How to set up TikTok custom audiences (4 types)

Open TikTok Ads Manager → Audiences (or Assets → Audiences) → "Create Audience" → "Custom Audience."

Select "Website Traffic." Pick the Pixel you installed.

Layer 1 — All Visitors (last 30 days): captures everyone who hit your site. Use for broad retargeting.

Layer 2 — ViewContent (last 14 days): users who viewed a product page. Higher intent than all visitors.

Layer 3 — AddToCart (last 7 days, exclude Purchase): users who added to cart but did not buy. Highest-converting retargeting audience.

Layer 4 — Purchase (last 180 days): existing customers. Use to EXCLUDE from prospecting (you are not trying to acquire them) and INCLUDE in upsell/cross-sell campaigns.

Prepare a CSV with columns: email, phone, first_name, last_name, country (more columns = better match rate).

TikTok will hash PII automatically on upload — you can upload plain text, TikTok hashes before storing. But for security, hash SHA-256 yourself before upload if possible.

Audiences → Create Audience → Customer File → upload CSV.

Map columns to TikTok's fields. Confirm hashing method (plain text → TikTok hashes, or already hashed → mark as such).

TikTok processes the file in 30 min - 2 hours. Match rate of 30-60% is normal (TikTok cannot match users who never logged in on TikTok with that email).

Build segments: All Customers, High-LTV Customers (top 20% by revenue), Churned Customers (no purchase in 180 days). Each gets a different campaign strategy.

Requires your TikTok Business Account to be linked to Business Center.

Audiences → Create Audience → Engagement → "TikTok Account."

Available segments: All users who engaged, video viewers (any video), profile visitors, followers, comments/shares/likes engagers.

Most useful: "Watched 75%+ of any video" in the last 60 days. These are warm without being purchase-intent — perfect bridge from organic to paid.

Combine with Website audiences for cross-funnel layering. Example: "Saw 75% of an organic video AND visited the site in the last 30 days" = high-intent warm audience.

Audiences → Create Audience → Engagement → "Video."

Available thresholds: viewed any video, watched 25%, 50%, 75%, 95% of the video.

Recommended layering:

• 25% video viewers — broad warm audience for sequenced retargeting (show a follow-up creative).

• 50% video viewers — engaged enough to consider a stronger offer.

• 75%+ video viewers — high attention, push toward conversion.

Select which ad accounts/campaigns/ads contribute. Useful when running multiple creative tests — segment audiences by which hook landed.

Use these for retargeting sequences. Example: anyone who watched 50% of Ad A in last 14 days gets retargeted with Ad B (the social-proof variant).

Audiences → Create Audience → Lookalike Audience.

Source audience: pick a HIGH-QUALITY seed (Purchase audience >1,000 users, or top-LTV customer file).

Country: pick one country per lookalike. Multi-country lookalikes have diluted match quality.

Audience size: 1% (most similar, smallest), 2-3% (good balance), 5-10% (broadest, lowest match quality).

Start with 1% lookalike of Purchase audience for prospecting. If that audience converts, expand to 2-3%.

NEVER build a lookalike from a small seed (<500 users) or a noisy seed (e.g., "all video viewers" without filtering). Garbage in, garbage out.

In every prospecting campaign, exclude: existing customers (Purchase audience), recently engaged users (already in your retargeting), and irrelevant segments.

For e-com: exclude Purchase last 90 days from prospecting (so you do not acquire your own customers as "new").

For SaaS: exclude trial users + paying customers from acquisition campaigns.

For service businesses: exclude job seekers (use negative interest categories).

Audience exclusions are set at the ad group level. Configure them in EVERY campaign — they do not propagate.

Monthly: review match rates and audience sizes. Lookalikes drift as your seed grows — rebuild from updated seed.

Monthly: rotate retargeting audiences. The "AddToCart last 14 days" gets stale; refresh by re-creating with current data.

Quarterly: review excluded audiences. Customer file audiences need re-upload as your customer list grows.

Annually: build new lookalike variants. Test 1% vs 2% vs 3%, evaluate by CPA over 30 days.

Audience operations is the difference between "set and forget" (typical DIY result) and compounding-quality targeting (the specialist edge).