Why sales change the TikTok setup
Awareness and consideration campaigns on TikTok can run before measurement exists. Sales campaigns cannot. The conversion objectives ask delivery to find people likely to complete a purchase event, and the system can only learn from events it receives, deduplicated and carrying the right value. A shop whose events are doubled, missing or valued differently from its order system is teaching TikTok the wrong lesson every day the campaign runs. The ecommerce conversion hub treats the shop's side of the purchase; this page treats the signal that connects it to TikTok.
The creative condition is just as strict. On a shopping results page the shopper has already decided to look for the product. In the TikTok feed nobody was looking, and a product photograph with a price rarely interrupts the scroll. Products sell on TikTok when a person shows what the product does, in a setting the viewer recognises. The setup therefore treats creative production as part of the campaign build, not an input to it.
The online sales objective hub sets the KPI set every sales channel is judged on, and the paid social channel page for online sales covers how much of the plan social ads should carry across networks.
Recommended setup for sales
| Setting | Recommended for sales | Why it differs from other objectives | Feature page |
|---|---|---|---|
| Objective | Product sales with a catalog for wide ranges or TikTok Shop; Website conversions for a few hero products | Delivery predicts purchases, which reach and traffic objectives never ask it to | Campaign structure |
| Events | CompletePayment with value and currency, sent by pixel and Events API with a shared event identifier | Purchase optimisation and return reporting depend on one clean, valued event | Pixel and Events API |
| Optimisation step | Add to cart while purchases are too rare to learn from; purchase once volume allows | A conversion goal that rarely fires leaves the ad group learning indefinitely | Campaign structure |
| Creative | Creator demonstrations of the product in use, promoted as Spark Ads or added to catalogue ads | The feed rewards showing over displaying | Shopping and catalog ads |
| Product page | Fast mobile page that opens on the product shown, checked in the in-app browser | The purchase happens on a page the advertiser controls, under feed-level impatience | Product page optimisation |
Product sales with a catalog or Website conversions, chosen by how the store sells
The help centre documents Product sales for advertisers with a product catalogue or a TikTok Shop, where ads can be assembled from catalogue items and matched to the products people viewed, and Website conversions for advertisers optimising towards events on their own site. The choice follows the shop. A store with a wide range, where most revenue comes from many products, gains from a catalogue: one setup covers the range, and the shopping and catalog ads guide explains why feed quality then decides results. A store whose revenue rests on a few hero products gains more from Website conversions with creative made for each one. The campaign structure guide shows how the objective and promotion type fix what each ad group can do, so pick before building.
Purchase event with value, deduplicated between pixel and Events API
Fire CompletePayment once per order, at the moment payment confirms, carrying value, currency and the product identifiers. Send the same event from the server through the Events API with the same event identifier, so TikTok keeps one copy. The pixel and Events API tutorial walks through each step, including the test events view that confirms deduplication. Agree with finance what the value includes, before or after tax and shipping, and send that consistently; return on ad spend (ROASRevenue generated per unit of ad spend.) in Ads Manager is only comparable with the shop's figures when both use the same definition. Product identifiers that match the catalogue are what let catalogue ads recognise which products a shopper saw.
Optimisation stepped to add to cart while purchases are rare
The help centre states that an ad group needs a number of conversions within its opening days to leave learning, and gives the current figure. A shop that will not record that many purchases from TikTok on its budget should not optimise on purchase yet. Step up to AddToCart, which happens more often and still signals buying intent, and judge the campaign on purchases regardless. When purchases from the campaign arrive steadily, add a new ad group on purchase rather than editing the learning one, and move budget across as it settles. This step is the main difference between a sales setup that works on a modest budget and one that stalls.
Creator demonstrations rather than catalogue images
The product has to be seen doing its job. A creator using a product in a real setting, saying what surprised them, does what a catalogue image cannot. Brief for the demonstration, not the endorsement: what the product solves, the moment it does so on screen, and the honest limits. Promote the posts that held attention organically as Spark Ads, and where catalogue ads are used, add video to the catalogue so the assembled ads are not stills. The RACE framework applied to sales is a helpful lens on where demonstration sits between reaching people and converting them.
Product page checked for the in-app browser shopper
The visitor arrives from a demonstration, inside TikTok's in-app browser, ready to buy the thing they just saw. Land them on that product, not the home page, and make the page fast and clear on a phone: price, delivery, returns and the buy button visible without hunting. Check that express payment options work in the in-app browser, since some depend on the device's own browser. The product page optimisation guide covers the page in depth; the TikTok check is to open it from a test ad on a phone and buy something.
A worked sales setup
Snapgrip, an invented phone-accessory shop
Snapgrip sells magnetic phone grips and stands with an average order value of £24 and a gross margin of 60%, so its break-even return on ad spend is about 1.7. It spends £100 a day on TikTok.
| Setting | First month | After the change |
|---|---|---|
| Objective | Website conversions | Website conversions |
| Events | Pixel only; CompletePayment on page view of the thank-you page | Pixel and Events API, order number as event identifier, value from the data layer |
| Optimisation step | Purchase | Add to cart for 3 weeks, then a new ad group on purchase |
| Creative | Product photos on a white background, animated | 5 creator videos: one-handed scrolling, desk stand in use, car mount |
| Product page | Collection page | The product shown, with express checkout tested in the in-app browser |
In the first month the ad group recorded 21 purchases from TikTok's view and never left learning; the shop counted 16 orders attributable by UTM, a return of 0.13 on £3,000. In the next month on add to cart, TikTok recorded 640 add-to-cart events and 118 purchases, and deduplication brought the purchase count to within 6% of the shop's own figure. In the third month, on the new purchase ad group, the campaign recorded 205 purchases for £3,000, a return of 1.9 at an average order of £27.80.
The first month's problem was not the budget. Purchase optimisation had too few events to learn from, the events it had were unreliable, and the creative showed objects rather than uses. Fixing the signal and stepping the goal gave delivery something to learn; the demonstrations gave it something worth showing.
Measuring sales in TikTok
Ads Manager reports Purchases and Cost per purchase for the CompletePayment event, with Purchase value and ROAS where value is sent. Read Add to cart events beside them, because the ratio of carts to purchases shows whether the product page and checkout are converting the interest the ads create. Events Manager's deduplication diagnostics show how many events arrived from each route and whether they were matched; a purchase count that exceeds the shop's orders usually means mismatched identifiers. Why that mismatch lowers the reported cost per purchase is explained in CPA in TikTok Ads. Video watch metrics on the sales creatives, average watch time and the share passing the opening, explain why one demonstration sells and another does not.
The business measures sit in the shop and analytics: orders, revenue, average order value (AOVAverage revenue per order.) and contribution after ad cost, read against the break-even return the margin allows. TikTok's figures follow its AttributionAssigning conversion credit to touchpoints. settings, so agree which settings the team reports and keep them fixed. Review event health and cost per purchase twice a week, return and creative weekly, and reconcile TikTok with the shop monthly.
Where the sales setup goes wrong
The setup fails when purchase optimisation starts on a budget that cannot produce enough purchases to learn, and the team concludes that TikTok does not sell. It fails when pixel and server events carry different identifiers, so reported purchases double and delivery chases phantom buyers. It fails when the value sent to TikTok is not the value finance reports, and the return argument never ends. And it fails on product pages that work on desktop but break in the in-app browser, where the shopper who was ready to buy meets a checkout that will not load. The TikTok Ads launch QA checklist checks each of these before a sales campaign goes live.
When Ads Manager reports more purchases than the shop took, the page on TikTok conversions higher than orders matches events to order numbers to separate real duplicates from attribution.
Before paying creators for authorisation at scale, the Spark Ads test comparing creator and brand handles checks whether the creator's identity lowers cost per purchase once fees are counted.
Common mistakes
- Purchase optimisation from day one on a small budget. Step up to add to cart until purchases flow.
- Purchase fired on a page view. Refreshes and revisits become sales; fire once per confirmed order.
- Pixel and Events API without a shared identifier. Both copies count, and every report inflates.
- Catalogue stills as the only creative. Show the product in use, in a person's hands.
- Landing on a collection page. Send the viewer to the product they just watched.
