Analytics
5K members online now
5K members online now
Learn how to set up goals and Ecommerce tracking, verify accuracy of attribution traffic sources, and discuss funnel visualization, Multi-Channel Funnels, the Goal Flow Report, and Enhanced Ecommerce
 
Guide Me
star_border
Reply

Making Sense of Conversion Attribution

Visitor ✭ ✭ ✭
# 1
Visitor ✭ ✭ ✭

Hi, 

 

I have Adwords import from Analytics the transactions as conversions. They end up not being the same and I understand that Adwords and Analytics apply different rules like attribution, cookie/lookback window lengths and timestamps. In trying to understand this even better, I'm laying all the data for one day side by side to see which transactions Analytics will count as source/medium google/cpc and which ones Adwords would claim as being responsible for. 

 

As I understand, Analytics uses by default the Last non-direct click attribution model. Adwords tries to find any campaign click in the conversion path and claims it (with the last Adwords click if there are more than one). 

 

So far so good, but then it gets confusing. 

 

In Analytics I pick one day and then I:

 

- go to Acquisition > All Traffic > Source/Medium and of 418 transactions that one day, 142 get attributed to google / cpc. 

- go to Acquisition > AdWords > Campaigns and see 142 transactions of 418 attributed to AdWords

 

142/418 seems the consensus as far as Adwords sales go...

 

- go to Conversions > Multi-Channel Funnels > Assisted Conversions and set Primary Dimension to Source / Medium and we see google / cpc clocking 126 "Assisted Conversions" and 98 "Last Click or Direct Conversions". This 98 is the first number I don't get. "Last Click or Direct Conversion" means "The number of conversions for which this channel was the final conversion interaction." So why is it 98 here and not 142?

 

No I go dig a little deeper by exporting and manipulating some reports. For example I:

 

- go to Conversions > Multi-Channel Funnels > Top Conversion Paths and set the Primary Dimension to "Source/Medium Path", up top I ensure I only select the right Conversion Type: Ecommerce Transaction and Path Length 1 or more. Type All and Lookback 30 days. The summary says 418 conversions again, correct. I get 241 unique source/medium paths. I export this to CSV and count which would score the Analytics attribution for google/cpc and which would match the Adwords attribution. 

 

Take for example a simple path, just google / cpc. On this day, 47 conversions match this path. According to Analytics' Last non-direct click attribution model, indeed 47 would be matched to Adwords. According to Adwords' any-Adwords-click model, these also get matched to Adwords. That one is easy. 

 

Now take a path like google / organic > google / cpc. This is also easy, both Analytics and Adwords will claim this as an Adwords sale. 

 

 

Then take a path like this: google / cpc > (direct) / (none) > awin / affiliate Good for 2 conversions. Analytics will look at the last step of the path and say this belongs to the Awin Affiliate program. Adwords will recognise the first step and claim it as a conversion. 

 

Continue counting them like this and of 418 conversions, I count 98 that Analytics would attribute to google/cpc and 197 that Adwords would claim its stake. 98 and 197. Elsewhere Analytics attributes 142 to Adwords. In another spot, on the assisted conversion report, this 98 number also surfaced. So where does 98 come from, and how does 197 make any sense in all this?

 

Let's look at it from another angle. I go to Conversions > Ecommerce > Transactions and set as Secondary Dimension the source / medium. Again, the total of 418 transactions seems consistent. Download this, count how many have secondary dimension google / cpc and I do get 142. So this agrees with other reports. But how does Analytics here pick google / cpc as the secondary dimension for that transaction? According to the Last non-direct Click attribution model? When I manually check that, I only count 98. 

 

I suppose one could place two orders in one session, but I highly doubt some 50% of the people do this. 

 

So my problem is that Google Analytics thinks 142 of 418 transactions belong to google / cpc in one place, but 98 in another place. But I cannot verify these totals manually. It gets worse when we import this as a conversion into Adwords, and Adwords then sees just 129 coming in from Analytics. 

 

I'm trying to get my head around the differences by manually seeing it, but I'm not getting there. How can I retrace the steps Analytics takes to attribute these sales to Adwords?

 

Thanks for bearing with me in this long story. 

1 Expert replyverified_user
1 ACCEPTED SOLUTION

Accepted Solutions
Marked as Best Answer.
Solution
Accepted by topic author Johan T
May 2017

Making Sense of Conversion Attribution

Rising Star
# 4
Rising Star

Hey Johan,

 

I don't think a good article exists but it comes down to how they treat direct traffic; in standard reports; direct is ignored if a previous touch was already credited to something else.  The other inherit problems are the limited lookback and the fact that these really aren't users -- it's a portion of the users journey the portion that happened on the converting client. 

 

Personally, I like using the AdWords conversion pixel with an appropriate cookie length for the business and optimize off of that and GA to look at the 93 days I can see in MCF to see assists so that I get a better feel for whats happening.  When client's have the appropriate technology, we can layer in custom dimensions with CRM data to give better insight and a longer lookback. 

 

Google Attribution could help immensely but we still don't have specs on lookback, cross device etc.

 

Regarding your transactions, you'll always have some that you wont track in GA but if you get to 5% of transactions; a serious audit is needed.

 

 

Best,

 

Theo Bennett

Analytics Evangelist at MoreVisibility | Contact Me
Connect on LinkedIn

View solution in original post

Making Sense of Conversion Attribution

Rising Star
# 2
Rising Star

Hi Johan,

 

First step to MCF nirvana is to understand that MCF and Standard GA reports have different attribution and Accept the different totals and move on.  Smiley Happy

 

Second is that depending on your lookback and your cookie length, you can be looking at different things. 

 

For example, I could have come to the site 99 days ago via Google CPC, then came today via Organic and bought. 

 

Here is how the conversion will show:

 

GA : google/organic

AdWords : No Credit (depending on cookie length using AdWords conversion tracking)

GA MCF: Google Organic with No Credit for Google AdWords (outside of maximum lookback)

 

Throw in the real case where multiple touches occur and each one would get an assist in MCF and you can start to see why it varies so much.

 

Regarding retracing the steps you really can't without some very sophisticated tracking of campaign touches in your CRM using custom dimensions and some custom JS.

 

The good news is that Google just made Google Attribution available for free; well they've announced it (two days ago) and within the next year we will be able to do more complex and real attribution by linking with AdWords and GA and have Attribution sort it all out.

 

Best,

 

Theo Bennett

Analytics Evangelist at MoreVisibility | Contact Me
Connect on LinkedIn

Making Sense of Conversion Attribution

Visitor ✭ ✭ ✭
# 3
Visitor ✭ ✭ ✭

Thanks Theo, 

 

Google does a great job with thousands of support articles, yet still some aspects feel like a black box. 

 

You say MCF differs in attribution from other reporting, I didn't know that. Can you please elaborate on the exact model, or point me to an article that has already explained the difference? 

 

The frustrating thing is, at the source of the e-commerce data, already not every single sales order gets logged as a transaction. Some never make it back to the Thank You page, lingering within the Payment Solution Provider process. Others have some form of blocking in place. So some go 'missing' right there and then. Next, we see three different numbers; in AdWords - despite being imported from Analytics, in Analytics standard reporting, and in Analytics MCF reporting. 

 

To work out some approximation towards true ROI and ROAS, we need to decide which numbers we trust most. A black box doesn't offer enough transparency to build up that trust. 

 

That new Google Attribution might be a step closer to this understanding and trust, thanks for mentioning it. 

 

Regards, 

 

Johan

Marked as Best Answer.
Solution
Accepted by topic author Johan T
May 2017

Making Sense of Conversion Attribution

Rising Star
# 4
Rising Star

Hey Johan,

 

I don't think a good article exists but it comes down to how they treat direct traffic; in standard reports; direct is ignored if a previous touch was already credited to something else.  The other inherit problems are the limited lookback and the fact that these really aren't users -- it's a portion of the users journey the portion that happened on the converting client. 

 

Personally, I like using the AdWords conversion pixel with an appropriate cookie length for the business and optimize off of that and GA to look at the 93 days I can see in MCF to see assists so that I get a better feel for whats happening.  When client's have the appropriate technology, we can layer in custom dimensions with CRM data to give better insight and a longer lookback. 

 

Google Attribution could help immensely but we still don't have specs on lookback, cross device etc.

 

Regarding your transactions, you'll always have some that you wont track in GA but if you get to 5% of transactions; a serious audit is needed.

 

 

Best,

 

Theo Bennett

Analytics Evangelist at MoreVisibility | Contact Me
Connect on LinkedIn

Making Sense of Conversion Attribution

[ Edited ]
Visitor ✭ ✭ ✭
# 5
Visitor ✭ ✭ ✭

Thanks again Theo, 

 

For posterity, this article goes some way to explain this difference in attribution:

 

http://online-behavior.com/analytics/multi-channel-funnels 

and

https://support.google.com/analytics/answer/1319312?hl=en&topic=1191164&ctx=topic#Direct 

 

Interesting that you choose the AdWords built-in conversion pixel. To me it feels like it is overly generous in attributing the sale to AdWords. Which leads to an overly optimistic notion of ROAS and hence ultimately a lower ROI. Importing conversions from Analytics may lead to the PPC people having to work that little bit harder to earn their stripes, and be more decisive on so-so campaigns. 

 

For this client, indeed the focus must now be on reducing the non-tracked transactions. Sub 5% is what I see in other accounts too so this one sticks out indeed. 

 

Thanks for taking the time to read and answer my long post. 

 

Regards, 

 

Johan

Making Sense of Conversion Attribution

Rising Star
# 6
Rising Star

No worries Johan!  Yes I choose the AdWords Pixel (I don't; however, manage campaigns) as even if it is a "hungery" conversion indicator, it is used for automated bidding (CPA targets etc) which I think should be used sparingly etc. Even with it's flaws, it's still a consistent trend indicator and your overall analysis can still be done in GA using MCF.  Let's hope the new Google Attribution (free version) helps shed light on this as there are still many holes due to users not really being users in GA.  Smiley Happy

 

-Theo

Analytics Evangelist at MoreVisibility | Contact Me
Connect on LinkedIn

Making Sense of Conversion Attribution

Visitor ✭ ✭ ✭
# 7
Visitor ✭ ✭ ✭

Sounds like part of your issue here may be that the anchor date for Adwords reporting is the date of add click where the date reported in GA is the actual conversion date