Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

TikTok Marketing › Troubleshooting

TikTok Events API Events Missing

Why TikTok Events API server events go missing while browser events arrive: a test code left on, another pixel code, a lapsed integration, payloads or consent.
Troubleshootingintermediate14 min readUpdated

First, read each event by route for the same closed days

Open Events Manager on the pixel the ad accounts actually optimise on, and check that its pixel code is the one in the site's container and in the server configuration. The pixel and Events API tutorial, linked under the first cause below, describes the view to read: for each event, how many copies arrived from the browser, how many from the server, and whether pairs were deduplicated. Write down, for the events that matter, the browser count, the server count, whether pairs were marked as deduplicated and the last day on which the server count looked normal. Silence from the first day means the route was never finished. Silence that began part-way through means something changed, and the day it went quiet points at whatever was released, reconnected or handed over that day. Where the browser column is empty too, the browser-side diagnosis for a silent TikTok pixel, linked under the cause on another pixel code below, is the place to start, since one release can silence both routes.

Some patterns look like a broken server route and are nothing of the kind.

What Events Manager showsWhy that can be normalWhat to read instead
No server copy of a form, contact or sign-up event on a shop that relies on its platform's integrationThe hosted route the tutorial describes sends the standard commerce events; whether it carries anything else is for the setup record to sayThe setup record, to see which route was meant to carry that event
Unique events below the browser and server counts added togetherDeduplication keeps one copy of each pair that shares an event name and identifierWhether pairs are marked as deduplicated; pairs that fail to match are the subject of the diagnosis for conversions higher than orders
Server events below the orders the shop tookThe tutorial's consent step gates both routes behind the visitor's choice, so a refusal leaves nothing to sendWhether a session that accepted produces a server copy
A server event missing minutes after a testNew activity takes a short while to appear in Events ManagerThe same view again before anything is changed

Then compare outside TikTok. For a closed week, set the server count for the purchase event beside the order system's count for those days, since the server route should report every order from a consenting visitor whether or not an ad was involved. A large gap that the expected patterns do not explain is a fault. Last, push a single event along the route with the test code set, and watch the test events view. If it arrives, the route can reach this pixel, and the fault lies in the live configuration or in the event's own fields; if it does not, the break sits earlier: in the pixel code the route addresses, the integration, the container or the consent check. The TikTok measurement hub explains how the browser, server, app and offline sources feed one pool of events, which is why a silent server route matters long before anyone opens a report.

Diagnosis table

Diagnosis

SymptomCheckLikely causeFix
Real orders appear in test events from the server; the overview shows no server copiesEvery live setting able to hold the test codeLive sends still carrying the test codeTake the code out of every live setting and hold it only where preview runs
The integration or container reports sends; this pixel shows noneThe pixel code in the server configuration against the code shared with the ad account; other pixels in the business's listServer events addressed to another pixel codePoint the route at the business-owned pixel, with a token generated in that pixel's settings
Server events stopped on a known date while browser events carried onShop app changes, reinstalls, platform moves and changes of staff or agency access that dayA shop integration that stopped sendingReconnect the integration under the business's own login to the right pixel
Browser tags fire; the server container shows no incoming request, or no outgoing request to TikTokServer container preview: request received, client that claimed it, tags fired, responseA server container with nothing coming in or going outMend the broken hop first and the TikTok server tag second
Server events arrive under an unexpected custom name, or without value and currencyOne test order's server event set beside its browser twinPayloads that do not match the pixel's eventsBuild both routes from one data layer contract with the documented names
The server column emptied on the day of a consent banner changeA test order placed in a clean browser after accepting consentA consent check that blocks every server eventSend the recorded choice with each order and let the server act on it

Fix, verify and prevent each cause

Live sends still carrying the test code

TikTok lets a server payload carry a test event code, which sends that event to the test view instead of live reporting, and the verification step of the pixel and Events API tutorial depends on it. The fault is a code that stayed in place after testing: written into the server tag or hard-coded in a backend call. Every real event then goes to the test view, where it looks healthy to anyone who checks there, and the overview stays empty.

Fix: remove the code from every live configuration and keep it, as general practice, in a separate test configuration that only preview sessions use, such as a variable set in the server container's preview workspace rather than in the published version.

Verify: the next real order appears in the overview with a server copy, and the test view stops receiving orders that nobody on the team placed.

Prevent: keep the test code in one documented place, and make its removal a named line in the release notes of any change to the server route.

Server events addressed to another pixel code

Each server payload names a pixel code as its source, and the tutorial has the access token generated in that pixel's own settings and stored as a secret. Server events therefore arrive wherever the configuration points, which need not be the pixel the campaigns use. Three cases fit the pattern: the code of a pixel built for a Staging Environment that was copied into production, a second pixel created when an integration was set up again, and a pixel that lives in a former agency's Business Center. The browser-side diagnosis for a silent TikTok pixel describes the same trap for the pixel itself: a pixel made in one person's Ads Manager is that person's, not the business's. Check: compare the pixel code in the server configuration with the one in the web container and the one shared with the ad account, then open the other pixels in the business's list and look for the missing server events there.

Fix: point the route at the business-owned pixel, generate a token from that pixel's settings, store it as a secret, and either retire the stray pixel or record why it exists.

Verify: a test event with the test code lands on the right pixel, and real orders follow in the overview the next day.

Prevent: as general practice, one pixel per website, its code and token location in the setup record, and staging settings kept in a configuration that is never published.

A shop integration that stopped sending

On a hosted shop platform, one of the routes the tutorial describes is the platform's maintained TikTok integration, which sends the standard commerce events from the server with no code. An integration is a connection between two accounts, and as general practice it can end without anyone noticing: the shop's TikTok app is removed or reinstalled, the shop moves to another platform, a theme or checkout change is released, or the person who connected it loses access. The tutorial on setting up TikTok Business Center makes the underlying point about ownership: what is created under a personal login leaves when that login does. Check: set the date the server count stopped beside the shop's change history, its app list and the staff and agency changes of that week, and confirm in the shop's admin which account and which pixel the integration is connected to now.

Fix: reconnect the integration from a login the business controls, to the pixel in the business's Business Center, and if the shop has moved platform, rebuild the route on the new platform before the old one is switched off.

Verify: a test order produces one browser and one server copy of the purchase, marked as deduplicated.

Prevent: record which login connected the integration and when, and put the integration on the list of things to check whenever the shop platform, its apps or the people with access change.

A server container with nothing coming in or going out

Where the route runs through a server-side Tag Manager container, as one of the tutorial's three options, the TikTok server tag can only send what the container receives. Server-Side Tagging is a relay, and the diagnosis for a server-side container that receives no data follows one request from the page to the endpoint, the client and the tags. If every vendor behind the container has gone quiet, the break is upstream of TikTok, at one of the hops that diagnosis walks. If other vendors still receive events and TikTok alone does not, the break is in the TikTok tag; as general practice, look for a trigger that never matches, an empty pixel code or token field, or a tag paused in the published version. Check: with the server container's preview open beside a web preview, place a test purchase and follow it: did a request arrive, which client took it, did the TikTok tag fire, and what came back from TikTok.

Fix: repair the hop where the request stops and confirm it in preview, then the tag's trigger and fields, and publish.

Verify: the preview shows an outgoing request to TikTok for the test purchase, and the test events view shows the server copy.

Prevent: watch outgoing requests to each vendor as the server diagnosis recommends, and run the go-live checklist for server-side tagging before any cutover; the introduction to server-side tagging sets out the relay of endpoint, client and tags that each check walks through.

Payloads that do not match the pixel's events

The tutorial is precise about what a server payload carries: the pixel code as its source, the event name spelt as the pixel spells it, when the event happened, an event identifier, hashed user details, and properties that repeat the browser event's value, currency and contents. It also warns that a standard event name spelt differently is treated as a custom event. A backend that sends its own spelling therefore produces a purchase that never appears under the standard purchase event, and one that reads value from another field sends an event parameter that no longer matches the browser's. Check: place one test order and set its server event beside the browser event in the test view: name, value, currency, contents and event identifier.

Fix: build both routes from one contract, as the guide to feeding every ad platform tag from a single data layer does for several platforms at once, with TikTok's documented event names and the same identifier generated once at the moment of the action.

Verify: the test view shows the server copy under the standard name, with the same fields as its browser twin, and the pair is marked as deduplicated.

Prevent: write the event contract into the measurement plan, and change it only when both routes change together; the help centre is the reference for the fields TikTok requires, which this page does not restate.

A server does not see the banner, so the tutorial has the consent decision made again on the server, and the privacy and consent guide sets the rule: the server must receive each visitor's consent state and honour it exactly as the browser would. The fault is a check that cannot find the state it expects, perhaps because a banner update renamed the value or stopped passing it with the order, and so treats every visitor as having refused. Browser events from consenting visitors keep arriving and the server column empties on the day of the banner change. Check: in a clean browser, accept consent, place a test order and look for its server copy; then refuse consent and confirm that neither route sends anything.

Fix: pass the visitor's recorded choice with the order to the system that sends server events, and send or withhold each event according to that choice.

Verify: accepted sessions produce a server copy and refused sessions produce none.

Prevent: test both consent branches on both routes after every change to the banner, the consent platform or the checkout.

Prevention: a server-route check after every release

The server route breaks when something around it changes, so the check belongs to every release that touches it: a shop platform or app update, a checkout or theme change, a publish of either container, a consent banner change, a new pixel or Business Center change, and any change of the people or agencies with access. The routine is short. Before the release goes live, place a test order through each route and confirm in the test view that one browser copy and one server copy arrive under the standard name, with value and currency, marked as deduplicated, and that no test code remains in the live configuration. The day after, read the counts by route against the days before the release. Weekly, set the server purchase count beside the order system's count for the closed week. A move to a new shop platform can end the integration that carries the route, so the plan for keeping TikTok tracking and ads working through a website move treats it as a single planned job.

Keep a record that makes the route legible to the next person: the pixel code, the route chosen, the deduplication field and the date, as the tutorial's final step suggests, and, as general practice, which login connected any integration and where the token is stored. Enter each change in the measurement change log too, so later readers know which weeks were measured on one route. When the server count does fall, treat the gap as a Tracking Outage and work the tracking outage response SOP. Before each new campaign goes live, the pre-launch checklist for TikTok Ads walks a test journey again and confirms that the optimisation event arrives once, deduplicated across both routes.

Read results from the gap with care. With the server route silent, the ad groups optimised on browser events alone and counted fewer results, so a cost per result that rose in those weeks may be a counting change rather than a delivery one, which the cost-per-result diagnosis separates before anyone blames the creative. Teams running Meta beside TikTok can run the same routine for both server routes together, since Meta's diagnosis for server events that never reach Events Manager follows the equivalent hops on Meta, and one release can break both. The Conversions API on either platform is only as dependable as the routine that checks it.

A bonsai shop's server purchases go quiet (illustrative)

Illustrative example — not a real result

Zjornakev Bonsai finds where its server purchases went

Zjornakev Bonsai, an invented online seller of bonsai trees, tools and pots, sends browser events through Tag Manager and server events through its shop platform's TikTok integration. Reading Events Manager by route, the owner finds 212 browser purchase events over a fortnight but no server copies for the last 8 days, against 248 orders in the shop over the fortnight. Browser counts are steady, so the pixel itself is fine, and a test event with the test code never reaches the test view. The date matches a theme update after which the TikTok app was reinstalled. In the shop's admin, the reinstalled integration is connected to a second pixel created during the reinstall, and that pixel's Events Manager holds 132 server purchases. The owner reconnects the integration to the pixel in the business's Business Center from the business's own login, records the change and the stray pixel, and places a test order that arrives once from each route, deduplicated. The following week shows 118 browser and 131 server purchase events, 131 unique after deduplication, against 141 shop orders, and the cost per purchase in Ads Manager falls back towards its level before the gap, which the owner annotates rather than credits to the ads.

Common mistakes

  • Reading the wrong pixel. The missing events may be sitting on another pixel; check the code in every configuration before rebuilding anything.
  • Mistaking deduplicated pairs for lost events. A unique total below browser plus server is deduplication working.
  • Leaving the test code in. Server events then look healthy in the test view and never reach reporting.
  • Connecting the integration from a personal login. The route then hangs on one person's access; connect it from a login the business controls.
  • Changing the consent banner without testing the server branch. A renamed value can switch the server route off for everyone.
  • Spelling events differently on the server. A purchase under a custom name never appears under the standard purchase event.
  • Changing several hops at once. Repair one hop, verify it, then move on, so the cause is known.
  • Waiting for the monthly report to notice. Read the counts by route the day after every release.

The TikTok Ads troubleshooting hub routes the other measurement, delivery and creative symptoms to the page that diagnoses each one.

Frequently asked questions

How can I tell whether server events are missing or simply deduplicated?

Read the counts by route rather than the total. Events Manager shows how many copies of each event arrived from the browser and from the server, and whether pairs were deduplicated. A deduplicated pair counts once, so the unique total sits below the two routes added together, and that is the design working. Server events are missing when the server count itself is empty, far below the order system's count for that period, or stopped on a particular day.

Why do server events show in test events but never in the overview?

That is the pattern a test event code produces when it stays in the live setup. The Events API accepts that code so a team can watch server events in the test view without touching live reporting, and a code left in the live setup sends every real event there too. Take it out of every live setting, keep it, as general practice, only where preview sessions can reach it, and confirm that the next real order shows a server copy in the overview.

Do missing server events change what delivery learns, or only the reports?

Both. Delivery optimises on the events that arrive, so with the server route silent an ad group learns from browser events alone, which lose the purchases that blockers and short-lived cookies hide among visitors who consented. Fewer counted results also raise the reported cost per result with no change in the ads. Fix the route before changing bids, budgets or creative, and read the weeks of the gap as a measurement break.

Terms used on this page

Frameworks

Sources

  1. About Events API — TikTok for Business
  2. About TikTok Pixel — TikTok for Business

Related