If you run Facebook Ads and want to see how they perform in Google Analytics, you may have noticed some big discrepancies between the data available in Facebook Ad Manager and GA. Both systems use different ways to track clicks and visitors, so let’s unpick where the differences are.
There are two kinds of metrics you’ll be interested in: ‘website clicks’ = the number of Facebook users who clicked on an advert on your own site, and (if you do ecommerce) the transaction value which was attributed to that advert.
Website Clicks vs Sessions from Facebook
1. GA isn’t picking up Facebook as the referrer
If users click on a link in Facebook’s mobile app and your website opens in an in-app browser, the browser may not log that ‘facebook.com’ was the referrer.
You can override this (and any other link) by setting the medium, source, campaign and content attributes in the link directly.
Pro Tip: you can use GA’s URL builder to set the UTM tags on every Facebook campaign link for GA.
In GA, under the Admin tag and then ‘Property settings’ you should also tick the box saying ‘Allow manual tagging (UTM values) to override auto-tagging (GCLID values)’ to make this work more reliably.
2. The user leaves the page before the GA tag fires
There’s a time delay between a user clicking on the advert in Facebook and being directed to your site. On a mobile, this delay may be several seconds long, and during the delay, the user will think about going back to safety (Facebook’s app) or just closing the app entirely.
This will happen more often if the visitor is not familiar with your brand, and also when the page contents are slow to load.
By Facebook’s estimation the GA tracking won’t fire anywhere between 10% and 80% of clicks on a mobile, but fewer than 5% of clicks on a desktop. It depends on what stage in the page load the GA pixel is requested.
If you use a tag manager, you can control this firing order – so try firing the tag as a top priority and when the tag container is first loaded.
Pro Tip: you can also use Google’s mobile site speed suggestions to improve mobile load speed, and reduce this post-click drop-off.
It’s possible your page has a specific problem that prevents the GA tag firing only for mobile Safari (or Android equivalent). You’ll need to get your developers to test out the landing pages specifically from Facebook’s app. Luckily Facebook Ad Manager has a good way to preview the adverts on your mobile.
Facebook Revenue vs GA Ecommerce revenue
4. Attribution: post-click vs last non-direct click
Currently, Facebook has two types of attribution: post-view and post-click. This means any sale the user makes after viewing the advert or clicking on the advert, within the attribution window (typically 28 days after clicking and 1 day after viewing), is attributed to that advert.
GA, by contrast, can use a variety of attribution models, the default being last non-direct click. This means that if the user clicks on an advert and on the same device buys something within the attribution window (typically 30 days), it will be attributed to Facebook. GA doesn’t know about views of the advert.
If another campaign brings the same user to your site between the Facebook ad engagement and the purchase, this other campaign takes the credit as the ‘last non-direct click’.
So to match as closely as possible we recommend setting the attribution window to be ’28 days after clicking the ad’ and no ‘after view’ attribution in Facebook (see screenshot above) and then creating a custom attribution model in GA, with the lookback window at 28 days, and the attribution ‘linear’
The differences typically come when:
- a user engages with more than one Facebook campaign (e.g. a brand campaign and a re-targeting one) where the revenue will only be counted against the last campaign (with a priority for ads clicked vs viewed)
- a user clicks on a Facebook ad, but then clicks on another advert (maybe Adwords) before buying. Facebook doesn’t know about this 2nd advert, so will attribute all the revenue to the Facebook ad. GA knows better, and will attribute all (or part) of it to Adwords.
5. Facebook cross-device tracking
The main advantage Facebook has over GA is that users log in to its platform across all of their devices, so it can stitch together the view of a mobile advert on day 1 with a purchase made from the user’s desktop computer on day 2. Here’s a fuller explanation.
By contrast, unless that user logs into your website on both devices, and you have cross-device tracking setup, GA won’t attribute the sale to Facebook.
6. Date of click vs date of purchase
In Facebook, revenue is attributed to the date the user saw the advert; in GA it is to the date of purchase.
So if a user clicks on the advert on 1st September, and then buys on the 3rd September, this will appear on the 1st on Facebook – and on the 3rd in GA.
7. The sampling problem
Finally, did you check if the GA report is sampled?
In the top right of the screen, in the grey bar, you’ll see that the report is based on a sample. If that sample is less than 100% it means the numbers you see are estimates. The smaller the sample size used, the larger the possibility of error. So in this example, a 45% sample of 270,000 sessions could skew our results plus or minus 0.2% in the best case.
As a rule of thumb, Google applies sampling when looking over more than 500,000 sessions (even if you select the ‘greater precision’ option from the drop-down menu).
You can check your own sample using this confidence interval calculator.
Altogether, there’s a formidable list of reasons why the data will never be an exact match, but I hope it gives you a way to optimise the tracking.
Please let us know if you’ve seen other tracking issues aside from these.