DTC Operator Playbook 2026 · 10 个高手的 X 内容浓缩

00 · TL;DR (10 个 operator 浓缩为 8 条)

Q1 2026 brand CPC 跌 9% YoY, 反转了 Q1 2025 +19% 的暴涨. 现在是用 brand search 做 defensive coverage 的低成本窗口, 但不要因为看着便宜就放预算 scale.
PMax inventory 41% 已经在 non-Shopping placements (video alone 6%). 没视频素材就把这部分白送给 Google auto-generated 的烂内容. SparkO 必须出至少 1 条 15-20s 视频.
Optmyzr 2 万账户研究: partial pinning 全维度赢. CTR 11.9% vs 8.5%, CPA $13.68 vs $32.57. 只 pin 品牌名 H1, 其他全 unpin. Headlines <20 字符, descriptions 61-70 字符表现最好.
Brand 必须从 PMax 隔离. 这是 2026 已经共识的事 — Holiday, Plofker, Curry, Levy, Ryan 五个 operator 一致. 不隔离 = PMax 吃掉 brand 流量, 报表 ROAS 虚高, 你判断不出真增量.
金字塔顶是 CM3 dollars (不是 %, 不是 MER %). Plofker 2026 升级: 从 "优化 MER" 走向 "优化总贡献毛利美元". MER/aMER 当 sanity check, channel ROAS 当 diagnostic. Scoutlite 现阶段先建 MER + aMER 仪表盘, CM3 等成本数据打通再上.
PMax 出价小账户用 Max Conversion Value, 不上 tROAS. 月转化 < 30 单 tROAS 会勒死算法. Scoutlite 现在月 ~25-30 单, 正好在分水岭, 建议先 Max Conv Value 跑 4-6 周稳定后再上软 tROAS.
NCA Value mode (不是 Only). 用 Customer Match list 喂 Google 知道谁是老客, 设个 $15-25 bonus 让算法偏向新客. Only mode 会饿死小账户的 smart bidding feedback loop.
Asset group 按 commercial theme 分, 不按 SKU 不按 audience. Mike Ryan 数据: multi-asset-group themed builds 比 single asset group 和 feed-only 都好. Scoutlite 应该是 EDC Daily Carry / Outdoor / Safety 三个 themed groups.

01@bigalittlea · Tinuiti

Q1 2026 Tinuiti 基准数据 (Aaron Levy 解读)

Tinuiti 是美国最大独立 performance media 公司之一, 季度 benchmark 报告基于他们管理的数百亿美元广告花费, 是 DTC 圈最受重视的数据源. Aaron Levy 是这份报告的主要解读人. 2026 Q1 的关键转折点:

-9%

Brand CPC YoY

反转 Q1 2025 +19% 的暴涨. 当下是 brand search defensive coverage 的低成本窗口.

41%

PMax 非 Shopping 占比

从 30% (Q1 2025) 涨到 41%. PMax 正在变成 "Shopping 外壳的品牌视频频道".

6%

PMax Video 占比

从 3% 翻倍. 没视频素材, 这 6% 就白送给 Google 自动生成内容.

93%

Retailer 跑 PMax

2025 年底 PMax 已经不是可选项, 是 default. 控制能力反而是差异化关键.

-7%

PMax ROAS vs Std Shopping

PMax CPC 高 13%, ROAS 低 7%. PMax 更像增长工具, 不是效率工具.

14%

AI Max Search 平均提升

+14% conversions/value 在同 CPA. exact/phrase 重的账户能到 +27%. 但retail 被 Google 排除在 headline benchmark 外.

As more retailer spend flows into PMax, you can't cleanly separate brand search from non-brand and from upper-funnel media. PMax co-mingles all of them. Aaron Levy · 2026
Cheaper brand looks very efficient again on a last-click basis, but resist over-allocating; evaluate whether PMax and upper-funnel are driving incremental demand. Aaron Levy 解读 Q1 2026 brand CPC drop
Practitioners need to shift from granular keyword-level tactics toward system-level levers: audience signals, creative, budgets, and conversion modeling. Aaron Levy · 2026

这组数据有三个直接含义: (1) brand search 现在是布局好时机 — 拿便宜的 defensive coverage, 但不要因为 ROAS 看着漂亮就放预算 scale brand search; (2) PMax 已经把 41% 钱花在非 Shopping placements 上了, 没视频 = 把 6% 直接送给 Google 用 stock 素材自动剪; (3) AI Max for Search 在 retail 表现 Google 自己都不愿提, 暂时别开.

布局节奏: brand search budget 上限 $10-15/day (符合 $0.20-0.30 CPC × 24 clicks/day 的现状). PMax 不动预算, 但 4 周内必须出 SparkO 一条 15-20s 视频 asset, 不然继续被 algo 浪费 6%.

02@siliconvallaeys · Optmyzr

Optmyzr 2026 RSA 研究: 2 万账户的硬数据

Frederick Vallaeys (前 Google AdWords 产品负责人, 现 Optmyzr CEO) 在 2026 发布的 RSA 研究, 样本规模 ~20,000 账户, >100 万条 RSAs. 是目前最权威的 RSA 优化研究. 关键结论:

Pinning 策略表现对比

Pinning 策略	CTR	CPA	CVR	ROAS	结论
Partial Pinning (推荐)	11.9%	$13.68	10.55%	365	全维度赢
Full Pinning	8.5%	$32.57	~6.8%	~210	CPA 2.4 倍, CTR -29%
Zero Pinning	~10.1%	~$18.40	~8.9%	~290	没有 brand control 不可取

Copy 长度的最优区间

资产类型	最优长度	表现维度	Scoutlite 含义
Headline	< 20 字符	CPA / CTR / CVR 全部最优	"Scoutlite Official Store" 22字符已经超了, 改 "Scoutlite Direct" 16字符
Description	61-70 字符	CTR / CVR / ROAS / Conv per Impr 全部最优	当前给的几条 description 都 130+ 字符, 全部需要重写

Conversion rates suffer when you pin. Pinning should be treated as a creative/brand control choice, not a primary optimization lever. Frederick Vallaeys · Optmyzr 2025
Mid-tier 'Average' Ad Strength ads actually showed the best CPA and CVR. Ad Strength is not a good predictor of performance. Optmyzr 2026 Study
Across 1.7M+ ads, RSAs got ~1.6× as many conversions per ad group as ad groups without RSAs. Optmyzr 2026 Study
Take out a PPC insurance policy with automation layering — Google handles bids/matching/RSA combinations; you handle anomaly detection, kill switches, and budget guardrails via rules or scripts. Frederick Vallaeys · 2026

这是当前最有 actionable 价值的研究. 我之前给的 brand search RSA 草稿 15 个 headlines 几乎全部超过 20 字符, descriptions 都在 130+ 字符. 严格按 Optmyzr 数据 rewrite 一遍, 预期 CTR 从 ~8% 拉到 ~11%+, CPA 砍半.

Ad Strength 不重要这个发现是 counter-intuitive 但被数据反复验证: Google 后台显示"Excellent"的 ad 反而 CPA 高于 "Average" 的. 别为了凑 Ad Strength 把文案做长.

本周 action: 把 strategy 1-pager 里那 15 条 headlines 全部 rewrite 到 ≤20 字符, 4 条 descriptions 重写到 61-70 字符. 只 pin 一条 brand name H1 (例如 "Scoutlite Direct"), 其他全 unpin. 预期: brand search CPA 从估算的 ~$5 跌到 $2-3.

03@TaylorHoliday · Common Thread Collective

Taylor Holiday: Profit-First Metric Hierarchy

核心框架: Plot, Pivot, Profit: 每天 plot 多个指标 (spend / revenue / CM / MER / customer mix); pivot 是最小有效干预 (而不是大改); profit 是日目标不是月结果. 已经完全离开 ROAS 作为主指标, 转向 contribution-margin dollars 和 incremental marginal return (iMR).

Taylor 是 CTC (Common Thread Collective, 累计帮品牌做到 $3B+ 营收) 创始人. 在 2024-2025 完成了从 "MER % 优化" 到 "CM3 dollars 优化" 的关键升级. 他的整套系统在 YouTube "2026 E-commerce Masterclass" 视频里讲得最完整 (有专门一份知识库在 reports/knowledge/taylor-holiday-profit-system.html).

Profit is the objective. Incrementality is the measurement. Creative volume is the mechanism. Taylor Holiday · CTC 2025 (最被引用的一句)
The metrics that matter: contribution margin and channel incremental marginal return, number of ads launched, cost per ad launch. Taylor Holiday · CTC 2025
Most problems are either volume problems or efficiency problems. Taylor Holiday · CTC Podcast 2025
Growth is strategy to the power of execution. Taylor Holiday · 2025
Build a backlog of ideas before you need them. Taylor Holiday · CTC 2025 (recurring)

Taylor 的"creative volume is the mechanism"这句话直接打中 Scoutlite 的关键瓶颈: 你账户结构其实没那么糟糕, 真正缺的是 足够多的 ad iteration. 一个 SparkO PMax asset group 里 15 个 headlines + 5 个 long headlines + 4 个 descriptions + 15 张图 + 5 条视频 — 你现在能交付的可能只有一半. PMax 算法吃不饱素材就只能反复 serve 同一组组合, fatigue 来得快.

"smallest action that fixes the problem"这个 lens 适合你. 不要见到数据下跌就加 campaign / 加预算 / 改结构. 先问: 这是 volume problem 还是 efficiency problem? Volume problem 加素材, efficiency problem 找漏点.

Taylor 的金字塔在 Scoutlite 简化版: (1) CM3 dollars (年底接入, 现在没有完整 cost breakdown), (2) MER + aMER (现在就能开始追), (3) channel ROAS 当 diagnostic, 出问题才看, 不当指挥棒.

04@CodyPlofker · ex-Jones Road / Plain

Cody Plofker: Brand Defense + Attribution Skepticism

核心框架: 两个支柱: (1) 归因结构性吵闹, 趋势是 incrementality + blended view; (2) 平台 over-credit lower-funnel/brand search, under-credit upper-funnel prospecting. 2024-2025 完成了从 "margin %" 到 "CM3 dollars" 的升级. 是 Northbeam 重度用户.

Cody 之前是 Jones Road Beauty 的 CMO, 现在做咨询 + Plain. 在 X 和 newsletter 上长期发声的核心议题是 不要相信平台 ROAS. 他对 brand search 和 PMax 关系的判断, 现在是 DTC 圈共识的基础.

Attribution overvalues low funnel and undervalues top of funnel. Cody Plofker · X (2024-2025 recurring)
PMax looks amazing when you let it eat all your branded search, but that doesn't mean it's actually growing your business. The platform will happily take credit for conversions that would have happened anyway. Cody Plofker · 2025 (paraphrased)
Make budget moves at the MER/CM3 level, sanity-check with Northbeam, and only then look at platform ROAS for creative and tactical decisions. Cody Plofker · 2025
If two brands have the same revenue, the one with higher contribution margin is usually healthier, even if its ROAS is lower. Cody Plofker framework synthesis · 2024

Plofker 的 ROAS 层级

层级	指标	用途	调整频率
顶	MER + CM3	决定整体预算大小	周 / 月
中	Northbeam modeled / incremental	判断渠道是否真增量	周
底	平台报 ROAS (Meta / Google)	素材和战术微调	日

Cody 的 "separate brand search, cap it, don't scale on its ROAS" 是这次 Scoutlite 重构的中心思想. Brand search 是 defensive line, 不是 growth engine. 同样的逻辑要求 PMax 排除 brand: 不 fence 就被 PMax 吃掉, 报 ROAS 虚高, 你判断不出是否真 acquire 新客.

$1.5-3.5K/月规模上完整 Northbeam (起步 ~$1K/月) 太奢侈, 内化的纪律才是关键: 决预算看 MER/aMER, 判增量看时间维度的 holdout (停渠道 2 周看全店是否等比例跌), 看平台 ROAS 只用来调素材.

本月 brand search 预算硬上限 $15/day, 即使 ROAS 显示 20x 也不加. 月度 review 看 Shopify aMER (新客 revenue / Google Ads cost) 而不是 Google 后台 ROAS. 季度做一次时间维度 holdout 测真增量.

05@AndrewJFaris · AJF Growth / ex-4x400

Andrew Faris: Setter-Hitter + Creative Iteration

核心框架: Creative testing 是 operational system, 不是 "throw stuff and see what sticks". Identify the core winning message from top spenders, then restate that message as differently as possible across hooks/visuals/formats. 2025年6月明确说自己"几乎不做 iterative ad creative changes" — 因为 Meta 把近似变体当同一条 ad 处理.

Figure out the core message in the ads that spend, then restate that message as differently as possible. Andrew Faris · AJF Podcast June 2025
Bad testing methodology is one of the biggest ways ecom brands have operational inefficiency and financial waste. Andrew Faris · X 2024-2025
Media buying is the setter, creative is the hitter. You need both but creative scores the point. Andrew Faris · Motion App event
Use AI for angle ideation, then systematically re-work your best formats with new angles to find fresh winners instead of constantly reinventing ad types. Andrew Faris · 2025

Setter-Hitter 在 Scoutlite 的映射

角色	是谁	职责	产出节奏
Setter	Skyler	写 brief, 决定 angle, 选择渠道结构, 看归因	周
Hitter	Higgsfield + Jimeng + GPT image gen (AI 工具)	每个 brief 出 10 个变体, 少数变体成为 winner	每个 brief 1-2 天
裁判	Meta/Google 算法	选出 winner 给量	自动

Andrew 的 "restate the message differently" 规则直接套到 PMax asset group: 不要做 20 个 SparkO clip-on hook 变体, 要每个 angle 做 5 个变体, 跨多个 angle:

(1) 磁吸 (magnetic clip), (2) 解放双手阅读 (hands-free reading), (3) 给孩子讲睡前故事 (kid bedtime), (4) 修车场景 (auto repair under the hood), (5) 露营/夜跑.

借力 Reddit research 里的 user vocabulary (reddit_research/insights/02_user_vocabulary.md). 5 个 angle × 5 个文案变体 = 25 条 headlines, 配上 PMax asset group 需要的 15 张图 + 5 条视频 (按 angle 分配). 工具: Higgsfield 是 Hitter, 你是 Setter.

06@BrettCurry · OMG Commerce

Brett Curry: YouTube + Google + RSA Pinning

核心框架: YouTube + Google 适合 "visually demonstrable" 产品. PMax 是 "powerful but misunderstood tool", 需要高质量素材 (视频+图) + 干净 conversion tracking + 结构好的 feed. RSA 大部分不 pin, 只 pin compliance.

YouTube must do all the work — hook fast, demonstrate, overcome objections, social proof, clear CTA. Brett Curry · OMG 2025
60-90 second video length is the sweet spot for conversion-focused YouTube; up to 3 minutes still viable. Brett Curry · 2025
Voiceover and messaging quality matter more than polished visuals. Brett Curry · 2025
Treat PMax as a way to scale and discover incrementality, not a set-and-forget black box. Brett Curry · OMG 2025

Brett 的 YouTube 视频结构 (60-90秒)

段	内容	时长	关键
Hook	抓眼球的问题或痛点	0-3s	无声也要看懂
Demo	产品在使用场景里运作	3-30s	SparkO 切4色温/4种安装方式
Objections	用户最常问的疑虑, 提前回答	30-50s	"30天无理由退" / "Kickstarter背书"
Social Proof	评论 / KS backers / KOL片段	50-70s	1-2个证据点
CTA	明确指令 + URL	70-90s	"Buy at scoutliteworld.com"

SparkO 是 "visually demonstrable" 的完美样本: magnetic clip, 4 色温, 4 种安装. Brett 这个 60-90s 结构直接适用. 但他建议的 YouTube 投放预算 ($500-1,000/day "ideal", $100/day 起步) 对 Scoutlite 现阶段不可行.

但视频本身仍要做 — 不投独立 YouTube campaign, 而是把 60-90s 视频作为 PMax asset, 它会 serve 在 PMax 的 6% 视频 inventory 里 (那块现在没视频就被 Google 自动剪).

本月可以做的最小可行视频: 1 条 60-90s SparkO demo (Skyler 自拍 or 找现有 KOL 视频剪辑), 上传到 PMax asset group. 后期再做 BlinkO/Vario 的版本. 不需要 high production value (Brett 强调 voiceover + messaging 比 visual polish 重要).

07@mikeryanretail · Smarter Ecommerce (smec)

Mike Ryan: PMax Structure + Signals

核心框架: PMax 不再可选, 跟算法合作而不是 recreate manual Shopping. Single-asset-group full builds 输给 feed-only 和 multi-asset-group builds. Multi-asset-group full builds 数据上赢, 当 asset 按主题分好. Search themes 和 audience signals 是 nudge 不是 filter.

Don't fight [the algorithm]. Instead, train your AI campaigns with better conversion tracking, audience signals, and clear objectives. Mike Ryan · LinkedIn 2025
Audience signals are signals, not filters; they inform the model but do not restrict who can see your ads. Overloading them often brings little incremental benefit. Mike Ryan · 2025
Seed each main PMax campaign with 10-30 carefully chosen themes that mirror your core categories and hero use cases, not every possible query. Mike Ryan · 2025 framework
This ain't your grandaddy's PMax! Mike Ryan · LinkedIn 2025 (on expanded PMax controls)

Mike Ryan 的 PMax 结构对比

结构	ROAS	AOV	适合
Feed-only (无 asset)	中	低	纯 Shopping 替代
Single-asset-group full	低	低	避免
Multi-asset-group themed	中-高	高	推荐

Scoutlite 的 3 个 themed asset groups (推荐)

Asset Group	产品	Landing Page	Search Themes (5-10)
EDC Daily Carry	SparkO + small Varios	/pages/buy-sparko	edc flashlight, pocket flashlight, magnetic clip light, hands-free reading light, work light nurse, photographer light
Outdoor / Camping	CUPO 系列 + Vario 大流明	/collections/headlamps	camping headlamp, hiking light, trail running light, outdoor adventure lamp, night ride bike light
Safety / Visibility	BlinkO + 反光配件	/products/blinko	safety light dog walking, runner safety light, kid bike safety, dual visibility light, 270 degree safety

Mike Ryan 的 multi-asset-group themed 数据非常 clean — 不要按 SKU 分 (那会变成 SKAG-style 过度细分), 不要按 audience 分 (audience signals 是 nudge 不是 filter, 分得再细 algo 也会扩散). 按 commercial intent / use case 分.

"Audience signals are signals, not filters" 是这章最重要的认知更新. 我之前给的策略文档建议 4 个 signals — 按 Mike 的逻辑应该是这 4 个而不是更多 (更多反而稀释).

现在 PMax 1117-SparkO 是 single asset group full build (per Mike 的数据是最差的结构之一). 重构到 3 个 themed asset groups 是本次 PMax refresh 的关键动作. 每个 asset group 给 15+ headlines, 5+ long headlines, 4 descriptions, 15+ images, 5+ videos (target).

08@SeanEcom · Ridge

Sean Frank: Brand-First + Margin Moats

核心框架: 三支柱: brand-first 不 channel-first; 不信平台归因; "few things done extremely well". Margin 是 brand 的 fuel: 主张 80% gross margin (COGS < 20%) 才能买得起 attention. DTC moat 来自 brand preference, 不是 media arbitrage.

DTC is no longer 'Facebook ads plus a product'; it's building a system that buys and manufactures attention across channels. Sean Frank · 2024-2025
A brand is an attention machine; performance spend is how you buy attention. Sean Frank · X recurring
If you do not model incrementality yourself, you will believe their numbers and over-invest or under-invest in the wrong places. Sean Frank · Operators Podcast
Even with new platforms, I still expect Meta to be the biggest bucket of spend, Google second, then a long tail. Sean Frank · 2025
Many influencer platforms are expensive UI layers that add friction. Sean Frank · Operators Podcast (Ridge has 8+ people doing creator outreach manually)

Sean 的 margin 论点对 Scoutlite 关键: SparkO 售价 $45.99, gross margin 估计 50-65% (含 COGS + 运费), 低于他认为的 healthy 区间 (80%+). 这意味着你的素材成本必须严格控制, bundle 提 AOV 是非选项. 你已经在做了 (Scoutlite 项目里有 bundle widget), 这点 align.

"Few things done extremely well" 这句对 Google Ads 重构有直接指导: 不要同时跑 Brand Search + Multi-PMax + Demand Gen + YouTube + Display. 选 1 个 PMax (3 themed asset groups) + 1 个 Brand Search, 把这两个做到完美.

本次重构后, Google Ads 账户只有 2 个 active campaign — 这本身就是 Sean Frank 思路的实例. Demand Gen / 独立 YouTube / Discovery 全部 defer 到 spend 翻倍后再考虑.

09@nik_sharma · Sharma Brands

Nik Sharma: Performance Branding

核心框架: Performance branding — performance 和 brand 不是 either/or 而是 orchestrated together. 重度倡导 advertorial-style / "Trojan horse" landing pages (看着像内容, 实际是 direct response). 主张 lean teams + 早期 "athlete" generalists 后期再叠 specialists.

Performance AND brand, orchestrated together. Nik Sharma · X recurring 2024-2025
Brand work should make performance cheaper over time, while performance channels validate and distribute the brand story at scale. Nik Sharma · 2025
Treat the audience like a product on subscription — a strong audience makes paid performance more efficient. Nik Sharma · 2025
Build multiple targeted landing pages rather than only optimizing a generic homepage. Nik Sharma · 2024-2025
Every touchpoint should both build brand and move a key metric. Nik Sharma · "performance branding" doctrine

"Brand work should make performance cheaper over time"是这章可量化的核心. 这正好可以用 Branded Search Lift 来追: Scoutlite 跑非品牌 PMax → 用户看到 → 几天后搜 "scoutlite" → brand search volume 涨 → 即使 Google Ads 没归因到, 真增量发生了.

Nik 的"multiple targeted landing pages"规则正好对应 Scoutlite 现在的 LP 结构: SparkO 有专属 LP (buy-sparko), BlinkO 也有专属页面. 已经在做了. 下一步是 advertorial-style 长内容页 (像 Mighty Patch 那种 "I tried 47 magnetic flashlights and only one worked").

月度 KPI 加一个 Branded Search Lift: 从 GSC 拉 "scoutlite/sparko/blinko" 月度搜索量, 跟 Google Ads 非品牌 PMax spend 做相关. 如果 non-brand spend 涨, brand search 也涨, 说明 PMax 在帮品牌建认知, 不只是收割.

10@bigalittlea · Tinuiti / Optmyzr

Aaron Levy: System-Level Levers

核心框架: Google Ads 是三个业务的合体: (1) upper-funnel 视频/Display, (2) "classic" text/Shopping search, (3) 越来越自动化的 PMax 黑箱. 实际操作要从 keyword-level tactics 升级到 system-level levers: audience signals, creative, budgets, conversion modeling.

Aaron 已经从 Tinuiti SEM Group Director 转去 Optmyzr 做 evangelist 角色. 他对 Tinuiti benchmark 报告的解读是 DTC operator 圈第一手数据源. Q1 2026 数据细节在 Chapter 01, 这里聚焦他的 system-level lever 框架.

Aaron 的 4 个 system-level levers (代替 keyword-level tactics)

Lever	含义	Scoutlite 操作
1. Audience Signals	用 Customer Match + in-market segments 喂 algorithm 知道好客户长啥样	上传 Shopify customer list (24个月), Scout Rewards Beam+Blaze 高 LTV 客户, Kickstarter backers
2. Creative	给 algorithm 足够多的素材去探索 inventory	每个 asset group: 15+ headlines, 4+ descriptions, 15+ images, 5+ videos
3. Budgets	大方向预算分配 (campaign 间 / channel 间), 不是 daily micro-tweaks	Brand Search $10-15/day cap, PMax 保持当前预算让其重学
4. Conversion Modeling	Enhanced Conv + Customer Match + new_customer parameter	Shopify Google & YouTube app 自动配, 加 Customer Match 月度刷新

Aaron 的洞察很重要: 2026 的 Google Ads 不再是 keyword-level 游戏. 你不能再像 2018 年那样调 bid adjustment by device or by hour of day. PMax 把这些 micro-control 抢走了. 留给你的杠杆就是上面 4 个 system-level lever — 但每个都很 powerful.

这跟 Brett Curry 的 "PMax is powerful but misunderstood" 一致, 也跟 Mike Ryan 的 "don't fight the algorithm" 一致. 三位 operator 不约而同收敛到同一个结论.

你的 Brand Search + PMax 重构, 4 个 lever 全部都在动: (1) 加 Customer Match signal, (2) refresh PMax asset group, (3) 重新分配预算 (Brand Search 独立), (4) New Customer Value mode + Enhanced Conv. 完成后你就把 system-level levers 全调对了.

11@HighBrowMarketr · ex-True Classic

Bryan Cano: TPS Lifecycle Structure

核心框架: "Stair-step" launch sequence: 用 phase 控风险和学习. Variant types 按 product category / hero vs long-tail / lifecycle (Testing-Profit-Scale) / offer 分. 避免 SKAG-style 过度细分 — 别为每个 micro audience 建一个 asset group.

Maximum creative variety inside the asset group: 15+ headlines, 5+ descriptions, 10-20 mixed image formats. Bryan Cano · True Classic 2024-2025
Rotate losers out based on asset-level performance instead of constantly adding new asset groups. Bryan Cano · 2025
Don't build a new asset group for every micro audience; fold multiple audiences into a handful of strong groups to preserve volume. Bryan Cano · 2025
Cut the fat, double-down on winners. Bryan Cano · True Classic 2024 rebuild mantra

TPS (Testing-Profit-Scale) Lifecycle Structure

阶段	放谁	预算	更换频率
Testing PMax	新 SKU + 新 creative	小 (15-25% 总预算)	频繁 swap, 7-14 天判断
Profit / Evergreen PMax	已验证 SKU, CM 达标	中 (40-50%)	稳定, 30-60 天大调
Scale PMax	top 5-20 SKUs, 最赚钱的	大 (30-45%)	少改结构, 主要调预算

TPS Lifecycle 对 Scoutlite 2026 roadmap 非常重要: 6月 headlamp 新品 / 9月 Arc lighter / 10月女性安全 flashlight + SparkO upgrade — 每个新品启动都应该走 Testing PMax 模式, 跑到 CM 达标后升级到 Evergreen, 不是另开 campaign.

当前 Scoutlite 只有 1 个 active PMax, 还没到需要 TPS 三层的规模. 但当 6 月 headlamp 上线时, 应该用 Testing PMax 的模式: 单独建一个 campaign 跑新品, 小预算, 频繁 swap asset, 30-60 天后评估是不是 graduate 到现有 main PMax.

6 月 headlamp launch 的 Google Ads 结构 (提前 plan): 建一个新 PMax campaign "Testing - 0625 Headlamp", $30/day 起步, 单 asset group 聚焦新品. 30天后看 CM 是否达标 — 达标 graduate 到现有 main PMax 的 "Outdoor / Camping" asset group; 不达标暂停.

12@siliconvallaeys · Optmyzr

Frederick Vallaeys: Automation Layering

核心框架: 数据驱动的 RSA/PMax 研究 (硬数据在 Chapter 02) + automation layering framework: 在 Google 的自动化上面再叠你自己的 automation, 防止算法跑偏. "Take out a PPC insurance policy with automation layering."

Take out a PPC insurance policy with automation layering. Frederick Vallaeys · Optmyzr 2025
Google's own internal data: ~20% lift when advertisers enable Smart Bidding together with broad match keywords. Frederick Vallaeys citing Google
Mid-tier 'Average' Ad Strength ads actually showed the best CPA and CVR. Optmyzr 2026 RSA Study

Automation Layering 分工

归 Google 管	归你管
Bid / match / RSA combination optimization	Anomaly detection (异常告警)
Audience expansion based on signals	Kill switches (满足条件自动暂停)
Search Themes 信号扩散	Budget guardrails (预算上下限)
Asset combination experimentation	规则 / 脚本 (Google Ads Scripts)

这是 Task #6 (weekly automated reporting) 的理论基础. 我会写的 weekly script 就是 automation layering 的一部分: 拉数据, 检测 anomaly (CPC 月环比涨>30%, ROAS 掉>25%, asset fatigue), flag 给你看. Google 在底下跑算法, 我在上面盯着不让它跑偏.

更激进的 layering 是写 Google Ads Scripts (账户里跑的 JavaScript): 比如某 keyword 连续 7 天 CTR < 0.5% 就自动 pause. 现在阶段不用 — 你账户太小, anomaly 我手工 flag 就够.

Task #6 weekly script 必做的检查项: (1) brand search Search IS < 85% 报警, (2) PMax 任一 asset group CPC 周环比 +25% 报警, (3) 总账户 ROAS 周环比 -20% 报警, (4) 任一 ad group 0 impressions 连续 3 天报警 (说明 ad 被拒). 这 4 个 anomaly detector 就是 Scoutlite 当前阶段的 automation layer.

13

共识 + 真分歧 (Synthesis Across Operators)

已经共识的事 (3+ operators 一致)

1. Brand 必须从 PMax 隔离

Holiday + Plofker + Curry + Levy + Ryan 一致. 不隔离 = PMax 吃 brand 流量 = 报表 ROAS 虚高 = 你判断不出真增量. 2026 已经是默认操作.

2. CM dollars > ROAS as operating metric

Holiday + Plofker + Frank + Sharma 收敛到 CM 美元 (不是 %, 不是 MER %). ROAS 是 diagnostic. MER 是 business 级别 sanity check.

3. Audience signals 是 nudge 不是 filter

Ryan + Cano + 小预算/PMax setup consensus 一致. 一个 asset group 里 stack 几个 signals 比拆成多 asset groups 好. 也不要 stack 太多 signals.

4. Multi-asset-group themed 结构最好

Ryan 数据 (smec analyses) + Cano True Classic 经验一致. Themes 按 product role 分, 不按 audience 分.

5. Partial RSA pinning 赢

Vallaeys (硬数据, 2万账户) + Curry (理念) 一致. 只 pin compliance-critical messaging.

6. AI / PMax 放大你给它的信号

Curry + Ryan + Vallaeys + AI Max consensus 收敛: 好 feed + 好 conversion tracking + 好 audience signals = AI 工作. Sloppy inputs = 烧钱的学习实验.

2026 平台变化 (everyone is talking about)

变化	含义	Scoutlite action
PMax inventory 41% non-Shopping	PMax 在变 brand/video 频道	必须出视频 asset
Brand CPCs -9% YoY Q1 2026	defensive coverage 便宜期	Brand Search 现在布局
AI Max for Search 出 beta	retail 表现不明确	defer 3-6 个月
PMax Channel Performance Reports	能看 surface mix 了	月度 review 加这项
PMax negative keyword 上限 10,000	从 100 升到 10,000	本次 brand exclusion 直接用
Customer Match 最低 size 放松	不再有 strict minimum	立即上传 Shopify list

真分歧 (operators 还在吵)

NCA Value vs Only mode

Consensus: Value 是 default. 真分歧在 bonus 设多少. 一派设到等于 LTV uplift (AOV × 1.5-2), 一派用 "+$0.01" 小 nudge.

平台 ROAS 信不信

Frank + Plofker 说是 noise. Ryan + Curry 说 directional 可用. Holiday 中间: ROAS 是诊断, 不是优化目标.

归因栈选哪个

Northbeam (channel-level lift), Triple Whale (操作型 dashboard), Haus (geo/time incrementality). $1M 内 — Triple Whale 够; $1-3M — 叠 Haus; $3M+ — 标准 geo testing.

Brand Search 预算: cap 还是 scale

Plofker + Levy: 硬 cap, brand 是 defensive 不是 growth. Curry + Ryan: 不artificially cap 如果竞品在抢.

Ad Strength 追不追

Vallaeys 数据: 别追. Google: 推荐功能跟 ad strength 绑定. 共识: 别为 ad strength 优化, 但 Poor 要修.

素材迭代频率

Faris (June 2025): "几乎不做 iterative hook swap" 因为 Meta 把近似变体当同一条. Holiday: ad volume 是 mechanism. 共识: 迭代message, 不迭代 hook tweaks.

14

Scoutlite Action Plan: 8 个 Tactical Moves

从 10 个 operator 的 framework 浓缩出 8 个针对 Scoutlite $1.5-3.5K/月规模的具体动作. 每条带数据依据和负责 operator.

01

建 1 个 PMax 含 3 个 themed asset groups, 不是 3 个 campaign 也不是 6 个 asset groups

Asset Group 1: EDC Daily Carry (SparkO + small Varios → /pages/buy-sparko). Asset Group 2: Outdoor / Camping (CUPO + 大 Varios → /collections/headlamps). Asset Group 3: Safety / Visibility (BlinkO → /products/blinko). Bundle 进 hero 所在 group, 用 custom label 过滤 feed.

Source: Mike Ryan multi-asset-group thesis + Bryan Cano variant-type structure

02

Brand Search 硬 cap $10-15/day, partial-pinned RSAs

Q1 2026 brand CPC -9% YoY 是低成本布局窗口, 但不要因为 ROAS 漂亮就 scale. 只 pin "Scoutlite" 在 H1 位置 1 一条 RSA, 其他全 unpin 让 Google 测组合. PMax 用 brand exclusion list 把 brand 流量 fence 给 Brand Search.

Source: Vallaeys partial-pinning data + Plofker brand-cap rule

03

PMax bid strategy 头 30 天用 Maximize Conversion Value, 不上 tROAS

月转化 20-30 单时 tROAS 会勒死算法. 4-6 周后稳定再加 soft tROAS (target 60-70% of observed average), 然后逐步收紧.

Source: small-budget DTC consensus + Mike Ryan algorithm-feeding

04

开 New Customer Acquisition Value mode (+$15-25 bonus)

Scoutlite 10% repeat purchase 目标意味着老客仍有意义. Value 模式 (不是 Only) 保留 easy-win 转化, 同时偏向新客. $15-25 = SparkO AOV 的 30-50%, 在 moderate 区间不扭曲报表.

Source: NCA consensus + Plofker/Faris circle standard

05

上传 3 个 Customer Match list 作为 audience signal

List 1: 全 Shopify 24个月内客户. List 2: Smile.io Beam+Blaze 高 LTV. List 3: SparkO Kickstarter backers (7000+人). 每 asset group 加 1-2 个 in-market segment (Outdoor / Auto / Photography).

Source: 2026 PMax consensus + Sean Frank first-party data emphasis

06

4 周内出 SparkO 第一条 15-20s 视频 asset

PMax 41% spend 在 non-Shopping placements, video 占 6%. 没视频 = 让 Google 用 stock 素材自动剪. Brett Curry 结构: hook 0-3s, demo 3-30s, objections, social proof, CTA. 不需要 high production, voiceover + messaging 比 visual polish 重要.

Source: Tinuiti Q1 2026 + Brett Curry creative blueprint

07

所有 RSA + PMax 文案 rewrite: headline <20 字符, description 61-70 字符

Optmyzr 2万账户研究最优区间. 我之前 strategy 1-pager 里给的 15 条 headlines 几乎全部超 20 字符, descriptions 都在 130+ 字符. 全部重写, 预期 CTR +30%, CPA -50%. 文案 hook 来自 reddit_research/insights/02_user_vocabulary.md 真实用户词.

Source: Vallaeys Optmyzr 2026 RSA Study (20K accounts)

08

建立 weekly review 三指标: MER + aMER + CM3

MER (总 Shopify revenue / 总 Meta+Google spend) → 业务级别 sanity check. aMER (新客 revenue / 总 ad spend) → 真增量信号. CM3 (revenue - COGS - 运费 - fees - ad spend) → 利润 dollars. 不要看 platform ROAS 做决策, 只用作 diagnostic.

Source: Taylor Holiday + Plofker metric hierarchy

15

Quote Bank · 引文库 (按主题分组)

40+ 条 verbatim quotes 浓缩为可扫读. 按主题分组, 用来快速 lookup 任何争论的"圈内权威说法". 完整数据源在 raw digest.

利润优先 (Profit-First Measurement)

Profit is the objective. Incrementality is the measurement. Creative volume is the mechanism.Taylor Holiday · CTC 2025
The metrics that matter: contribution margin and channel incremental marginal return, number of ads launched, cost per ad launch.Taylor Holiday · CTC 2025
Most problems are either volume problems or efficiency problems.Taylor Holiday · CTC Podcast 2025
If two brands have the same revenue, the one with higher contribution margin is usually healthier, even if its ROAS is lower.Cody Plofker framework · 2024
If you do not model incrementality yourself, you will believe their numbers and over-invest or under-invest in the wrong places.Sean Frank · Operators Podcast
Make budget moves at the MER/CM3 level, sanity-check with Northbeam, and only then look at platform ROAS.Cody Plofker · 2025

Brand Search vs PMax Cannibalization

Optmyzr saw keyword overlap in 91% of accounts between PMax and Search.Frederick Vallaeys · Optmyzr 2025
Real-world data shows that when performance between overlapping PMax and Search traffic diverges by more than 10%, the Search campaign is more often the better performer.Optmyzr 2025
Attribution overvalues low funnel and undervalues top of funnel.Cody Plofker · X recurring
PMax looks amazing when you let it eat all your branded search, but that doesn't mean it's actually growing your business.Cody Plofker · 2025
Cheaper brand looks very efficient again on a last-click basis, but resist over-allocating; evaluate whether PMax and upper-funnel are driving incremental demand.Aaron Levy · Q1 2026

PMax Structure

This ain't your grandaddy's PMax!Mike Ryan · LinkedIn 2025
Don't fight [the algorithm]. Instead, train your AI campaigns with better conversion tracking, audience signals, and clear objectives.Mike Ryan · 2025
Audience signals are signals, not filters; they inform the model but do not restrict who can see your ads.Mike Ryan · recurring
Seed each main PMax campaign with 10-30 carefully chosen themes that mirror your core categories and hero use cases, not every possible query.Mike Ryan · 2025
Maximum creative variety inside the asset group: 15+ headlines, 5+ descriptions, 10-20 mixed image formats.Bryan Cano · 2024-2025
Cut the fat, double-down on winners.Bryan Cano · True Classic 2024 rebuild
Treat PMax as a way to scale and discover incrementality, not a set-and-forget black box.Brett Curry · OMG 2025

RSA + Ad Copy

Conversion rates suffer when you pin. Pinning should be treated as a creative/brand control choice, not a primary optimization lever.Frederick Vallaeys · 2025
Partial pinning: CTR ~11.9%, CPA ~$13.68, CVR ~10.55%, ROAS ~365. Full pinning: CTR ~8.5%, CPA ~$32.57.Optmyzr 2026 Study, 20K accounts
Headlines under 20 characters show better CPA, CTR, and CVR.Optmyzr 2026
Descriptions in the 61-70 character range perform best across CTR, CVR, ROAS, and conversions per impression.Optmyzr 2026
Mid-tier 'Average' Ad Strength ads actually showed the best CPA and CVR.Optmyzr 2026

Creative Iteration

Figure out the core message in the ads that spend, then restate that message as differently as possible.Andrew Faris · AJF Podcast June 2025
Bad testing methodology is one of the biggest ways ecom brands have operational inefficiency and financial waste.Andrew Faris · X 2024-2025
I'm making almost no true iterative ad creative changes.Andrew Faris · June 2025 podcast (old-style hook swaps; now broader variation)
Use AI for angle ideation, then systematically re-work your best formats with new angles.Andrew Faris · 2025

Brand + Attention

DTC is no longer 'Facebook ads plus a product'; it's building a system that buys and manufactures attention across channels.Sean Frank · 2024-2025
A brand is an attention machine; performance spend is how you buy attention.Sean Frank
Performance AND brand, orchestrated together.Nik Sharma · X recurring
Brand work should make performance cheaper over time.Nik Sharma · 2025
Every touchpoint should both build brand and move a key metric.Nik Sharma · performance branding doctrine

Platform Meta-Environment (Q1 2026)

Q1 2026: advertiser brand keyword CPCs on Google declined about 9% year over year.Tinuiti Q1 2026 Benchmark
By Q1 2026, video and other non-Shopping placements accounted for 41% of PMax spend, with video alone at 6%.Tinuiti Q1 2026
By end of Q1 2025, 93% of retailers were running PMax.Tinuiti Q1 2025
PMax CPCs were 13% higher and overall ROAS from PMax was about 7% lower than ROAS from standard Shopping.Tinuiti Q1 2025
Google's own internal study indicated about a 20% lift when advertisers enable Smart Bidding together with broad match keywords.Vallaeys citing Google
Advertisers activating AI Max in Search typically see around 14% more conversions or conversion value; campaigns heavy on exact/phrase can see ~27% uplift.Google AI Max launch data

YouTube + Demand Gen

YouTube must do all the work — hook fast, demonstrate, overcome objections, social proof, clear CTA.Brett Curry · 2025
60-90 second video length is the sweet spot for conversion-focused YouTube.Brett Curry · 2025
Voiceover and messaging quality matter more than polished visuals.Brett Curry · 2025
YouTube allowable CAC sweet spot: $50-$150.Brett Curry · 2025

Attribution Stack

Use attribution/MTA to decide which ad to pause today. Use incrementality tests to decide whether to keep spending millions on a channel this quarter. Use MMM (calibrated by incrementality) to plan next quarter's channel mix.2026 attribution-stack consensus
Anchor internal debates around tests, not dashboards.2026 attribution consensus

Lean Teams + Operator Economy

Bias toward 'athlete' generalists who can own strategy, creative direction, and basic execution across multiple channels, then layer in specialists as channels prove out.Nik Sharma · 2025
Pick three levers you'll be world-class at and ignore the rest until those are maxed.Sean Frank synthesis
Many influencer platforms are expensive UI layers that add friction.Sean Frank · Operators Podcast (Ridge has 8+ people doing creator outreach manually)