Think about the last time you bought something online. It isn't likely that you bought that product or service immediately. It is more likely that you were impressed by a link or advertisement and visited a website, then you exited that website and you did some research before purchase. It is important for digital marketers to know which touchpoints during this journey convinced you to purchase.
If you are in the digital marketing industry, you know that everyone talks about campaign optimization. But before campaign optimization, you need to assess the value of your campaigns. Two main questions that arise when you are trying to understand the conversion journeys are:
- How to attribute conversion success to marketing spend?
- Where to spend the next marketing dollar?
How do you answer these questions?
Google Analytics has some methods to help you. The common methods are:
1. Last Interaction/Last Click Attribution model.
All standard reports in Google Analytics give one hundred percent of all the credit to the last touchpoint. This is easy to understand, easily implemented, and it might work for some emotional purchases which don't need any research or are simple “one touch purchases” . But, you might look at the purchase data and see some customers bought your product directly. Let’s think about that. What if those purchasers who bought directly, had visited your website before via email, search ads, organic search and so on. We should figure out a way to give credit to these channels as well.
2. Last Non-Direct Click Attribution Model.
All the direct channels are ignored and all the credits are assigned to the last channel that the customer clicked before conversion. This model is imprecise and ignores all the channels which might deserve some credit.
3. Last AdWords Click Attribution Model
The Paid Search channel click receives 100% of the credit for the conversion. This model overvalues the adwords' importance. Your customers might be convinced to buy your product via other campaigns rather than the last Adwords they clicked.
4. First Interaction/First Click Attribution Model.
This model is the reverse of last click and gives all the credit to the first click rather than to the last click. But, if the first channel was that perfect to deserve the credit why do customers bother to enter your website via other channels.
5. Linear Attribution Model.
Each touchpoint in the conversion path shares equal credit for the conversion. However, this is not true, you should understand which channels give you more value compared to others.
6. Time Decay Attribution Model.
The touchpoint nearest the conversion would receive the highest credit. and the touchpoint prior to that will get less credit. This can be less incorrect than previous models ( well, at least this one does not run counter to common sense!) but still it isn't precise.
7. Position Based Attribution Model
In this model, 40% credit each is assigned to the first and last interactions, and the remaining 20% credit is distributed evenly to the other interactions. This model, is one idea that might or might notwork for particular types of businesses.
8. Data-Driven Attribution Model
Finally, get rid of the assumptions! If you are a happy owner of Analytics 360 (formerly Google Analytics Premium) Google will calculate your real Attribution model based on your data. The model is only valid for a week and that's the beauty of it: it's dynamic. Google's Data-Driven Attribution Model uses your Default Channel Grouping to attribute impact on conversion. You can also Download the full model and process data sa you need it using Excel or GSheets.
The above mentioned models are commonly used because of their simplicity. But, I doubt any of marketers believe that this is the best model that they can use. There must be some better solutions to attribute success fairly to different touchpoints . Perhaps a model based on analysis of the underlying data can help? A data-driven model that uses statistical analysis. But the key question is what kind of statistical analysis? If you ask statisticians, they will say it seems straightforward. Why not use a simple regression model? But once they become more familiar with digital marketing, they immediately find that it is not a simple problem and the data is not available as they would wish. Most of the time they only have access to detailed data about those customers who converted; and even if they have detailed data about those who didn't convert, they do not know how likely customers are to convert in the future.
So, what is the solution? I think most of you believe your answers will eventually come from Google. I agree with you and encourage you to read this blog post.