Search the guides

TikTok Ads

Keep the TikTok click connected to the conversion

Understand the separate jobs of click IDs, website events and order records before adding another redirect.

One purchase needs a consistent story

The shop has an order, but TikTok's report lacks the purchase you expected. First check three separate steps: the website reported the action, the receiving system accepted it, and the report connected it to advertising.

TikTok Click ID, shown as ttclid, is a value attached to the ad's landing address. It helps connect a website action to a click. It is not a campaign name, customer account or receipt.

An event is a message about an action, such as a purchase. Attribution links a result to advertising. A real order can exist even when one of those measurement steps is missing.

Give each system one job

Event flowThree records, three jobs
  1. Click contextWhere this visit came from
  2. Website eventWhat actually happened
  3. Order recordWhat the business accepted

Do not use a click label as a receipt or a button tap as proof of a completed order.

The Pixel reports from the visitor's browser. Events API sends event information from a server. Both describe actions; the order system confirms whether the business accepted a purchase. A button press alone is not that confirmation.

When both routes report the same purchase, the integration needs to recognise one action rather than count two. This is deduplication. Check the current identifier requirements in TikTok's documentation and keep the two reports consistent.

Find the point where context disappears

Hypothetical example: a product URL contains ttclid, but an automatic language redirect removes it before the measurement integration reads the page. Checkout works and the order exists. Investigate that first lost handoff before rebuilding checkout.

Follow the ad preview, landing request and move to checkout. In a controlled test, compare the permitted click value at each required step. Seeing it in the address does not prove that the integration captured or sent it.

  1. Confirm the tested ad uses the intended final destination.
  2. Observe whether the click parameter reaches that destination.
  3. Check how permitted click context is retained for the relevant visit.
  4. Trigger one documented test action.
  5. Confirm the browser/server event and business record refer to that action.

Interpret differences before correcting them

Decision pathWhere should the investigation go?
A business result exists

Find the corresponding controlled test order.

  • No event found

    Check the trigger and delivery.

  • Event rejected

    Inspect validation diagnostics.

  • Event accepted

    Check matching and report definitions.

Start from one known test action, then follow the first missing piece of evidence.

Compare the same action and period. An attribution window is the time allowed for linking a result to an earlier ad interaction. Reporting delays and different system clocks can also shift where a purchase appears.

Separate an event that never arrived from an accepted event assigned differently in a report. The first needs a delivery check; the second needs a matching and reporting check. Changing a working redirect will not solve both.

A server connection does not override the visitor's data choices. Decide what information the business may share before choosing how to send it.

Write the event definition before inspecting the count

Agree what 'purchase' means: order created, payment authorised or payment settled. Choose one business milestone for the event and write it down for marketing and engineering.

Define the transaction reference, value and currency needed for that milestone. Send only necessary, permitted information. An event specification is not a reason to copy every customer field.

In an illustrative setup, the browser reports on the thank-you page and the server reports after payment confirmation. A visitor may close the page earlier. Explain the different triggers before interpreting the difference as fewer orders.

Distinguish a retry from a new action

A delivery request may be retried after a delayed response. Find out whether the repeated message describes the same purchase or a new one. Similar timestamps alone cannot answer that.

Test one action through the intended reporting routes, then a repeated delivery in the supported test environment. Write down the expected count before looking at the result; do not manufacture real orders for the test.

Document how the business order reference relates to the event identifier. An integration change should not silently give the same purchase a new identity.

Two approachesTwo messages or two purchases?

One action, repeated reports

  • The browser and server describe the same purchase.
  • Their event identity must remain consistent.

Two real actions

  • The business has two distinct purchases.
  • Each requires its own traceable identity.

Deduplication means recognising repeated reports about one action. It should not merge genuinely separate orders.

Inspect the external handoff

Mark an external checkout or payment service as a boundary in the route. Which system receives the click information, which knows the order result, and how are those facts linked?

Ask how the actual integration carries permitted information across that boundary. Browser storage does not automatically follow every move to another domain or app. An undocumented handoff is a question to resolve, not a reason to add another redirect.

Test completion and a return without completion. Seeing checkout is not a purchase. Name an owner for the whole handoff when the shop, payment and analytics systems belong to different teams.

Produce a redacted trace of one action

Keep a controlled test's time, route, event name, delivery result and private reference linking it to the order. Redact identifiers before sharing with people who do not need the raw values.

State what each observation proves: receipt, acceptance or attribution. A successful delivery response does not establish how every report will display the event. Use the platform's diagnostics for the next stage.

Repeat this test after checkout or measurement changes. Record untested payment methods and pending report results so the handover does not imply complete coverage.

Keep the route understandable

Write down who owns the ad address, landing page, event delivery and order record. Explain where click information is captured and what business result triggers the event.

Remove obsolete redirects only after checking dependencies. A clear route makes discrepancies easier to investigate; it also prevents a reporting gap from being mistaken automatically for fraudulent traffic.

Questions answered

With Events API, do I still need to check the browser?

Yes, still check the browser. The visitor must complete the page task, the click information needs a defined route, and browser/server reports must not count one action twice.

ADSTRAFFIC / ENAll guides →