Attribution is a subject I’ve been obsessed with since I started working in digital media, so I am glad to see it turn into one of the hottest topics of 2017.
I had the chance to attend the largest gathering of SEM specialists in the world, HeroConf 2017 in Los Angeles, and the subject of attribution was omnipresent throughout most of the sessions.
Just a few days after the event, Google made an announcement to its partners about the use of the Last-Click attribution model, declaring that this type of model would soon be obsolete. Why change the default model we’ve used for years from one day to the next? Is optimizing for Last-Click In-Platform a thing of the past?
If you aren’t familiar with attribution models, I would suggest starting with this excellent resource.
The majority of the digital media industry in 2017 is still using the default model, which is Last Non-Direct. When you use standard Google Analytics reports (Audience, Acquisition, Behavior and Conversion) you are looking at reports based on Last Non-Direct.
This model is still entirely viable if you’re looking for an overall understanding of your performance. However, if you are using these reports purely to judge the performance of your channels and tactics, they might be leading you to make decisions that will actually hinder your campaigns. For example, you might cut an ad group because they weren’t performing in Last Click, without realizing that these ads were a valuable first point of contact with your brand for many users. There could be hidden treasures in your data!
It’s not for nothing that the default attribution model in AdWords has been Last Click for years. This model is true to the vision of Avinash Kaushik, who has always been prone to saturating the bottom of the conversion funnel before investing in the top of the funnel. This is a model that allows you to optimize for the keywords and tactics that close sales.
The question to ask, then, is this: do I only want to optimize for the things that close sales, or do I prefer to look at all the points of contact that might eventually lead to a sale?
There are currently three attribution models available on AdWords and Google Analytics that track the multiple points of contact that might generate a sale or lead. These Multi-Touch models have the advantage of considering the entire spectrum of points of contact in the Buyers’ Journey.
I invite you to take a look at the three attribution models in question: Linear, Time Decay and Position Based.
This model is particularly useful when you are looking to reach a new in-market audience.
When you really think about it, the majority of AdWords investments are made on generic campaigns that serve to introduce the brand to users who are actively searching for a product or service. To do that, we want to exclude users who have already visited the site, and exclusively target new users to maximize our investment. In this situation, optimizing for First-Click makes a lot of sense, because the primary objective of the campaign is exactly that: to fill the top of the funnel!
The user in question might do a search, click on an ad, visit the site, compare it with competitors, then finally convert. These touchpoints are just a few examples of a typical user journey.
To come back to our example, using a Last-Click model, no conversion value would be attributed to the generic campaign. Whereas if we approached the example with a First-Click model, we would be able to optimize the campaign to generate more first interactions, and thereby increase overall sales for this product or service.
Here’s an example of a user journey that includes far more than three or four interactions before a conversion.
Before jumping into a media campaign and setting goals for each channel, you have to choose the right attribution model.
To help you, here’s an infographic of the different attribution models, broken down by objective.
Unfortunately there is still no miracle recipe for choosing the right attribution model. In order to choose the best model for you, I would suggest looking at your various tactics and campaigns through the lens of different models. You can do this both through Google AdWords and Google Analytics.
Another important thing with attribution models is to choose the right analysis period. The analysis period represents how many days back the model will go to look for the first interaction. This is very important depending on your industry. Buying certain products or services, like a T-shirt, might take only a few minutes, while buying a car might happen over a period of weeks, even months.
Once you’ve analyzed your data and landed on a particular model, you can change the attribution model directly in AdWords.
During the conference, we learned that Google had launched a new attribution model, Data-driven attribution that, using Machine Learning, can dynamically attribute the right conversion weight to each player. However, there are some fairly major pre-requisites for using this new model. In order for the algorithm to have enough information to work, it requires 15,000 clicks and 600 conversions over the previous 30 days.
In short, there is no perfect model, which means it’s important to look at your attribution model in relation to your marketing objective. Looking at your data from a different perspective might help you see that your campaigns are more profitable than you had imagined!