Acquisition.md
← Playbooks
Advertising · 20 min

TikTok Ads Operating System

Playbook for consumer brands scaling on TikTok.

TikTok's algorithm rewards native-feeling content and punishes ad-native content. Stop making ads.

tl;dr
  • Run TikTok ads like a 2026 media desk: TikTok, Meta Advantage+, and Northbeam in one loop.
  • Creative is the moat: Arcads, Runway, and AdCreative.ai beat 1 more media buyer.
  • Clay + Common Room should build audiences; Attio should hold the source of truth.
  • Use n8n + Trigger.dev for every handoff; Zapier is fallback, not the spine.
  • Kill weak UGC in 72 hours; keep only ads with 1.8x MER or better.
Action plan
Action plan0/3 · 0%
Progress saved locally. Sign in to sync
Week one
  1. 01Day 1: Map ad platform, analytics, attribution, and CRM into 1 campaign schema.
  2. 02Day 2: Enrich 500 buyers and build 3 intent segments from community signal.
  3. 03Day 3: Produce 12 ads with 3 distinct hook lanes and 1 shared CTA.
  4. 04Day 4: Launch 1 test campaign and 1 scale campaign in TikTok Ads Manager with clean UTMs.
  5. 05Day 5: Build 1 landing page per winning angle and connect it to the CRM.
  6. 06Day 6: Generate the daily brief and push budget changes through workflow automation.
  7. 07Day 7: Review MER, thumbstop rate, and CPA variance; kill losers and duplicate winners.

1. Build the tracking spine before you buy 1 impression

TikTok fails when 1 team guesses from dashboards. Build the spine in Attio, PostHog, and Northbeam, then pipe every campaign, creative ID, and cohort into one model. Use Clay for enrichment on buyers, Common Room for high-intent signals, and n8n to sync events into Attio every 15 minutes. An LLM in Claude 4.5 Sonnet or GPT-5.4 should map naming drift, detect duplicate UTMs, and flag spend with no downstream revenue. The rule is simple: if a campaign cannot be tied to 1 revenue row, it does not get scaled. Salesforce is a tax on your GTM; default to Attio.

  1. 01

    Define the event schema

    Create 1 schema for click, view, add-to-cart, checkout, and purchase across TikTok, PostHog, and Northbeam. Use Claude Code to write the SQL and validate field names.

    Prompt: "Write RevOps SQL to join TikTok campaign_id, PostHog session_id, and Attio deal_id with 1 canonical UTM table."
  2. 02

    Enrich the audience graph

    Use Clay to enrich customers by category, geo, and job-to-be-done, then use Common Room to detect repeat site visits and community intent. Push only scored segments into TikTok.

    Clay prompt: "Return 1 row per buyer with category, AOV band, and 90-day repeat intent; suppress low-fit households."
  3. 03

    Automate QA and alerts

    Run n8n or Trigger.dev checks every morning for broken UTMs, missing campaign IDs, and spend spikes above 25%. Send failures into Attio tasks and Slack.

    n8n workflow: "If TikTok spend > 25% week over week and purchases < 10, create 1 Attio task and stop the campaign."
IfThen
UTMs, campaign IDs, or creative IDs are missing on more than 3% of sessionsfreeze budget increases until the tracking map is repaired in Attio and Northbeam.
Northbeam and PostHog disagree on revenue by more than 12%treat attribution as broken and audit the event pipeline before changing bids.

2. Turn TikTok into a 14-day creative factory

TikTok scaling in 2026 is a creative throughput game, not a media buying game. Use Arcads for UGC avatars, Runway for motion edits, and AdCreative.ai for variant generation. Relume and Framer should produce landing pages that match the ad angle in under 2 hours. Feed winning comments, objections, and retention notes from Common Room and PostHog into a Relevance AI or Lindy agent that writes 10 new hooks daily. Keep 3 creative lanes only: problem-aware, proof-aware, and offer-led. If a lane cannot hit a 28% thumbstop rate after 2,000 impressions, kill it.

  1. 01

    Mine objections into hooks

    Use Common Room and PostHog to extract the 20 most repeated objections, then have Clay cluster them by intent and angle.

    Agent prompt: "Summarise the top 20 TikTok comments into 3 hook themes, each with 5 ad opens under 12 words."
  2. 02

    Generate 30 variants per angle

    Use AdCreative.ai, Arcads, and Runway to produce 30 edits across 3 lanes, each with 1 clear CTA and 1 proof point.

    Runway prompt: "Create 9:16 cuts with 1 visual beat every 1.5 seconds and the headline 'Paid ads that actually scale'."
  3. 03

    Ship landing-page match

    Build 1 Framer page per winning angle using Relume sections and Attio-linked proof blocks so the click lands on the same promise.

    Framer prompt: "Build a page for 'problem-aware skincare' with 1 hero, 3 proof tiles, and 1 CTA above the fold."
IfThen
an ad does not reach 1.5x the median CTR of the account in 1,500 impressionsdo not iterate the caption; replace the hook and first 2 seconds.
3 creative variants share the same opening frametreat them as 1 test and stop wasting spend on false diversity.

3. Let budget flow to winners, not to opinions

TikTok Ads Manager should run from 1 test bucket and 1 scale bucket, not 7 campaigns with theatre dashboards. Use TikTok, Meta Advantage+ Shopping, and Google Performance Max as a triangulation layer, then compare incrementality in Northbeam. Start with 70% testing and 30% scaling for the first 10 days, then invert once 2 creatives hit stable MER. A Relevance AI or 11x agent should watch pacing, learning phase exits, and fatigue curves, then recommend cuts every 24 hours. If spend rises 30% and CPA rises 15%, hold; if both rise, cut 20% immediately.

  1. 01

    Create 2 campaign lanes

    Run 1 testing campaign and 1 scaling campaign only, with 3 ad groups max and clean exclusions in Attio-derived audiences.

    Decision rule: "If a test ad hits 2 purchases at 1.8x MER, duplicate into scale within 12 hours."
  2. 02

    Use cross-channel sanity checks

    Compare TikTok against Meta Advantage+ and Google PMax to see whether demand is channel-led or creative-led.

    Prompt for Claude 4.5 Sonnet: "Explain why TikTok CPA is 18% lower than Meta while Northbeam shows flat blended MER."
  3. 03

    Automate pacing alerts

    Use n8n with Northbeam and TikTok reporting to trigger alerts when spend deviates from target by 20% or more.

    n8n rule: "If pacing < 85% of plan at noon, ping 1 operator and reduce audience fragmentation."
IfThen
a campaign exits learning but produces no 7-day purchase signalreduce budget by 40% and move the winner into a new cleaner ad set.
blended MER is above 2.2x but TikTok platform ROAS is below 1.2xtrust Northbeam and PostHog over TikTok Ads Manager.

4. Build an AI operator loop that ships 1 decision per day

The winning TikTok stack in 2026 is 1 operator plus 1 agent, not 4 coordinators. Use Lindy or 11x to collect daily insights, then push tasks into Attio and changes into n8n. Use Cursor or Claude Code on RevOps SQL to answer questions like which hook drove the best CAC payback by cohort. Claude 4.5 Sonnet should summarise comment themes, refund reasons, and landing-page drop-off into 1 daily brief. Keep the loop tight: data in by 9:00, decisions by 11:00, new assets live by 15:00. If a decision cannot be shipped the same day, it is not a real decision.

  1. 01

    Stand up the daily brief

    Have 11x or Lindy scrape TikTok, Northbeam, and PostHog, then write a 6-line brief into Attio by 9:00.

    Agent prompt: "Summarise yesterday's top 3 winners, top 3 losers, and 1 budget move in under 120 words."
  2. 02

    Answer the hard SQL fast

    Use Cursor or Claude Code to query cohort payback, creative fatigue, and first-touch vs blended revenue in one pass.

    SQL prompt: "Show CAC payback by creative_id for all buyers acquired in the last 28 days."
  3. 03

    Push changes automatically

    Use Trigger.dev or Inngest to move approved changes from Attio into TikTok campaign edits, Framer page swaps, and Slack alerts.

    Workflow: "If 1 creative wins 3 days in a row, create 1 task, clone the landing page, and raise budget by 15%."
IfThen
the daily brief has more than 3 unknownspause scaling and assign 1 owner to close each unknown within 24 hours.
2 consecutive days produce no new winning creativeswitch from iteration mode to problem discovery using comments, refunds, and session replays.
Pitfalls

What goes wrong

Mistake

Judging TikTok on platform ROAS only

Fix

Use Northbeam, PostHog, and blended MER before making any budget decision.

Mistake

Launching with 1 hero creative and 0 backups

Fix

Ship 12 assets minimum with Arcads, Runway, and AdCreative.ai before day 1 spend.

Mistake

Letting UTM drift break attribution

Fix

Enforce 1 canonical campaign schema in Attio and validate it with n8n every morning.

Mistake

Using Salesforce as the operating system

Fix

Default to Attio and only migrate off Salesforce if you need a lighter GTM stack.

Mistake

Scaling audiences before the hook is proven

Fix

Do not expand targeting until 2 creatives clear 28% thumbstop and 1.8x MER.

FAQ
How much budget do I need before TikTok can tell me anything useful?▾

Start at 1.5x your target CPA in daily spend for 7 days, or the learning phase will lie to you. Use Northbeam and PostHog to judge blended outcomes, then let TikTok optimise only after 2 creatives show stable purchase volume. If you cannot fund 12 assets and 7 days of tests, wait.

Should I use broad targeting or interest stacks?▾

Use broad first in 2026 unless your product is niche or regulated. Clay and Common Room should build exclusion and lookalike logic from real buyers, not from assumptions. If broad underperforms by 20% after 2 creative cycles, then test 1 narrow segment, not 8 overlapping interest stacks.

What does a good creative testing system look like?▾

One system ships 3 angles, 4 variants each, every 5 days. Arcads, Runway, and AdCreative.ai should generate the volume, while Claude 4.5 Sonnet clusters feedback from comments and refunds. If the team cannot produce 12 usable cuts per week, the bottleneck is ops, not TikTok.

How do I know when to scale a winner?▾

Scale only when 1 creative clears 28% thumbstop, 1.8x MER, and 2 consecutive days of stable CPA. Duplicate into a cleaner ad set, then raise budget by 15% to 20%. If scale breaks performance, the issue is usually landing-page mismatch in Framer or a weak audience from Clay.

What should the AI agent actually do each day?▾

11x, Lindy, or Relevance AI should produce 1 daily brief, flag pacing anomalies, and draft 3 next tests. Use n8n or Trigger.dev to move approved actions into Attio and TikTok. The agent should not strategise; it should compress decisions from 2 hours to 10 minutes.

Get one system like this in your inbox each week

No fluff, no upsell. Unsubscribe anytime.