Discrepancy Between TikTok Ad Data and Adobe Analytics - Need Help

[Luciano_Rizzuti]

Hi everyone,

We are experiencing an issue on our site that we hope you can help us with.

Our company regularly runs TikTok ads, and we are, of course, interested in analyzing the campaign data.

However, the data displayed in the TikTok Business Manager is significantly higher than what we see in our dashboard, with hardly any traffic visible.

Our website is built on AEM and also uses Adobe Analytics.

Our consent manager is Usercentrics, and we have the TikTok pixel integrated into our site.

For comparison - TikTok shows 290 pageviews for a sample campaign, while Adobe Analytics shows no pageviews, only 4 occurrences.

Usercentrics shows 45 consents via the cookie banner, with a consent rate of 100%.

We can definitely rule out our firewall blocking traffic from the website, and traffic is not being redirected to another site.

We use a link parameter to track traffic, which we embed in the link. Although it’s known that TikTok alters the link, the parameter remains intact in all known cases.

We have also searched for the extension in Adobe Data Collection, but it is no longer visible in the catalog. We are referring to the following article: TikTok Server Side Extension.

Does anyone have any ideas on what we could work on?

Thank you very much in advance for your help.

Best regards,

Luciano

[Jennifer_Dungan]

This is an issue that affects all pixel trackers like this, TikTok, Meta, Google Ads, etc...

The problem is that they can only see their data, not the whole of what is happening on your site.... you may also want to look at what attribution your TikTok pixel is set to.

Basically, when someone clicks on the ad, the service will drop a cookie on the user's device (this may be a 30 day cookie, or a 7 day cookie, etc).. and if the user converts within that time frame then a conversion against your ad will be recorded...

But in reality, that TikTok ad was likely not the last driver, so Adobe will record the attribution very differently, as it can see everything that is driving traffic to your site.

[bjoern__koth]

Fully agree with @Jennifer_Dungan 

Also, typically tictac will rather be classified as a marketing tag, whereas Adobe Analytics would be a performance tag, so differences in cookie consent may also occur (even if that is rather rare).

we have seen the same problem with vendors like LinkedIn which claim an almost 10x times higher click-through rate than what we can see reaching on the website which, cannot be explained with cookie consent or ad blockers.

Frankly, it sometimes feels like they might inflate their numbers to look better and justify the money they charge you for displaying your ads, but that may just be my opinion... 

we also sometimes saw redirects that would cut off query params you set on the ad URLs, so there are a bunch of things that may go wrong

[Jennifer_Dungan]

Lol.. 

Frankly, it sometimes feels like they might inflate their numbers to look better and justify the money they charge you for displaying your ads

---

Yeah, that seems accurate to me!

Anyway, all of these pixel trackers are "greedy"... trying to get the attribution... if you have a lot of them, and a lot of cross channel traffic you will see that they all try to take attribution.. and if you add those up the number of conversions would be highly inflated.

Our philosophy is to basically treat these like Participants, but not necessarily the Direct Driver...

[BlorgeHaffles32]

This is a very backwards and quite frankly subjective take on attribution analytics. The reality is that last click attribution models only benefit PPC campaigns due to the nature of tracking - if you look at the number of clicks on TikTok Ads manager you'll likely see a much higher number of users are engaging with your content and clicking through to your site (and only a small number of these users will experience issues accessing your site).

If you use a more complex attribution model (i.e. view through, multi touch), you'll see that paid social is often under-represented in the like of GA4 or Adobe - sometimes by a factor of >22x (see here) The truth is use should use different models of attribution to triangulate something closer to the truth. Unfortunately attribution is not always an exact science, but it's important to keep an open mind and appreciate that every platform measurement will differ. 

[Jennifer_Dungan]

Literally no one here said to ignore multi-touch attribution... but last touch will give you your direct traffic... not something potentially that happened a month ago (as most social media default to a 30 day model)... them trying to take 100% of the credit (as per their reporting) is pretty laughable.

But we were telling anyone to ignore that data, merely explaining why Adobe shows a very different value to what the advertiser is saying.

Any GOOD marketer is going to look at all the data, from all sources...  and yes, Adobe can look at Participation models as well, giving that multi-touch view of drivers. Or an even better model might be a Time Decay... so that the touch points get more attribution the closer they are to the conversion (or less the farther back they are).

But to be perfectly honest... if I clicked on a ad 30 days ago, then come to the site today directly and make a purchase, I can assure you... I don't remember that original ad and I am most likely not making a purchase for the same item that was in the advertisement. I am lucky if I even remember clicking on a ad... let alone what that ad was advertising. So that advertiser patting themselves on the back for "driving my conversion" is completely inflated.... It's one of the reasons we changed many of attribution models from the default 30 days to 7 days (as that it much fresher, more likely to still be in the user's mind, and therefore relevant information)

However, you are welcome to disagree with that assessment.

But the fact remains, we answered the question as to why the values don't match... How each system "determines" the attribution. How each company chooses to use all their data sources and how to weigh things is up to them to decide.

[BlorgeHaffles32]

I never said that you said to ignore MTA or VTA, I just wanted to mention that there are other attribution philosophies that exist, and that these could be taken into consideration outside of last click. 

I would say this whole take on attribution feels pretty narrow and doesn’t really capture the complexities we’re dealing with. Just because a last-click model shows “direct traffic” doesn’t mean it’s the best representation of actual user journeys. Social platforms don’t operate in a vacuum, and to your point, they use 30-day models because user engagement is a lot more complex and layered than that (also most social platforms use 7-day/1-day models fyi!)

When we stick to just last-click, we ignore the fact that most of these touchpoints are cumulative, not isolated. Using something like Time Decay or a multi-touch view better reflects the reality that clicks and engagement happen across multiple touchpoints. 

Sure, no one’s saying last-click is useless, but treating it like the gospel truth for conversions is a misstep. In reality, a layered approach gives a more accurate picture of what drives conversions, especially in complex campaigns where attribution is anything but black and white.

(I have worked across social, PPC and programmatic and social so I'm not biased, just helping Luciano make an unbiased media mix decision!) 

[Jennifer_Dungan]

Again, no one here said to ignore multi-touch,. we explained why the values don't match... how the systems work (both of which are based on last touch for each system... and yes, as far as the paid social media is concerned, when they only look at themselves, they are essentially an "isolated last touch".

So thank you for adding to the conversation reminding people to look at multi-touch attribution, just as they would look at direct.

But again, this user didn't ask what information to use, or how to interpret it.. they asked why it didn't match... 

So again, I very much appreciate what you are saying.. and I very much agree that multi-touch attribution is important... it is up to each business to determine what model works best for them.

[BlorgeHaffles32]

I personally feel that terminology like describing an attribution model as "greedy" or saying things like "but that may just be my opinion..." doesn’t necessarily provide the most objective perspective that clients deserve. I hope they continue to diversify their investment across multiple channels and see success for their business as a result.

[MandyGeorge]

@BlorgeHaffles32 it's ironic that in the same sentence you give your opinion and also say that giving an opinion isn't objective. This is a community forum, the OP isn't a client of anyone here, no one here is getting paid. This is a community where we all give our opinions to one another to try and help solve problems. There is absolutely nothing wrong with the answers from @Jennifer_Dungan or @bjoern__koth. They addressed the question and provided the information to help understand what is happening. In a lot of cases there will be more than one potential solution to a problem, everyone's implementation is different, so some solutions might work better than others. Without knowing all the details, an opinion is all you can give sometimes.