I’ve recently been looking at Google Analytics traffic sources in detail for one of our clients and decided it was time to finally get to the bottom of a reporting ‘discrepancy’ which has been bugging me for some time - cookie traffic.
As far as I’m concerned GA is a site-based Analytics tool, it picks up visitors on their arrival at the site, judges to the best of its abilities where they’ve come from, and monitors their behaviour from there. It differs in several ways from 3rd party solutions such as Dart, Atlas, and the search engines’ own offerings such as Google Conversion Tracking.
These latter tools monitor traffic from one, or in some cases a few, but not all sources of traffic. This means that if they drop a cookie and that same user comes back through another source they have no way of knowing this. To give an example:
A user clicks on a Google PPC listing and is picked up by Google Conversion Tracking (GCT), they visit the site, browse and then leave. They later click on a natural listing to come back to the site, at which point they convert. As this user has a GCT cookie on their machine GCT views this as a GCT conversion, in spite of the fact that their last click came from a different source.
Google Analytics is different
In theory Google Analytics (GA) is different. Because it picks users up on their arrival at the site, and as it is not associated with any particular source it can see where all traffic has come from. What this means is that we shouldn’t see any cookie based activity, because a user always has to come from somewhere, and GA should track that visitor based on their most recent source. To look at this based on the example above:
A user clicks on a Google PPC listing, visits the site and is picked up by GA. They browse and then leave. They later click on a natural listing to come back to the site, at which point they convert. As GA knows that this most recent click came from a natural listing it will be recorded as a natural visit, and a natural conversion, no cookie joy for Google PPC!
So what has always bugged me is how we can see visits from campaigns that have long since been paused, what we would normally assume to be cookie based traffic, but surely that shouldn’t be possible with a site-based analytics tool? This is where GA is different to all of the other analytics packages I’m aware of, and as you would imagine it favours PPC!
Google scratches its own back
The cookie activity is all down to one subtle difference - the way direct traffic is reported. When a user directly inputs the URL, or uses a bookmark they are picked up as direct, if they then return later via a PPC link (or any other source) they will be picked up as PPC (or that other source), as we would expect. However, if a user first visits through PPC (or any other source) and later returns through direct they will still be recorded as PPC (or that other source). Essentially direct will not overwrite any other source, so if a user returns multiple times through direct they will always be recorded as their previous non-direct source.
What this means is that when you see traffic from paused activity still being recorded in GA you know this is actually direct traffic. What you can’t see however is what proportion of your traffic from live activities is actually direct, and also you can never get a true picture of what your direct traffic volumes are - useful for measuring things such as brand strength and the impact of offline promotions.
I can see why Google have done this as to an extent it is more useful to see how a user first became aware of the site before they started directly inputting the url, and this also goes back to the ongoing debate about the last click model. That said, if I have a site-based analytics tool I expect it to tell me what the last referrer was, it is my decision from there if I want to try to attribute activity to anything other than this last referrer, I would rather know exactly what my results are telling me without any kind of skew - and be able to properly monitor each individual traffic source.
It is interesting how this overstates Google’s own contribution. I know they traditionally take existing approaches and standards in new directions, but if this was the traditional way analytics programmes worked I find it hard to imagine that Google would have launched a tool which gave attribution purely to the last referrer.