TikTok ads for ecommerce work when three things line up: a product that sells itself on camera, a steady supply of native creative, and measurement that looks past TikTok’s own ROAS column. Keep the account structure simple, run Spark Ads from real creators, send purchases through both the Pixel and the Events API, and judge the channel on blended results. In my experience, accounts that stall after the first month stall on creative supply or measurement, not targeting.
Is TikTok right for your brand and price point?
TikTok is a discovery channel: people are scrolling, not searching, and your ad has a second or two to stop them. That favors products with a visible payoff: a before and after, a satisfying demo, a problem solved on camera.
Use this as a rough fit check, not a benchmark:
| Situation | TikTok’s likely role | Judge it on |
|---|---|---|
| Impulse-friendly price, product demos well on video | Primary prospecting channel alongside Meta | New-customer MER and CAC |
| Mid-range price, needs some consideration | Demand creation; purchases land later through search, email or direct | Blended MER, branded search, survey answers |
| High price, long research cycle | Awareness, plus creator content you reuse on other channels | Holdout tests and assisted signals, not last click |
| Restricted category (certain health, weight-loss or financial claims) | Check ad policies first; many claims won’t run | Whether compliant ads still sell |
Two more filters. Margin: TikTok usually looks worse in Ads Manager than in reality, so thin-margin brands have less room to wait. Creative capacity: if you can’t source several new videos a week, the account stalls once the first winners wear out.
Account and campaign structure
TikTok Ads Manager uses the same three layers as Meta: campaign, ad group, ad. The mistake I see most often is a separate ad group for every interest, age band and placement. TikTok’s delivery needs concentrated conversion volume to learn, and splitting a modest budget across many ad groups starves all of them.
A structure that holds up:
| Campaign | Purpose | Setup | Starting share of spend |
|---|---|---|---|
| Prospecting | Main engine for new customers | Conversion-focused objective, optimized for purchases, broad targeting, 1-2 ad groups | 70-80% |
| Creative testing | Find new winners without disturbing prospecting | Same optimization event, its own budget | 10-20% |
| Retargeting (optional) | Recover engaged viewers and cart abandoners | Video viewers, profile engagers, site visitors | 0-10% |
Treat the shares as starting points. At low spend, skip retargeting: the pools are small and broad prospecting already reaches warm viewers.
Settings worth getting right:
- Optimization event. Optimize for purchases if each ad group can get enough of them. TikTok’s own guidance points to roughly 50 conversions per ad group within a week to exit the learning phase. Far below that, optimize for add to cart or initiate checkout, then switch.
- Bid strategy. Start with maximum delivery (lowest cost). Add a cost cap or minimum ROAS goal only once performance is stable, because tight caps limit delivery.
- Automated campaigns. TikTok’s fully automated campaign types handle targeting, creative mixing and budget for you. They can work with proven creative and clean data, but they’re a poor place to start: you learn little about what works.
The testing cadence itself, including how many concepts to run and when to call a winner, is in the ad creative testing framework.
Spark Ads and working with creators
Spark Ads let you run an existing organic post, from your brand account or a creator’s, as an ad. The ad shows the original account, and likes, comments and follows accrue to the original post. Viewers see a real person’s video, not a brand ad, which is the point.
The creator publishes the video, then generates an ad authorization code in the app (choosing how long you can run it) or approves your request through TikTok’s creator marketplace. You add the post in Ads Manager with your call to action and landing page.
Where to find creators, roughly from cheapest to most expensive:
- Your customers. People who already post about you or leave detailed reviews.
- TikTok search. Creators already making videos about the problem your product solves.
- TikTok’s creator marketplace. Filter by audience, category and past performance, with payment built in.
- UGC creators. Creators who make content for brands, often without a big following. Good for volume; pair them with creators whose posts you can Spark.
Follower count matters less than you’d think: for paid use, you’re buying someone convincing on camera, not their audience.
Put these in every creator agreement:
- Paid usage rights and Spark authorization, for the same period
- Raw footage and a version without on-screen text, so you can recut it
- At least three hook variations per video
- Content disclosure turned on, as TikTok’s branded content policy requires for paid partnerships
- Fee per video, plus any fee for extending paid usage
Native creative: hooks, formats and creator briefs
Creative that works on TikTok looks like TikTok: vertical 9:16, sound on, shot on a phone, a person talking to camera or using the product. Polished brand cutdowns tend to get skipped. Keep text and product clear of the right edge and the bottom of the frame, where TikTok’s buttons and caption sit.
Hooks
The first one to two seconds decide whether anyone sees the rest. Hooks that work for DTC:
- Problem call-out: “If your [problem] does this, watch this.”
- Result first: open on the outcome, then show how you got there.
- Objection: “I thought this was overhyped until…”
- Side by side: your product next to the thing it replaces.
Write every concept with three hooks. Swapping the first two seconds is the cheapest test on TikTok.
Formats
| Format | What it looks like | Works for |
|---|---|---|
| Talking-head testimonial | Creator explains why they switched | Most products, with a convincing creator |
| Demo or tutorial | Hands and product, step by step | Beauty, kitchen, tools, gadgets |
| Before and after | Result shown in the first frame | Visible results; check policy for health and beauty |
| Green screen | Creator in front of reviews or a product page | Social proof and comparisons |
Creator briefs
Keep the brief to one page and leave room for the creator’s voice:
- One message: the single thing viewers should believe afterward.
- Audience and problem: in the customer’s own words, pulled from reviews.
- Must-show and must-avoid: the product in use and the result; claims you can’t make.
- Hooks to try: two or three options, plus permission to write their own.
- Deliverables: length range, hook variations, raw files, and music from TikTok’s commercial library or none, since trending sounds often aren’t licensed for ads.
Skip word-for-word scripts. A creator reading lines sounds like an ad, and people scroll past ads.
Pixel and Events API setup
TikTok optimizes toward whatever you report as a conversion, so broken tracking trains the algorithm on the wrong thing. Run the browser Pixel and the server-side Events API together, sending the same events, deduplicated.
On Shopify, the TikTok sales channel app handles most of it: set data sharing to the highest level, which adds the Events API alongside the Pixel. Elsewhere, use a server-side tag manager or a direct integration. Then check:
- ViewContent, AddToCart, InitiateCheckout and purchase events fire from both Pixel and Events API, with value and currency
- Browser and server events share an event ID, so TikTok deduplicates instead of double counting
- Advanced matching is on, sending hashed email and phone where available
- First-party cookies are enabled in the Pixel settings
- The ttclid click ID survives to the landing page (no redirects stripping parameters)
- Only one integration sends purchases, with no leftover manual pixel
- A test order shows up once in Events Manager, with the right value
Before scaling, I compare a week of purchase events against actual store orders. A large gap either way means duplicate or missing events.
Measuring beyond in-platform ROAS
TikTok’s reported ROAS is one input, not the verdict. It drifts from reality in both directions:
- It misses delayed and cross-channel sales. Someone watches, doesn’t click, then searches your brand on Google or buys on Amazon days later. Click-based reporting can’t see most of that.
- It can over-claim. View-through and engaged-view attribution credit TikTok for purchases by people who watched but might have bought anyway.
Know which attribution windows your account uses, and compare TikTok and Meta on matching windows. Then measure at the business level:
- New-customer and blended MER. When TikTok spend rises, do new-customer revenue and branded search follow? The MER vs ROAS breakdown covers targets and a weekly scorecard.
- Post-purchase survey. Add TikTok to “How did you hear about us?” If it appears often in answers but rarely in attribution, it’s doing more than Ads Manager shows.
- Holdout test. Once spend is meaningful, pause TikTok in a set of regions for a few weeks, or run TikTok’s conversion lift study if your account qualifies.
Scaling without collapsing efficiency
Most TikTok accounts don’t hit a targeting ceiling. They hit a creative ceiling, because the feed rewards novelty and winners tend to wear out faster than they do on Meta.
- Raise budgets in steps, for example 20-30% every few days, rather than doubling overnight, which can destabilize delivery.
- Scale with creative, not ad groups. Add new winners to the existing prospecting ad group instead of cloning ad groups that compete for the same people.
- Keep the pipeline ahead of spend. More budget needs more concepts in rotation, so order creator content a month ahead.
- Watch marginal returns. If spend rises 30% and new-customer revenue rises 5%, the extra money isn’t paying, whatever TikTok’s ROAS says.
- Automate last. With several proven concepts and clean data, test an automated campaign or value-based bidding against manual prospecting.
This is how I run TikTok inside performance marketing engagements: one simple structure, a steady creator pipeline and decisions made on contribution, not dashboard ROAS.
Get it built
If TikTok looked great for a month and then stalled, the cause is usually creative supply, tracking or the metric you’re judging it on. The Growth Audit covers all three, and it’s $1,500 fixed and credited if we continue. See pricing or get in touch.