04.05 Off-Site Paid Media & Search Growth SOP (Google / Meta / TikTok)
Governance metadata | Owner: Head of Marketing | Last reviewed: 2026-10-22 | Next review: 2027-01-31
Core principle: off-site paid media is not a by-product of the brand team; it is the main engine for cold start and scaling. But media buying only puts the right creative in front of the right people—attribution and review belong to 05.19 / 05.21.
1. Boundary: 04 Buys, 05 Attributes
| Stage | Owning document | Deliverable |
|---|---|---|
| Account structure, creative testing, budget scaling, stop-loss | 04.05 (this document) | Daily media report, creative library, budget plan |
| UTM/pixel/events, attribution model, ROI review | 05.19 / 05.21 | Attribution report, channel contribution |
| On-platform PPC (Amazon) | 05.01 | On-platform ad report |
2. How the Three Platforms Actually Differ: Know What You Are Buying
The three platforms are not "one set of creatives placed in three places". They differ in auction mechanics, user intent, attribution windows and review strictness — which determines account structure, creative approach and budget split.
| Dimension | Google (Search / PMax / YouTube) | Meta (FB / IG) | TikTok |
|---|---|---|---|
| What you buy | Keyword auction — expressed intent | Audience auction — interest that may be sparked | Interest/content auction — discovery and impulse |
| User state | Actively searching, comparing | Passively browsing, being seeded | Consuming entertainment, very short attention |
| Main play | Brand defence + category capture + competitor terms | Demand creation + lookalike scaling | Native content + creator assets + impulse conversion |
| Targeting logic | Keyword match + audience layering | Interest/behaviour/lookalike/retargeting | Interest/creator/Spark Ads/automated targeting |
| Creative fatigue | Slow (search leans on copy) | Medium (refresh every 2~4 weeks) | Fast (decays in 1~2 weeks) |
| Default attribution window | 30-day click / 1-day view | 7-day click / 1-day view | 7-day click / 1-day view |
| Review sensitivities | Trademark use, comparative ads, efficacy claims | Health/efficacy, before-after imagery, inducement | Content tone, minors, exaggerated promises |
| Learning-period constraint | Smart bidding needs conversion volume | Each ad set needs stable conversions to exit learning | Same, and creative decays faster |
| Best-fit stage | When search demand exists (harvest) | Awareness to consideration (seeding) | Cold-start exposure / impulse categories |
⚠️ Attribution windows follow each platform's current defaults; the values above are common defaults and change with platform policy. Unify the window before comparing ROAS across platforms, or the conclusion will be wrong (see 05.21 Paid Media Attribution in Practice).
Three rules this implies:
- One creative cannot be run in three places. TikTok needs native feel (the first 3 seconds decide), Meta needs the benefit up front, Google search needs copy-to-landing-page consistency. Re-cropping one video is the main source of wasted budget.
- Do not split budget evenly. Allocate by intent strength — search captures existing demand, Meta creates it, TikTok competes for attention. The ratio depends on whether the category is search-driven or discovery-driven.
- Never add attribution across platforms. Each platform tends to over-credit itself; summing their ROAS produces a fictitious total efficiency.
3. Account Structure
Three-level naming; "Campaign 1 / Campaign 2" is prohibited:
Campaign: {market}_{channel}_{objective}_{month}
Ad Set : {audience}_{placement}_{bid strategy}
Ad : {selling point}_{format}_{version}Four keyword groups: brand terms, category terms, competitor terms, and use-case terms. Competitor-term campaigns must clear legal first (trademark use and comparative-advertising boundaries).
4. Creative Testing: 3×3 Matrix
| Selling point \ Format | 15s short video | Static/carousel | UGC talking head |
|---|---|---|---|
| Pain point | A1 | A2 | A3 |
| Outcome | B1 | B2 | B3 |
| Trust | C1 | C2 | C3 |
Kill rule (empirical; calibrate by category): once a creative reaches the minimum sample, stop it if CTR is below 50% of the account median and CVR shows no advantage; keep 2–3 winners for scaling.
5. Budget Ladder
| Stage | Budget share | Goal | Scaling rule |
|---|---|---|---|
| Testing | 20% | Find creatives and audiences | Judge only after each creative reaches the minimum sample |
| Scaling | 50% | Amplify winning combinations | Increase 20%–30% at a time; observe 2–3 days |
| Steady state | 30% | Protect ROI, explore new audiences | Increase only when ROI is on target; cut back after 2 consecutive days below the red line |
6. Core Metrics and Three-Tier Thresholds
The following are empirical thresholds. They must be calibrated by category, AOV, and market, and must not be cited as industry standards.
| Metric | Watch | Warning | Circuit breaker |
|---|---|---|---|
| CTR | Above account median | 30% below median | 50% below median |
| CPC | Within target range | 20% above target | 50% above target |
| CVR | On target | 30% below target | 50% below target |
| CAC | ≤ target | 100%–130% of target | >130% of target |
| TACOS | ≤ target | 100%–130% of target | >130% of target |
| ROAS | ≥ target | 70%–100% of target | <70% of target |
7. Stop-Loss Rules
Stop if any condition is met:
- Zero conversions for 3 consecutive days;
- Cumulative spend reaches the per-creative test cap with no conversion;
- CAC exceeds the circuit-breaker line for 3 consecutive days;
- A brand-safety/compliance incident occurs (infringing creative, false claims).
Within 48 hours of stopping, produce a "failure review" with one conclusion each for creative, audience, bid, and landing page.
8. Search Growth
- Brand defense: own the top position for brand terms to prevent competitor interception;
- Category capture: prioritize by "search volume × conversion rate × competition";
- Competitor terms: run only in markets where legal permits; creative must not disparage;
- SEO/ASO synergy: feed high-converting terms discovered in paid media back into the DTC website content and App Store keywords.
9. Data and Attribution Interface
- UTM standard: all five elements—
utm_source/medium/campaign/content/term; - Pixels and events: unified event names for purchase, add-to-cart, sign-up, and subscription;
- Align with 05.21: priority order is "payment facts > platform reports > modeled attribution."
10. Budget Model: Work Backwards from the Revenue Target
1. Top-down: the marketing rate is the ceiling
Monthly media budget cap = monthly target units × per-unit marketing cost
Per-unit marketing cost = MSRP × marketing rate (14% in this library's basis)For AstroNext (MSRP USD 399, 14% marketing rate):
| Item | Calculation | Result |
|---|---|---|
| Per-unit marketing cost | 399 × 14% | USD 55.86 |
| 500 units / month | 500 × 55.86 | USD 27,930 / month |
| 1,000 units / month | 1,000 × 55.86 | USD 55,860 / month |
This rate is the same assumption used in 07.06 marketing 14% and 00.02 CAC <14% MSRP. It is not an industry benchmark — it is this master case's assumption; recompute for any other category.
2. Bottom-up: CPA is the brake
Affordable CPA = per-unit marketing cost = USD 55.86Once actual CPA exceeds this, every extra sale loses more money. This is the hardest line on the buying side and more intuitive than ROAS, which still needs a margin conversion.
3. Allocation: three ladder tiers × three platforms
First split by the ladder in §5 (testing 20% / scaling 50% / steady state 30%), then split by intent strength:
| Stage | Goal | Budget share | Platform emphasis |
|---|---|---|---|
| Testing | Find creatives and audiences | 20% | TikTok + Meta for fast creative tests, Google for search terms |
| Scaling | Amplify winners | 50% | Push proven assets into Meta lookalikes and PMax |
| Steady state | Protect ROI, open new audiences | 30% | Search captures brand/category demand |
At 500 units/month (USD 27,930): testing 5,586 ｜ scaling 13,965 ｜ steady state 8,379.
Discipline: the above is a calculation structure, not a recommended split. Real allocation must follow whichever platform beats the CPA ceiling; cut the rest regardless of platform prestige.
4. Minimum viable cold-start budget
Minimum test budget ≈ target CPA × conversions needed per creative per platform × number of creativesThe conversion requirement depends on the learning period (see §2). If the budget cannot get a single ad set past learning, open fewer platforms and run fewer creatives rather than spreading thin — data from an ad set that never left learning is pure noise.
11. Creative Specs and Review Pitfalls
1. Specs (follow current platform documentation)
| Platform | Primary ratios | Duration | Key constraint |
|---|---|---|---|
| Meta | 4:5 (feed) / 9:16 (Reels, Stories) | 6~15s to scale, 30s to explain | Benefit in the first 3 seconds; always subtitle (silent playback) |
| TikTok | 9:16 | 9~21s | Native feel first; anything ad-like gets swiped |
| Google PMax | 1:1 / 1.91:1 / 4:5 multi-size | 6~30s | Needs matching headlines, descriptions and logo asset group |
| YouTube in-stream | 16:9 | 6s non-skippable / 15~30s skippable | The first 5 seconds decide the skip |
2. Frequent rejection reasons
| Category | Typical issue | Mitigation |
|---|---|---|
| Efficacy claims | "cures", "permanently", "guaranteed" | Rewrite as verifiable functional description |
| Before/after | Weight-loss, beauty, cleaning comparisons | Replace with process demonstration |
| Trademark & comparison | Using competitor logos or disparaging by name | Route through Legal; see §3 competitor-term discipline |
| Price & promotion | Fake discounts, fabricated urgency | Promo price must match the site |
| Minors | Inducement in child-directed creative | Add age gate and parental-consent notices |
12. Attribution Windows and Definition Gaps
The three platforms have different default windows (see §2), which means "one order counted three times". Three rules:
- Unify the window before comparing: pull all platforms to the same window (7-day click recommended) before comparing ROAS; otherwise the conclusion does not hold.
- Payment facts win: platform reports are attribution; backend orders are facts. Priority: payment facts > platform reports > modelled attribution (see 05.21).
- Make unknown attribution explicit: when unknown attribution exceeds 40%, do not use "last-click ROAS" alone to cut brand or social — what gets cut is usually the upstream seeding.
13. Execution Checklist
[ ] Three-level account naming convention established
[ ] Four keyword groups completed; competitor terms cleared by legal
[ ] Three-platform differences aligned: auction logic, intent, attribution window, review strictness (§2)
[ ] At least one creative per cell of the 3×3 matrix
[ ] Minimum sample and kill rule written into the media plan
[ ] Budget ladder and scaling rules confirmed
[ ] Monthly budget derived from the revenue target: target units × per-unit marketing cost, not a round number (§10)
[ ] CPA ceiling written into the media plan: affordable CPA = per-unit marketing cost; stop above it
[ ] Minimum test budget computed: enough to exit learning, otherwise open fewer platforms
[ ] Creative specs produced per platform: 4:5 / 9:16 / 1:1, no re-cropped single video
[ ] Rejection risks self-checked: efficacy, before/after, trademark, fake promo, minors
[ ] Attribution window unified: aligned before any cross-platform comparison (7-day click recommended)
[ ] CAC/TACOS/ROAS thresholds calibrated by category
[ ] Stop-loss rules and review template ready
[ ] UTM five elements and event names aligned with 05.21
[ ] Weekly report covers creative, audience, bid, and landing page
14. Related Documents
- 04.00 Brand Marketing Overview
- 05.01 Online Global Playbook
- 05.19 Omnichannel Multi-Touch Attribution
- 05.21 Paid Media Attribution in Practice
- 02.03 Hardware + SaaS and LTV/Churn
Review record (2026-10-22): This document is a media-buying framework. All thresholds are empirical and must be calibrated by category; it does not constitute official platform guidance.