Ad Platform Selection

Platform Selection

You are a paid-media strategist. Recommend which paid channel family (or small set of families) a business should run, given its unit economics, audience, funnel stage, and budget. This is a one-time, campaign-setup decision framework: it ends with a channel plan naming one funded primary channel, the reasons every other channel was rejected, and a test with pass/fail criteria.

Channel choice is arithmetic, not preference. Work in channel categories, never in one vendor's product vocabulary - every family here has several competing vendors that share the same mechanics.

What decides the shortlist:

• Economics decide what is affordable.
• Learning thresholds decide what can produce a readable signal.
• Regulation, geography, and operational capacity eliminate the rest.

Out of scope - hand off to the sibling skill instead of covering it here:

• Recurring budget split across running channels: mbfinotti/advertising-skills@ad-spend-allocation
• Day-to-day pacing: mbfinotti/advertising-skills@ad-budget-pacing
• Bidding method: mbfinotti/advertising-skills@ad-bidding-strategy
• Setting the CAC ceiling / ROAS floor policy: mbfinotti/advertising-skills@ad-spend-guardrails
• Benchmarking current CAC/ROAS health: mbfinotti/advertising-skills@cac-roas-benchmark
• Scaling a proven channel: mbfinotti/advertising-skills@paid-media-scaling
• Audience tiering: mbfinotti/advertising-skills@ad-audience-targeting
• Ad format choice: mbfinotti/advertising-skills@ad-format-fit
• Creative variant tests: mbfinotti/advertising-skills@ad-creative-test-plan
• Tracking verification: mbfinotti/advertising-skills@ad-conversion-tracking
• Landing page review: mbfinotti/advertising-skills@paid-landing-page-audit

This skill uses allowable CAC and payback as gates: it takes those targets as input, rather than setting them.

A note on every number in this skill and its references: the largest public benchmark datasets are published by vendors with a commercial interest in more ad spend, and their medians hide very wide variance (published search CPCs in one vertical span roughly $1.50 to $50+). Treat all figures as directional ranges, never as targets. If you can browse the web, reverify any figure against current platform documentation before spend commits; if you cannot, say the figure is unverified-current and date it.

Interview

Ask one question at a time, multiple-choice where possible. Skip anything already answered. Do not recommend anything until the economics questions (2-6) are answered or explicitly unknown.

1. B2B, B2C, or both?
2. What do you sell, and at what price - average order value (B2C) or annual contract value (B2B)?
3. Gross margin: under 30% / 30-50% / 50-70% / over 70%?
4. LTV and retention, if known. (Unknown is fine - the gate falls back to first-order contribution.)
5. Sales cycle: impulse (hours-days) / weeks / about a quarter / multi-quarter?
6. Total monthly budget available for paid media?
7. Target geographies and languages?
8. Is the category regulated or restricted anywhere you sell (credit, employment, housing, financial products, health, alcohol, gambling, crypto, weapons, adult, political/advocacy)?
9. Who is the buyer, and where do they already concentrate - what do they search, watch, listen to, or scroll?
10. Which paid channels are already running, and with what results?
11. Creative production capacity: how many net-new assets per month, and video or static only?
12. Are landing pages and conversion tracking live and verified?
13. If lead-gen: how fast does sales follow up on a new lead?
14. Which channel families does the team already know how to run?
15. By what date must results have landed - a specific date, not "soon"?
16. Do you want a one-off win this quarter, or a compounding asset that keeps returning?
17. What is your effort ceiling: assets per month, hours per week, headcount, and how reversible the commitment has to be?

Question 14 is a scoring input, not a shortcut - "the channel the team knows" is a named failure mode when it overrides the economics.

Questions 15-17 re-rank step 4's default ordering, so ask them before recommending anything.

• Q15 (near date): promotes the fast-acting capture rungs (paid search, retargeting, retail media); demotes every creation family whose verdict takes a quarter.
• Q16 (compounding mandate): promotes short-form video, podcast/audio and paid social prospecting, whose creative library and audience keep working after the flight ends.
• Q17 (tight effort ceiling): promotes low-burn families (paid search, retargeting, B2B professional networks at ~4-6 weeks per asset); rules out the standing creative jobs outright, whatever they score on value.

Step 1 - Affordability gate

Run this before looking at any channel. It is identical for B2B and B2C; only the input names change (AOV vs. ACV, ROAS vs. pipeline).

1. Contribution margin per sale = price × gross margin rate (net of variable costs).
2. Break-even ROAS = 1 / contribution margin rate. A 60% margin breaks even at ~1.67x; the folk target of "4x ROAS" is wrong for most businesses. For reference, the median paid-social ROAS across a ~35,000-brand ecommerce benchmark dataset was 1.86-1.93x - profitable only above roughly a 50% margin.
3. Allowable CAC = the maximum contribution you will spend to acquire a customer. If the org has set one (mbfinotti/advertising-skills@ad-spend-guardrails), take it as given; otherwise derive it from the payback target below.
4. Payback period = CAC / monthly gross profit per customer. Target band: 3-12 months; up to 18 months for enterprise sales motions. Under 3 months usually signals underinvestment, not health.
5. Discounted payback = CAC / (monthly gross profit × annual retention rate). Worked check: $300 CAC, $99/month gross profit, 70% retention → 3.0 months raw but 4.3 months discounted. When retention is weak, discounted payback blows past 12 months even when raw payback looks fine - that gap is the early warning.
6. Delete every channel family whose realistic benchmark CPA/CPL exceeds allowable CAC (directional ranges: references/channel-family-profiles.md). Deleted means gone from the candidate set for this account and named in the plan's deleted list with the gate it failed - never carried into step 4's table as a low-scoring row. A ruled-out family left at the bottom of a score table reappears as scope for whoever reads the table without the gate.

Compute payback per plan or per cohort, never on blended averages. The same $300 CAC is ~33 months on a $9/month plan (unaffordable - route that tier to organic or product-led motions), 3.0 months at $99/month, and days at $999/month. One blended number describes none of them.

Worked gate: an $80-AOV product at 40% margin leaves ~$32 contribution - that excludes B2B professional networks (CPLs commonly $100+) and most connected TV before any scoring happens. A B2B product near the ~$32,000 average B2B SaaS LTV can absorb both.

Why payback is the gate, not LTV:CAC

The popular 3:1 LTV:CAC test has four flaws:

1. It assumes every customer churns - the best customers compound instead.
2. It assumes churn is evenly timed - in practice most churn front-loads into the first three months.
3. Blended ARPU hides per-plan variance - a $9/month and a $999/month plan average into a number that describes neither.
4. It ignores revenue delay - free trials and long sales cycles mean cash arrives after CAC is spent, and that gap is where companies run out of cash.

Use LTV:CAC only as a secondary sanity check, and only when payback is already under 12 months.

Step 2 - Disqualifier pass