
• Optimization calls — which channels to increase (high ROI, unsaturated), decrease (poor CAC, saturated), and test (untapped, stage-appropriate).
• One-page executive summary — total recommended spend, expected acquisition volume, blended CAC target, projected ROI, and the top 3 risks with mitigations.
• A quarterly review cadence, and an explicit list of assumptions needing validation against real data.

Escalate when

• Unit economics come out broken (LTV:CAC below 1:1, payback past 18 months) — the answer is a fix, not a budget.
• The revenue goal requires spend the business cannot sustain for six months.
• No CAC or LTV data exists and the user wants precise projections rather than ranges.
• The requested allocation contradicts the company's stage.
• The plan would be read as spend authorization rather than a recommendation.