by Louise

30/05/2018

AdWords, DoubleClick, Internet Marketing, Online Advertising

An Intro to Data-Driven Attribution in Google Search

‘Last click attribution is dead!’ is something we have all heard over and over again. But why? And if last click attribution is dead, then what attribution model should we be using?

Let’s start from the beginning – Attribution

Whether you’re looking to generate leads, or drive transactions online – it’s likely you’ll be measuring the success of your efforts by tracking conversions. Attribution is a set of rules that gives credit to touchpoints that led to a conversion. There are various attribution models that can be used to distribute this credit within an AdWords account:

  • Last Interaction – giving all the credit to the last touchpoint. They closed the sale after all!
  • First Interaction – giving all the credit to the first touchpoint. Because without that initial introduction, the sale might not have taken place.
  • Linear – sharing the credit across every touchpoint a user interacts with before purchasing.
  • Time Decay – the closer the touchpoint is to the sale, the more credit given.
  • Position-Based – the introducers and the sale closers getting the most credit, but the touchpoints in between are still be given a small amount.
  • Data-Driven – the new(ish) kid on the block. More on this later.

Looking for a more in-depth explanation for each? Click here to visit the AdWords article covering each model in more detail.

Really want to get stuck into the origin of attribution? Search for Fritz Heider and ‘The Psychology of Interpersonal Relations’ 🤓

So why has last click always been the go-to attribution model?

Is it because this is the default setup in AdWords? Possibly. Perhaps marketers have viewed the last touchpoint as the one that pushed a user over the line and eventually got that sale.

The same can be said for football; we all remember who had the last kick that led to a goal, but tend not to dwell on the journey that took place up until that point.

In reality, customers research, compare products, read reviews, and seek recommendations across a number of different touchpoints – each of which has an impact on the user’s final decision. As a result, we need to consider each user journey and look at the impact each touchpoint had.

Cue – data-driven attribution

Data-driven attribution uses data from your AdWords account to determine which ads, keywords and campaigns have the greatest impact on your business goals. It then distributes credit to each, depending on how much it contributed to driving the conversion.

How is the credit calculated?

Google compares the click paths of customers who convert to the paths of those that don’t to identify patterns among those clicks that lead to conversions. This allows the model to highlight specific steps along the way that have a higher probability of leading a customer to complete a conversion.

Here’s an example to put this into context:

  • The original path

200 users are exposed to the following path – 10 of which convert
Conversion rate = 10/200 = 5%

Attribution Path: Generic - Brand - Conversion
The above path is then compared with paths where one element is missing, for example, the brand keyword:

  • The original path, without brand

200 users are exposed to the following path – 8 of which convert
Conversion rate = 8/200 = 4%

Attribution Path: Generic - Conversion
When the brand keyword is removed, the conversion rate drops from 5% to 4%, giving the brand keyword a weight of 1%.

Then we look at paths where the other element is missing, for example, the generic keyword:

  • The original path, without generic

200 users are exposed to the following path – 4 of which convert
Conversion rate = 4/200 = 2%

Attribution Path: Brand - Conversion
When the generic keyword is removed, the conversion rate drops from 5% to 2%, giving the generic keyword a weight of 3%.

Using the above, we can see that the credit ratio for the generic keyword: the brand keyword would be as follows…

3 : 1
Or
75% : 25%

We can then translate the above credit ratio into conversion credit and apply it to the original path, where 200 users were exposed to the following path – 10 of which converted:

Total conversions: 10

Credit ratio

  • Generic keyword: 75%
  • Brand keyword: 25%

Conversions

  • Generic keyword: 7.5
  • Brand keyword: 2.5

And there you have it – an attribution model that is specific to you as an advertiser and that evolves with your customers!

It’s not perfect – I’m not sure anything can be when the decision to convert happens in a user’s mind. But it IS using data which, for us, is a very exciting step in the right direction.

Data requirements

As you’ve probably guessed from the above, the model requires a good amount of data to work with to be able to successfully create the data-driven model. As a result, there are a few thresholds that an account needs to meet in order to be given the option to switch:

  • 15,000 clicks on Google Search in 30 days.
  • A conversion action with at least 600 conversions within 30 days.
  • The above has been maintained for 30 consecutive days before the option is shown in an account.
  • Once the model has been applied, the account will need to generate at least 10,000 clicks and 400 conversions within 30 days to continue using it.
  • If your data drops beneath this threshold, you will receive an alert.
  • If the low data continues for 30 days, the attribution model for that specific conversion action will be switched to linear.

Keen to get started?

Check out the Google Partners video for more information about getting started with data-driven attribution: https://www.youtube.com/watch?v=74y0kRslmBA

Or visit the Google help centre: https://support.google.com/adwords/answer/6394265

Want to find out how to set this up using floodlight activities in DoubleClick Search?

Get more information from Google’s DS support: https://support.google.com/ds/answer/7015524

And find out how this can incorporate Data-Driven attribution in DS reporting here: https://support.google.com/ds/answer/7016041

 

Any questions? Get in touch!

Comments

Leave a comment

Your email address will not be published.