Google Drops All The Attribution Models That Make It Look Bad

On April 6th 2023, Google announced that four attribution models would be sunsetted across both Google Ads and Google Analytics starting May 2023.

Advertisers will be switched to Data-Driven Attribution (DDA) automatically, though Last Click (the original default) will still be available. The models Google’s dropping are First Click, Linear, Time-Decay, and Position-Based.

Ginny Marvin (@adsliaison) provided some context in a series of tweets:
“Use of rules-based attribution models has dropped significantly with the introduction of DDA 3 years ago. Fewer than 3% of conversion actions in Google Ads use them now. DDA has the broadest adoption & is available to all businesses, with no data requirements.”

Google has been pushing DDA for a while now, and most advanced advertisers already use it for optimization. The underlying tech is sophisticated and used to cost tens of thousands of dollars before Google bought Adometry and made it free. Machine learning based approaches like DDA have become more necessary in the wake of user privacy initiatives like the EU’s GDPR legislation and Apple’s App Tracking Transparency banners in iOS14.

However…

Marketers like me who have been burned before, couldn’t help but note that the models Google decided to drop were the ones that made it look bad. Google still makes about 80% of its revenue from advertising, with the vast majority of that coming from search. Bottom of funnel channels like search are overwhelmingly likely to get the last click before someone buys. This gives Google Ads a disproportionate share of the credit when using a last click model.

Data-driven attribution is a good model, but it’s also a black box that once cost me over $40,000. Applying machine learning to attribution has ultimately been a good thing for marketers, but increasing lack of transparency is a concerning trend. While most marketers didn’t use these other models for optimization, many of us compared them regularly to find insight into potential issues, and to form better assumptions about the relative importance of various channels.

What Are The Different Attribution Models?

Here’s an example I’ve adapted from the example Google provides in their documentation. I’ve changed it to incorporate other channels, so you get a sense of what we’re losing:

Say you own a fashion boutique called Chic Couture in New York City. Someone views a TikTok video featuring an influencer wearing your clothing, and clicks through a link in the comments to your blog post announcing the collaboration. Later in the week they click on a retargeting ad on Instagram, but are too busy to get their credit card out. A few days later they can’t quite remember your brand name, but find your site after performing these searches: 'fashion boutique nyc’, and then 'chic couture nyc’.

• In the 'Last click' attribution model, the last keyword, 'chic couture nyc’, would receive 100% of the credit for the conversion.
• In the 'First click' attribution model, the first touch, TikTok, would receive 100% of the credit for the conversion.
• In the 'Linear' attribution model, each touch leading up to the sale (TikTok, Instagram and 2x Google) would share equal credit (25% each touch) for the conversion.
• In the 'Time decay' attribution model, the keyword 'chic couture nyc’ would receive the most credit because it was searched closest to the conversion. The TikTok visit would receive the least credit since it occurred first. Instagram would also receive credit based on how close the timing was to the conversion.
• In the 'Position-based' attribution model, 'designer clothes nyc’ and 'chic couture nyc’ would each receive 40% credit, while TikTok, and Instagram would each receive 10% credit.
• In the 'Data-driven' attribution model, each keyword and channel (TikTok and Instagram) would receive part of the credit, depending on how much they contributed to driving the conversion.

As you can see, all of the models that Google is dropping are pulling money away from Google Ads. Upper funnel channels are important, because they help generate new demand for products, instead of advertisers all competing for the same small pool of people ready to buy. This move will make Google more money, because it’ll make them look better in reports to marketers who don’t know any better, but it’ll be harmful to true incremental value.

The Data-Driven model is the best of the bunch, but it’s also completely unaccountable – we don’t get to see how it works, what rules its following, or why it took certain actions. We’re destined to rely more and more on AI not just for attribution, but also for optimization and generating creative, but that doesn’t alleviate us of our responsibility to make sure we’re in the driving seat, and making the right decisions with our marketing budget.

What Can Marketers Do?

The solution is to not allow Google to grade their own homework, and invest in external validation on the incrementality of your advertising campaigns. Multi-touch attribution has become less reliable anyway, given the backlash against invasion of privacy, and should be taken with a pinch of salt. More privacy friendly techniques like Marketing Mix Models and Incrementality Testing (as offered by Measured) have been getting easier to implement, and are filling some of the gaps in attribution marketers are facing. You shouldn’t blame Google for acting in their own best interest, just make sure you’re acting in the best interests of your company and taking steps to verify what Google’s reports are telling you.

With Google's decision to remove certain attribution models, it's crucial for marketers to explore alternative solutions to ensure they're making data-driven decisions. One such solution is offered by Measured, a platform that provides Incrementality Testing to help you gain insights into your marketing performance. These privacy-friendly techniques can fill the gaps left by traditional attribution models while still delivering actionable information. Don't leave your marketing success to chance – take control by understanding the true value of your campaigns. To see Measured in action and learn how it can benefit your business, click here to visit our demo page and request a personalized demonstration.