Table of contents

    Google Ads is moving toward more automation and less autonomy for advertisers at the same time. In many cases, the push to adopt newer, more “simplified” features means discontinuing older ones that everyone is used to. Nomen omen, Google Ads itself for many years urged advertisers to use attribution models, which it is now abandoning.

    Attribution is a tool that marketers are slowly realizing must be a major part of their marketing analysis. According to Think with Google already in 2017. 76% of all marketers said they have or will have marketing attribution in the next 12 months.

    What can already be accepted officially is the discontinuation of first-click attribution models, linear models, time decomposition models and position-based attribution models. According to Google Ads, data-driven attribution was already the most commonly used model to automatically determine rates in April 2023. Similarly, less than 3% of conversions in Google Ads were measured using first-click models, linear models, time decomposition or position-based models. Are we losing anything in this case?

    google attribution models

    Existing attribution models in Google Ads

    When customers come close to making a conversion, they are likely to encounter multiple ads or several sources of traffic. Attribution models assume that you’ll be able to model what share of conversions each ad interaction should get.

    Statistically, these models can provide deeper insights into ad effectiveness and help you make improvements in media planning, ad optimization and ROI counting. There are three main benefits to using attribution model data:

    The data is also used in the following ways.

    1. The model will allow you to identify the main points of contact, their contribution and impact on the final conversion.
    2. With the model, you will recognize which ads work better at the first stage of the purchase funnel, and which ones are more apt to close the sale.
    3. The model will allow you to identify which ads work better at the first stage of the purchase funnel, and which ones are more apt to close the sale.
    4. At the end of the day, you will get a more complete picture of marketing analysis and move away from zero-one analysis of whether a channel sells or doesn’t sell.

    Until recently, several completely different models were available in Google Ads:

    • Last click: the share of conversion is attributed to the last clicked ad and the corresponding keyword.
    • First click: the share of conversion is attributed to the ad that was clicked first and the corresponding keyword.
    • Linear: the share of conversions is distributed evenly across all ad interactions along the path, for example, for five interactions each ad received 0.20 conversions
    • Time distribution: the share of conversions is distributed according to a 7-day half-life, in which a higher share is attributed to ad interactions that occurred closer to the conversion. In other words, the closer to the purchase the lower the share
    • Based on item: share distribution: 40% of the share for both the first and last ad interaction and corresponding keywords, and the remaining 20% is distributed among other ad interactions on the path.
    • Based on data: allocates conversion share based on previous data for that conversion-causing activity. It uses data from your account to calculate the actual share of each interaction on the conversion path. In theory, the model, with enough data, should have the ability to work out the “best” shares

    What is data-driven attribution?

    Data-driven attribution is a model that evaluates and analyzes customer behavior to identify patterns of people who convert versus those who don’t. Simply put, it’s a statistical model that analyzes data and models probabilities.

    Google offers data-driven attribution in both Google Ads and Google Analytics. These tools will compare the paths of customers who convert and those who don’t, to find patterns in their interactions. Instead of having to analyze the data yourself, data-driven attribution makes it faster, more efficient and more convenient to present findings from the modeling process.

    How-data-driven attribution works in Google Ads?

    Google Ads analyzes all the data in an ad account to determine which keywords, ads and campaigns have the greatest impact on conversions made. They use a data-driven attribution model:

    • learn which keywords, ads, ad groups and campaigns generate the most conversions
    • you will optimize rates based on performance data
    • in the reports you will see which ads have the greatest impact on your marketing goals

    If you’re using an auto-scoring strategy to increase conversions, the system will use statistical data and your chosen attribution model.

    Minimum requirements for using DDA in Google Ads are:

    • 3,000 interactions with ads on supported networks in the last 30 days
    • 300 conversions in the last 30 days

    To continue to use this model, you must reach the following minimum conversion threshold in the last 30 days:

    • 2,000 interactions with ads
    • 200 conversions

    Time to clean up your attribution model?

    Describe your needs, we will prepare an analytical audit and show you the way forward taking into account your attribution model, cookie requirements and automation.

    Let’s talk!

    How does data-driven attribution in Google Analytics 4 work?

    Since January 7, Google has released an update to Google Analytics 4: now the multi-channel data-driven attribution (DDA) model is available to all users. 

    What has changed in Google Analytics 4 compared to Universal Analytics? Universal Analytics used single-channel attribution models by default. These models included last indirect click, first interaction, last interaction, linear, time distribution and position-based. Almost everyone was using standard reports that actually presented data in the last-lick model.

    Unlike in previous versions, in Google Analytics 4 it is possible to change the attribution model for all reports:

    • Rule-based attribution including multiple channels (last click, first click, linear, position-based, time distribution)
    • Last-click attribution (from ads)
    • Data-driven attribution

    The GA4 help page additionally states that each data-driven model is specific to each advertiser and each (single!) conversion event, and explains how data-driven attribution works:

    “Data-driven attribution allocates conversion share based on data about each conversion event. It differs from other models in that it uses data from your account to calculate the actual share of each post-click interaction.” According to Krista Seiden in Google Analytics 4, the number of contact points used in the modeling reaches up to 50+, ensuring that none of your marketing activities are missed when calculating and assigning share.

    Worth noting is the fact that conversions can occur days after interacting with ads by making the expiration period settings very important! There are two options available in Google Analytics 4:

    • Conversion events related to acquisition (first_open and first_visit). The default validity period is 30 days. If necessary, it can be changed to 7 days.
    • All other conversion events. The default validity period is 90 days. It can be changed to 30 days or 60 days.

    Changes in the validity period apply to all reports in your Google Analytics 4 service.

    Google Analytics offers various attribution types, such as multi-channel attribution, multi-device attribution, online-offline attribution and hybrid attribution. Each type provides unique insights into user interactions and conversions, enabling comprehensive insights for customer analysis and profiling, especially in performance marketing

    To use data-driven attribution in Google Analytics 4, you need to have enough conversion and touchpoint data in your account. In particular, you need:

    • Make sure your data is properly tracked. This means you need to have tracking codes, tags and pixels implemented correctly on all pages or screens of the mobile app.
    • Have a sufficient number of conversion events in your account from which Google Analytics 4 can learn. Google recommends at least 500-1,000 conversions per month across all conversion events.
    • Have enough historical data in your account for Google Analytics 4 to analyze and learn from. For best results, Google recommends using historical data from at least 28 days

    What future awaits attribution models?

    For now, you can use Google Ads attribution models based on data or last click, or choose GA4 and use additional capabilities. In addition, the option remains to use the export of clean data from Google Analytics 4 and develop your own attribution model.

    In some digital marketing campaigns, the last-click model can be beneficial. These are primarily short, simple campaigns that you run in a single channel. This will keep your analysis simple and transparent. On the other hand, you will still be able to relate exactly which keyword and ad attracted a particular customer. In other cases, data-driven attribution will be better.

    What will we get in the future? Certainly, there s no going back to the old attribution models based mostly on guesswork, i.e. asking myself how in my business can new customers come? How many points of contact do I have and what weight should I assign to each channel? In many conversations with customers, we have heard: what is the correct attribution model for my business? Well, there is no correct one, there are different ones. So you can verify different approaches, see how the weight of the channel changes on different paths, what characterizes a particular campaign or source. In summary, attribution is a model that helps you understand the customer path, but never definitively defines it.

    Let's talk!

    Tomasz Starzyński
    Tomasz Starzyński

    CEO and managing partner at Up&More. He is responsible for the development of the agency and coordinates the work of the SEM/SEO and paid social departments. He oversees the introduction of new products and advertising tools in the company and the automation of processes.