< Back to previous page

How to pick the best Attribution Model for your AdWords campaigns

Google has recently announced an expanded beta testing of AdWords Attribution Modelling.  While attribution modelling is a staple part of data analysis, it may be unfamiliar to some AdWords marketers.

If you are suddenly seeing attribution settings in your AdWords account and want to know how to make the most of it, this is the post for you.
The attribution problem
When you look at your conversions column in AdWords, it shows you all the leads or conversions you have obtained from your campaigns, ad groups, ads, keywords or placements.  AdWords allocates these conversions based on a set of criteria called an Attribution model.
The default model is "Last interaction", which simply means that if a customer clicks on an AdWords ad, and then converts on your website or app, AdWords will report a conversion coming from that particular ad/keyword. In many cases, this makes perfect sense as the customer converted after clicking that ad so it must have been pretty convincing.
However, in reality, there may have been numerous events that contributed to the customer converting. As an example, they may have gone through the following process:
  • Read about your product/service on a blog post
  • Seen a display ad for your product/service while researching
  • Clicked on an ad to view product/service details as part of research
  • Viewed a product/service tutorial advertised on YouTube
  • Searched for your brand name – and clicked an ad to visit and purchase
In this scenario, the last click was clearly supported by the previous steps, but if you looked at the conversion stats, it is only the last campaign that gets credit for it.
When making decisions on your marketing spend, it will look like your YouTube and display campaigns are not performing; but when you reduce your budgets on these campaigns, your search campaigns also drop in performance because they aren’t getting the 'warm' leads that were being funnelled through your other channels.
What other models are available?
Rather than looking at the last click or interaction, you will want to attribute some value to all the touch points along the way, but what are the alternatives?
First interaction – This is the opposite of 'last interaction', where the first touch-point is assigned the conversion value. This model is great if your business focuses strongly on video or social media to engage people and generate leads, but it still ignores a lot of the 'in-between' steps.
Linear – This model allocates an equal amount of value to each touch-point on the way. In the example above, there are 5 steps, so 20% of the conversion value would be assigned to each step. This is a more equitable approach; allowing each channel to share equally in recognition.
Time decay – Time decay spreads out conversion value similar to linear, but assigns a larger percentage to the more recent touch-points, and less to the earlier touch-points. This model attempts to assign more value to the more recent steps in the funnel, based on the idea that the more recent touch-points have been more compelling than the earlier steps, otherwise the customer would have converted earlier.
Position-based – This model assigns 40% of the conversion value to the first and last interaction, on the basis that those clicks are most critical to converting clients. The in-between steps then share the remaining 20% equally.
Data-driven – This model attempts to distribute credit for the conversion across the various steps based on data available. This model is the best option, but you need enough data to analyse, and ideally, your own data analyst to make sense of that marketing data relative to your business model, and other customer behavioural data that you collect.
Which model should you use?
No single model is better than others; if there were, there wouldn't be all these options to begin with. Trying to decide which model is right for your company gets a little complicated, but never fear, I have an answer for everyone!
This is all too much!
Some businesses, especially small businesses who self-manage their AdWords campaigns or have in-house staff who manage campaigns as part of a broader role, will struggle with choosing a model.
If this is you, I recommend making an educated guess. Have a look at your client journeys in your Analytics account and see where those touch points are. Then choose either linear, time decay or position-based models. Whatever you select, it is better than the last interaction model, so you are making an improvement.
Not enough data
Data-driven is the best choice, but maybe you don't have enough data to enable that option. In this case, get your data analyst to have a look at your user journeys and make a recommendation. You may even want to assign different models for different types of conversion.
You may also want to try 2 or 3 models to see which one fits the data best. Your data analyst will be able to guide you with this. If the marketing agency you are working with doesn't have a data analyst, consider contracting an external consultant for this specific purpose.
If you have the data and resources, you should consider investing in a thorough analysis to develop a custom model. The data-driven option that Google provides is great but is limited to what data they collect. If your customers have touch-points that AdWords does not record (such as Facebook, email campaigns, and other sources), then the model AdWords uses is incomplete.
A data analyst can collect AdWords data, as well as your CRM data, website analytics data, and other sources of data to get a complete picture of your customer journey. The model should also be reviewed regularly as the ideal attribution may change over time.
Final thoughts
Many businesses over-invest in their 'direct-response' campaigns, under-valuing the brand awareness and engagement campaigns that feed leads into your conversion funnel.
Choosing any of the attribution models I have mentioned above will give you a more holistic view of which campaigns contribute to your company's profitability.
At Metrixa, we have invested heavily into our technology and specialist resources to ensure that all our clients’ campaigns have custom attribution models, which are regularly reviewed to ensure we have the most accurate understanding of campaign performances.
If your accounts have been green-lit for the attribution model beta, get onto it immediately! Or better still, talk to us about integrating all your data and building a custom data model for your conversion attribution.