TikTok Ads Fundamentals

Campaign Structure

Three levels, familiar from Meta: campaign sets the objective and overall budget, ad group sets targeting, placement, budget and bidding, and ads hold the creative.

Objectives span awareness, consideration and conversion. As with any algorithmic platform, choose the objective that matches the outcome you actually want — optimising for traffic when you want purchases produces cheap clicks and no revenue.

Why Creative Dominates

TikTok distributes content through the For You feed based on engagement signals rather than follower relationships. The same logic applies to ads: creative that holds attention gets distributed cheaply, creative that does not becomes expensive regardless of targeting precision.

What performs: vertical video shot on a phone, a hook in the first second, native pacing, on-screen text, trending audio where appropriate, and a real person speaking rather than a product montage.

What fails: repurposed TV commercials, landscape video, corporate voiceover, and anything that signals "advertisement" before the viewer has decided to watch.

Spark Ads

Spark Ads promote existing organic posts — either your own or, with permission, a creator's — rather than uploading a separate ad asset.

They frequently outperform standard ads for two reasons. The post carries its existing engagement, comments and follower count, which reads as credible rather than paid. And it appears in-feed as a normal post.

For most brands, promoting well-performing organic content or creator content through Spark Ads is the highest-return format on the platform.

Targeting

Options include demographics, interests, behaviours, custom audiences from your own data, and lookalikes. Also available is targeting by interaction with specific content categories or creators.

Broad targeting generally outperforms narrow targeting on TikTok, because the algorithm finds responsive users faster when given room. Over-constraining audiences starves delivery and raises costs.

Give the system creative variety and conversion signal rather than trying to define the audience yourself.

Measurement Realities

TikTok drives behaviour that is difficult to attribute. Users discover a product on TikTok and then search for it, or buy later on another device. Last-click measurement systematically undercredits the channel.

Use the TikTok pixel and Events API for what can be tracked directly, and supplement with post-purchase survey questions asking how customers heard of you, and with holdout tests where budget allows.

Setting Up So the Learning Phase Completes

The most common cause of a failed TikTok test is a structure that prevents any ad group from gathering enough data to optimise.

Fewer ad groups, more budget each. Splitting a modest budget across six audiences means none exits the learning phase, and six inconclusive results read as the platform not working.

Do not edit during learning. Changing the budget, the bid, the audience or the creative restarts it. The instinct to fix a slow start is exactly what prevents the recovery.

Optimise toward an event that happens often enough. A conversion occurring a handful of times a week cannot train a bid strategy. Optimise for a more frequent upstream event and measure the rare one separately.

Give it a fortnight before judging. Shorter reads are dominated by learning-phase volatility, and the decisions made from them are usually wrong in both directions.

Instrument before launching: the pixel or events API firing, the events you care about defined, and where available a server-side connection, because the browser-side signal is the weakest link.

Reading Attribution Honestly

Platform-reported performance is measured by the party selling the inventory, and on a high-reach video platform the standard settings are generous in a specific direction.

Separate click-through from view-through in every report. Blended, a view-through window on a platform with this much reach credits purchases that were already happening.

Shorten the default windows and compare against your own analytics. The gap between platform-reported conversions and what your own system sees is the number to understand, not to eliminate — they are measuring different things and both are internally consistent.

Watch total revenue, not attributed revenue. If platform-attributed sales rise while total sales do not, the channel is reallocating credit rather than creating demand.

And run a holdout for anything material. A geographic or audience holdout over a defined period is the only method that answers the incrementality question directly. It costs real sales and it is the only evidence that survives a sceptical finance review.

Sources

What each claim on this page rests on. Entries are typed so you can see which are primary.