How to debug wrong channel attribution in Google Analytics

Question

I have a client with an online shop whose backend system show some 50000 transactions for the last year. Google Analytics show 48000 Transactions (all numbers are by way of example only) which is close enough.

According to the Google Adwords reports some 45% of conversions are generated by Adwords. According to Google Analytics only 30% of those conversions are CPC conversions. Since GA reports a (broadly) correct number it looks like a lot of conversions are attributed to the wrong marketing channel. The numbers of clicks in the Adwords report and the number of sessions initiated by Adwords roughly match up (with a difference of about 5%).

However performance does not suspicious in the other channels. If anything is wrongly attributed it must have ended up either in organic traffic or direct traffic (the other channels have not enough volume to contain the difference). There is no sudden shift of conversions rates between channels, so whatever is wrong has been wrong from the start.

Now, this sometimes happens when campaign parameters (or the gclid parameter from Adwords auto-tagging) are lost (e.g. in redirects). When testing I don't see any evidence of that happening, though.

I'm a bit at a loss now. How would I proceed to debug the problem ? Any ideas welcome.

Answer 1

I find that the Goals system in Analytics is not as reliable as we wanted. Same thing happen to other features and that is mainly because of the source of the tracking used by Google: Javascript.

This means that the Code needs to be fired up in order to count as a Goal and depending on the user's security settings or device, that might not happen from time to time. Don't get me wrong, Analytics is great and Javascript is close to the best way to do it, however it does has its flaws.

If I compare the pageviews of one of my websites between Analytics and Cloudflare (that uses DNS) the difference is almost 10% with the last taking the lead.

I found that the best way to track conversions is with the backend by creating a simple tracking system (that judging by the amount of traffic you have, it's worth to implement).

If the system is counting the conversions in what looks like correctly, the difference you have is your error margin and the main data is enough to take conclusions on which campaigns are performing better and which to optimize.

On a side note, Google has been upgrading some parts if its Tracking code, you might want to take a look at that, can be better for this case, here's their Changelog -> https://developers.google.com/analytics/devguides/collection/gajs/changelog

Answer 2

I think that your "problem" is different type of attribution.

Google Analytics uses last-click attribution, AdWords prefers first-click. So... if your client ever clicked an campaign ad created in AdWords and make a transaction in another session, it will be counted in AdWords as "adwords conversion". But in Google Analytics it will be conversion from that source, which was session providing transaction (direct, if your customer comes from bookmark, for example).

It is funny that Google decides to use two different ways of attribution, but it's good for some reasons.

In GA you've got report of multichannel conversion - it is good to understand how long could be customer path.