Why are my Facebook and Google Analytics data so different?

“Why is my Facebook account showing me such different numbers from my Google Analytics account?” The question is a common one among marketers getting started with Facebook advertising. Just when pixel deployment is complete, the site is QA’d, and media is pushed live, the real questions start. For some reason, Facebook is suddenly showing much higher conversion numbers than Google analytics is. “What’s going on?” your CMO wants to know.

Luckily, because the question is fairly common, many people have already been working out the answers. In general, there are three causes that often make up the difference:

  1. Attribution Modeling
  2. Tracking Issues
  3. Traffic Loss

You can pretty much eliminate causes 2 & 3 from consideration and just focus on attribution if the amount of website traffic Facebook is reporting is within 5%-10% of the sessions from Facebook that Google Analytics is reporting.

Remember that Facebook reports on multiple types of clicks, and not all of them drive to your website. For most formats, you’re better off using ‘Link Clicks’ rather than ‘All Clicks’, as link clicks tracks clicks to the selected destinations only. For Canvas ads, you may need to look at Outbound Clicks.

Attribution Modeling

Attribution is far and away the most common cause for a discrepancy in reports. Even if everything is correctly set up, Facebook and GA use very different models for attributing conversions to online channels and will report very different numbers.

Facebook defaults to a 1-day post view and 28-day post-click attribution window. This means that if a person saw your ad and bought within 24 hours, Facebook would count that conversion regardless of whether or not they clicked your ad. Similarly, if they clicked your ad, Facebook will count any conversion in the next 4 weeks.

GA defaults to last interaction, with that last “interaction” almost always being a click. Using this model, GA would only count a conversion for Facebook if a person clicked on a Facebook ad, went straight to the site, and made a conversion during that session. If the person didn’t click an ad, or clicked, but decided to come back later – perhaps via search or directly typing in the URL – GA won’t count the conversion for Facebook.

For most advertisers, Facebook tends to operate close to the top of the funnel, driving awareness, interest, and brand affinity. The time between a Facebook impression and an eventual purchase is, on average, much longer than that between a search and click on a Google ad.  This means it is unlikely for Facebook activity to drive people straight to a purchase. In turn, there is almost always another touchpoint (including organic search or directly typing in a URL) between Facebook and GA, so GA will count zero conversions for Facebook.

Below, you can see the differences between the types of activity Facebook will record a conversion for vs. those that GA will record a conversion for. As you can see, there are quite a few gaps between Facebook and GA.

Image showing the differences between Google Analytics (GA) and Facebook (FB) MeasurementFacebook’s model may seem optimistic, but GA’s model isn’t quite right either. Evidence from Croud’s client base suggests that the truth probably lies somewhere in the middle. In fact, some case studies have demonstrated even greater impacts from Facebook advertising than Facebook’s own attribution model suggests.

While a full discussion of attribution models, conversion deduplication, and media mix modeling are beyond the scope of this post, some readers may find it helpful to consider using GA’s “Assisted Conversions” to get a quick feel for what performance would look like “somewhere in the middle.”

Tracking Issues

If you’ve evaluated clicks to your website, as reported by Facebook, and compared them to GA sessions, and found a discrepancy larger than 10%, then there may be other issues at play.

The next most common cause are tracking issues. These occur when traffic is coming through to your site but isn’t being recorded properly. To diagnose these problems can take a bit more work than comparing attribution models.

In approximate order of importance, we recommend checking that:

  • All Facebook ads have tracking (e.g. UTM parameters) correctly applied
  • The Facebook pixel is correctly implemented with a conversion event
  • No landing page URLs are being redirected – server redirects often strip tracking parameters
  • The GA profile you are using does not have any filters in place
  • GA is firing in the page header on all pages
  • GA is not configured to override UTM parameters
  • No other site components, such as tag managers, are interfering with the GA tag firing

Traffic Loss

Finally, if clicks/sessions are too low to be differences in attribution models and you’ve checked the tracking issues above, the problem may be traffic loss.

Traffic loss occurs when a user clicks an ad but ultimately does not see the landing page load. The most common drivers for this are dead links in the Facebook ad or long load times.

Manually checking that the landing pages work, by copy-pasting the links from Facebook Ads Manager into the URL bar, or using a bulk link checker,  can assess dead links.

Long load times can be assessed by GA page timings or other tools that measure site speed. Although there are differences between user-side and server-side speed, any GA timings (server-side speed) over 5-7 seconds may be a traffic loss driver. This is especially true for mobile traffic, where users are much quicker to abandon a page that is slow to load.

There are, for both traffic loss and tracking issues, a few much rarer causes that arise occasionally, but these require deeper technical expertise to diagnose and fix, such as parsing server logs and writing javascript. If you’ve gone through the steps above and none of the common causes seem to be driving the issue, you may need to consider more in-depth technical audits of both your tracking implementation and your site.

If you’re looking for help working through the causes above, please get in touch!

by Croud
5 October 2017



Related posts