TikTok and Meta Attribution: How to Reconcile Your Numbers

TikTok and Meta don’t measure the same thing: each credits itself with the conversions it saw, within its own window. In 2026, Meta applies by default 7 days after click, 1 day after engagement and 1 day after view; TikTok applies 7 days after click (options of 1, 7 or 14) and 1 or 7 days after view. Both therefore sometimes claim the same sale. The only reliable reference remains your back office: compare the platform totals against it, then calculate your discount coefficient per channel. By the end of this article, you will know which figure to put in your reporting, and which to throw away. Meta admits it itself: according to CB News (March 2026), the group acknowledges that its definition of “click-based” conversions can lead to an inconsistency between what the advertiser sees in Meta Ads Manager and what they see in third-party reporting tools.

Data collected in March 2026. Attribution moves fast: Meta changed its rules twice in three months. Everything that follows is dated and sourced, and we re-check it every quarter in the interfaces. If you are reading this page in 2027, check the default windows before applying anything.

Why does the sum of your TikTok + Meta + GA4 figures exceed your actual revenue?

Because each platform credits itself with the sales it saw, without knowing what the other saw. A single order can be claimed by Meta (a Reel viewed the day before), by TikTok (a click three days earlier) and classed as “direct” by GA4. Three systems, three truths, one single sale banked in Shopify.

The sentence we hear every month in steering committee: “it doesn’t match Shopify”. That’s normal. It isn’t a pixel failure, it isn’t a tracking bug, and in nine cases out of ten it isn’t a consent problem. It’s self-reporting. Meta marks its own homework. TikTok marks its own homework. Nobody talks to anyone. The subject is nonetheless closely watched: according to the CNIL’s 2025 sanctions review, 21 organisations were sanctioned in 2025 for breaches of the rules on trackers.

The typical journey of a €90 basket at one of our cosmetics clients, reconstructed from the post-purchase survey:

• Day 1, the person watches 8 seconds of a UGC video on TikTok. No click. TikTok records an engaged view.
• Day 3, they see an edit of the same video again as a Reel, click, visit the product page, don’t buy. Meta records a link click.
• Day 4, they type the brand name into Google, arrive organically, and buy. GA4, in the default data-driven model, mostly credits Google organic/Search here.
• Result: TikTok claims €90, Meta claims €90, GA4 credits Google. The back office records €90 once.

A gap therefore does not signal a malfunction. According to benly.ai (2026 data), a 20 to 40% gap between Meta Ads and GA4 is considered normal. Below 20%, be wary: either your windows are abnormally short, or your pixel only sends back part of the events. Above 40%, there is something else to look for, and it still isn’t necessarily a failure.

Before correcting anything, you need to know what each interface actually measures. Here is the matrix we print and stick on the wall of the accounts we manage.

Last line, the one that matters. A platform ROAS tells you whether to raise an ad set’s budget this afternoon. It is not for deciding the quarter’s media budget. Both uses are legitimate; mixing them is not. For official definitions, the TikTok Ads documentation (ads.tiktok.com/help) and the Meta Business Help Centre remain the only authoritative sources in the event of a disagreement with a provider.

What changed at Meta in 2026, and why have your conversions “dropped”?

Meta removed windows and moved conversions from one column to another: it isn’t a drop in performance, it’s a reclassification. According to goodmorningco.com (noted early 2026), Meta entirely removed the 7-day-view and 28-day-view windows in January 2026. And according to bloomanalytics.io, since March 2026 click-through only counts link clicks.

In practice: likes, shares, saves, comments and video views of 5 seconds or more are no longer “clicks”. They fall into a separate attribution type, engage-through, with a 1-day window. The former engaged-view only concerned video, from 10 seconds; according to theoptimizer.io, the new bucket covers all formats and lowers the video threshold to 5 seconds.

The example given by theoptimizer.io is the most telling. Before: someone shares your ad, buys five days later, the sale is counted under 7-day click. Now: that share opens only a 1-day window, and the purchase on D+2 is no longer attributed at all. The sale exists. It has simply become invisible in Ads Manager.

The wrong diagnosis is expensive. We have seen brands renew 100% of their creative pool in February-March 2026 to fix a drop that was accounting, not creative. An attribution reclassification isn’t treated like ad fatigue: the test that settles it is frequency and hold rate, not the “purchases” column.

Which attribution windows should you choose on TikTok and on Meta depending on your average basket?

The window must be set to your real decision cycle, measured in your back office, not to the platform’s default setting. If 80% of your orders land within 48 hours of the first visit, a 14-day window buys nothing but noise. If your basket exceeds €250, a 7-day window truncates your reading.

How to measure this cycle: export 300 orders, calculate the gap between the date of the first known session and the order date, look at the median and the 80th percentile. That figure is worth all the benchmarks. It takes an hour to produce.

As for the available settings: according to improvado.io (noted March 2026), TikTok natively offers 1, 7, 14 or 28 days after click, with 7 days as the default, and a view window that can be disabled or set to 1 or 7 days, with 1 day as the default. TikTok therefore natively covers up to 28 days after click; beyond that cycle length, you do have to go through an external attribution tool or the CRM.

A rule we apply without exception: never change a window mid-month. A window change makes the two halves of the month incomparable, and nobody remembers it four weeks later in committee. On TikTok, the rule is even enforced by the tool: according to the TikTok Ads Manager help centre, attribution settings can no longer be modified once the ad group is published, and you need to create a new ad group to use a different attribution window or a different event counting mode.

How do you calculate your platform discount coefficient in 30 days?

You divide the revenue actually banked by the sum of the revenues claimed by your platforms, and you get a coefficient that you then apply to each channel. A coefficient of 0.66 means your platforms collectively claim 1.5 times what you sold. That is neither a scandal nor fraud. It is the mechanics of self-reporting, quantified.

The protocol, in five steps, over a full calendar month:

• 1. Freeze the windows. Meta and TikTok, settings noted in writing, freeze date recorded. No change for 30 days.
• 2. Pull the figures three times. D+1, D+7 and D+30 on the same period. Platform figures come in after the fact, sometimes by a lot. If you pull them only once, you are comparing two different maturities.
• 3. Extract net back-office revenue. Returns deducted, cancelled orders deducted, taxes excluded if your pixels send figures excluding VAT. That line is the only one that doesn’t lie.
• 4. Calculate. coefficient = back-office revenue attributable to paid ÷ sum of claimed revenues. If you have strong organic, first remove the organic share measured by post-purchase survey.
• 5. Apply and display. The coefficient appears in the table, visible, with its date. Never applied on the quiet.

Full example, rounded illustrative figures, on a month with €100,000 banked and €33,000 of media.

What we observed at Esekai, internal data, October 2025 to February 2026, 14 French e-commerce accounts run in parallel on TikTok and Meta, average basket from €38 to €210, media budgets from €8,000 to €120,000 a month. Method: D+7 reading in each interface, Shopify revenue net of returns, windows frozen for a month. The sum of claimed revenues exceeded banked revenue by 28% to 61%, median around 38%. Between the D+1 reading and the D+7 reading, the revenue claimed by Meta rose by 6 to 14% on the same days. This isn’t a published study, it’s our data on our portfolio: treat it as an order of magnitude, not as a market norm.

The logic is the same as for media costs: a borrowed benchmark is worthless, a benchmark built in-house is worth everything. We detail this normalisation mechanism in our protocol on TikTok Ads CPM benchmarks in France.

A word of caution. This overall coefficient is a management-committee tool, not an attribution model. It assumes the three channels over-report in the same proportions, which is false. If you spend more than €150,000 of media a month, move to an MMM or an incrementality test. Below that, the coefficient does 80% of the work for €0 of tooling, and we would rather say so than sell you a software layer you don’t need.

How do you know which creative really sold when two platforms are fighting over the sale?

You never compare two creatives across two platforms in absolute value: you compare rankings within each platform, over the same window, from a shared asset identifier. The ROAS of a video on TikTok and the ROAS of the same video on Meta are not two measurements of the same quantity. Their rank, however, is comparable.

The prerequisite is stupidly logistical: one creative = one stable identifier, not two file names. Our naming convention, the one we enforce on every account we deliver: brand_angle_hook_variant_YYYYMM. Example: maisonlune_social-proof_direct-question_v3_202603. That same identifier appears in the ad name on Meta, in the ad name on TikTok, in the utm_content parameter, and in the source file name. Without it, no creative reconciliation is possible, and you lose two days a month stitching names back together by hand.

Platform attribution is only the top floor of creative measurement, and the noisiest. The three floors below, hook rate, hold rate, CTR, depend on no attribution window. They are measured in the same unit on both sides. That is why they are used to settle things first: the full method is in our 4-tier creative measurement protocol.

Three complementary actions, in order of profitability:

• The post-purchase survey. One question on the confirmation page: “how did you hear about us?”. Response rate observed among our clients: 25 to 45%. It is the cheapest and most honest data on the market, because it depends on no cookie.
• Consistent UTMs. Same convention on both platforms, distinct utm_source, utm_content = asset identifier. Without that, GA4 will be of no use to you in this exercise.
• Staggered launches. Launch a creative on TikTok a week before Meta, then reverse it on the next one. You won’t know everything, but you will see whether the sales curve moves before the platform claims it.

What we can’t do, and won’t pretend to do: cleanly isolate the contribution of a single creative in a six-channel mix. Nobody does it honestly below €150,000 of monthly media. What we can do is identify the top 3 and bottom 3 of a pool of 40 creatives, in a way that is stable from one month to the next. Once that top 3 is identified, the question becomes a production question: how many variations, at what pace. We have quantified it in the rule of 3, then 5 to 8 variants per angle.

Should you connect conversions APIs (CAPI and Events API) to make attribution more reliable?

Yes, but a conversions API repairs signal loss, not double counting. It sends back to the platform events that the browser no longer transmits. It will never tell Meta that TikTok has already claimed the same sale. Your gap with the back office won’t disappear after setup. It will shrink by a few points.

The context remains that of iOS 14.5: according to ruleranalytics.com, App Tracking Transparency requires explicit consent, a significant share of users refuse, and the corresponding conversions may never appear in Ads Manager. Improvado.io estimates that iOS users account for 50 to 65% of traffic for many brands. On our French fashion and beauty accounts, we are regularly above 60%. The Apple support page on App Tracking Transparency confirms it: since iOS 14.5, iPadOS 14.5 and tvOS 14.5, apps must ask permission before tracking your activity across other companies’ apps and websites.

What CAPI really repairs: events blocked by a browser, by a script blocker, by a loss of session between the app and the site. What it doesn’t repair: the fact that two platforms count the same sale, product returns, and the difference between click date and order date.

A French point, and a non-negotiable one. A server-side CAPI doesn’t exempt you from consent. If your CMP hasn’t collected agreement, you have no right to send the enriched event, whatever the technical route. The server is not a legal workaround. We have seen integrations that were technically clean and indefensible in front of the CNIL.

On event match quality (EMQ on Meta’s side), the recommendations circulating in the ecosystem vary by source and by interface version. We have no recent official threshold to quote you, and we would rather not invent one: open your Events Manager, note your current score, and work on improving it by adding hashed email, phone number and customer ID. The direction matters more than the absolute value.

Does TikTok really under-report its conversions compared with Meta?

TikTok is both under-credited in last-click and over-credited in view-through: both biases exist, they don’t cancel out, and they need to be measured separately. The received wisdom that “TikTok doesn’t convert” almost always comes from a last-click reading, which is the model least favourable to a discovery channel.

According to improvado.io, last-click systematically under-credits TikTok, and its 7-day post-click window undercounts its contribution when competing channels are read over 30 days. The same source describes TikTok’s engaged view as capturing the conversions of users who watched 6 seconds or more without clicking, then converted via organic search or direct access. This mechanism explains a large share of “direct” sales that are nothing of the sort.

The other side of the bias is just as real. Enable the 7-day view on TikTok with a large retargeting budget, and you will see your platform ROAS double without a single extra sale hitting the till. It happens every month. It is the trap on line 4 of our first matrix.

One figure circulates widely: dataslayer.ai claims that advertisers moving from a 7-day to a 28-day window would discover 63 to 79% more assisted conversions. We quote it because it is repeated everywhere, and we frankly put it in perspective: it is publisher data, the range is wide, and we have never reproduced it at that level on our accounts. In our readings, widening the window reveals far less than that. Take it as a directional signal, not as a promise.

The position we stand by, even when it complicates a sale: if your TikTok shows a platform ROAS of 1.2 and your overall MER has risen since TikTok launched, TikTok is working. If your MER hasn’t moved a point in three months, TikTok is only capturing sales you would have made anyway. The dashboard will never settle this debate. The MER curve will.

Which figure should go in the monthly report presented to management?

Three lines, not fifteen: net back-office revenue, the revenue claimed per platform with its coefficient displayed, and the total media cost that gives the MER. Everything else is operational steering and has no place in a management presentation.

The sentence to repeat until it sinks in: platform ROAS is for steering within the day, MER is for deciding the month’s budget. Never mix them in the same table. The day a managing director compares a Meta ROAS of 4.3 with a MER of 3.0 without understanding why both figures coexist, you spend twenty minutes defending your method instead of talking about growth.

What goes in the appendix, never on page 1: the campaign-level detail, CPM, CTR, engage-through conversions. What goes on page 1: the month’s MER, the MER of the three previous months, the media budget, and one sentence on what changed in attribution. That last sentence prevents half the misunderstandings.

Be careful when comparing ROAS across industries: a ROAS only makes sense alongside its attribution window, its model and its gross margin. Our ranges by brand type are detailed in what ROAS to expect from creative ads in 2026. A ROAS of 2.4 with a 70% gross margin is worth more than a ROAS of 4 with a 25% margin.

A last point on GA4, since the question comes up every time. According to durum.ai (French-language analysis, 2025), GA4 tends to over-attribute to Google, Search and YouTube, and to under-attribute Meta Ads. Trackingplan, for its part, points out that Meta includes view-throughs and GA4 doesn’t display them. GA4 is therefore not the referee. It is a third witness, with its own blind spots. The referee is the till.

If you were to retain only one action from this whole page: put a quarterly review of your attribution windows in your calendar, and note the date of the data pull at the top of every report. TikTok and Meta attribution moved twice in three months in early 2026. It will move again. The brands that find their way aren’t the ones with the best tool, they’re the ones that have dated their figures.

Frequently asked questions

How do you recover your TikTok account, and why is it the prerequisite for attribution?

Log back in via the associated email or number, or via the Google/Apple account used at sign-up; the “Find my account” option shows the profiles linked to the device. For a brand account, access is recovered through TikTok Business Center from the administrator. Without access to Business Center, you can see neither your pixels, nor your attribution windows, nor your conversions. It is prerequisite number zero.

Where do I find my TikTok link, and what is it used for on the advertising side?

Your link is tiktok.com/@yourusername. From the app: profile, Share button, Copy link. On the advertiser side, this handle and link are used to authorise an organic post as Spark Ads, via an authorisation code generated by the creator. It is also the way to link an organic video to its sponsored version, and so avoid counting the same video’s performance twice in your reporting.

How do you find the details of a TikTok advertising account?

In TikTok Ads Manager, the ad account ID appears at the top of the interface and in the account settings. Business Center lists the accounts, pixels, catalogues and associated users. This ID is required for any API pull, for connecting the Events API and for connecting a third-party tool. Note it down and keep it outside the interface: you will need it for every technical integration.

Does TikTok monetisation have anything to do with advertising attribution?

No. Creator monetisation depends on eligibility criteria on the personal or creator account side, age, number of followers, volume of views over 30 days, compliance with community guidelines, and these thresholds change by programme and by country. Advertising attribution concerns an advertiser account, a pixel and conversion windows. Two different accounts, two different logics, two different interfaces. Don’t look for your conversions in the creator space.

Why did my Meta ROAS drop in early 2026 without my sales falling?

Because Meta removed the 7-day-view and 28-day-view windows in January 2026, then restricted click-through to link clicks only in March 2026, according to goodmorningco.com and bloomanalytics.io. Likes, shares, comments and video views of 5 seconds or more moved into the engage-through bucket, 1-day window. Your conversions haven’t disappeared, they have changed column. Check your MER before touching your creatives.

What size of gap between platform and back office is abnormal?

According to benly.ai (2026), a 20 to 40% gap between Meta Ads and GA4 is considered normal. On our 14 French e-commerce accounts tracked from October 2025 to February 2026, the sum of claimed revenues exceeded banked revenue by 28 to 61%, median around 38%. Below 20%, suspect a loss of events. Above 60%, check your view windows and your retargeting.