Are you ready for the attribution changes coming to Google AdWords?
There is an old saying: the only constant in life is change! Word on the street and from Google is they’ll stop supporting last click attribution sometime this year. This means advertisers will have to opt into one of several other attribution models available in Google.
In this article, I’ll provide some general commentary on attribution as well as an overview of the different models available in Google AdWords.
What is an attribution model?
According to Google,
“an attribution model is the rule, or set of rules, that determines how credit for sales and conversions is assigned to touchpoints in conversion paths.”
Currently, the default in Google AdWords is last-click attribution. Last-click attribution gives the click just before the conversion event (like a purchase) 100% of the attribution credit. Every business has different marketing objectives so tailoring attribution models to those specific goals just make sense.
General pointers on attribution
To get started, here are some general pointers on attribution:
- Changing attribution models doesn’t change actual account performance, just your perception of it based on changes to what “counts” as a full or partial conversion. The various models report differently on existing account data so you are free to change models, as long as you inform all stakeholders of the major cosmetic changes and changes that may come with the change.
- A new attribution model won’t muck up your account. You can go to the attribution models page to view what the various models say about your data, without actually making a change to your default reporting.
Below is a screenshot and you get to the tool by going to conversions > attribution > model comparison tool in Google Analytics.
Rumor has it Google is considering making its darling, the data-driven attribution model, the account default. It will be up to the advertiser to opt into other attribution models. This is similar to how Google automatically opts you into ads that rotate for clicks, users searching in, around, about, etc.
In the sections below, I cover other options for attribution in Google AdWords.
Another option — one I won’t cover fully here — is a function in Google Analytics that gives different weight and credit to different channels like paid search, paid social, email marketing and direct channels. A simple example of this would be to assign 25% to each channel.
It’s interesting that Google’s new attribution technology may now be seamlessly baked into Google Analytics (GA). We have to “adopt” a model, rather than have a separate tool that watches the “watchers”.
Different attribution models
Below is an overview of the five different attribution models that you can choose from in AdWords. They are:
- Linear model. With this, every touch point that contributed to the conversion gets the same score. The first click gets the same amount of credit as the last click. Google seems to apply very partial credit in some cases, as little as 0.1 of a conversion. Does this mean there were potentially ten interactions by that user? Or does Google apply some weighting, even with the linear model? Google’s documentation on this isn’t extensive. This model is useful for any company that wants to sprinkle as much credit as possible around to any keyword that had a role in the user’s consideration process towards a conversion, so they can reduce the number of uninformative “zeroes” in the conversion stats. This can be especially important when we’re dealing with highly relevant, but low volume, long-tail phrases.
- Time decay model. With this, the touchpoints closest in time to the sale or conversion get most of the credit. The keywords consumers interacted with within a few hours of conversion would get the largest weighting. This model is the most similar to last-click and would be considered the most “conservative” change to an existing account. This is an excellent option if you want the same type of attribution that you’re getting currently with a last-click attribution account.
- Position-based model. With this, 40% credit is assigned to each the first and last interaction, and the remaining 20% credit is distributed evenly to the middle interactions. I’ve never heard of anyone using this model, but if they did, they’d be looking to move some credit away from the last click, but not give too much credit to repetitive searching in the middle research phase, in cases where consumers really dithered — but would never have done so, perhaps, without the power of that very first interaction. This model is actually very clever.
- First-click model. With this, 100% of the credit is given to the first touch point. This is generally used for when companies are looking for growth and focuses on new user/customer acquisition. An example of this would be a company whose goal was to introduce their offering to new prospects so they can remarket to them or place them on an email list and sell them from there.
- Data-driven attribution model. Data-driven attribution is the most black box out of all of the attribution models. It analyzes various data points to determine what the specific weighting should be when a conversion occurs. It redistributes credit in favor of converting ads and associated keywords, ad groups, and campaigns.
According to the Google AdWords Blog:
“DDA (data-driven attribution) is different from rules-based attribution models. It uses your account’s conversion data to calculate the actual contribution of each search ad click along the conversion path… The model observes what your customers do before converting, and what they do when they don’t convert, to measure what’s important. Using Google’s machine learning, the models continue to improve over time”.
Note: advertisers have to qualify to use this type of attribution.
Google claims there’s a 5%-10% increase in conversions from data-driven attribution and the company says Ford recently saw a cost-per-click (CPA) decrease of 25% with this model.
I’ve come to expect a CPA increase (of 10% to 20%) from using Google’s automated products, so be prepared.