In the digital marketing process, customers—especially during the evaluation phase—will access your website or app, and your product or service available on the marketplace.
At this stage, although there may be interactions (events or touchpoints), the conversion resulting from these interactions is quite low. This is a normal situation for a brand that targets customers through multiple channels during the research phase1. However, this activity’s performance during the evaluation phase—when we aim to express customer behaviors numerically—will present certain challenges. At this point, we provide a solution through an attribution model. An attribution model enables the evaluation of customers’ interactions with various sources on a measurable, common basis2.
Attribution Model (Attribution Modeling)
How do you connect your products and services—or developments related to them—with your potential customers or existing customers? How do you inform them, present your offerings: social media shares, organic searches, content published across different sources, videos, advertisements appearing across different media and formats, email newsletters? Some of you may say one or all of these. Indeed, digital marketing activities are typically conducted across multiple channels, and often even across device types. So, how can we evaluate these interactions within a common framework and analyze the movement of users along their conversion paths?
Attribution model (attribution modeling) is a comprehensive framework that determines how much credit or responsibility a particular source (such as search, ad clicks, etc.) deserves for a conversion, based on the interactions (e.g., searches, ad impressions, clicks) a customer undergoes during their conversion journey. This enables proper budget allocation, a deeper understanding of ad performance, and optimization throughout the conversion path.
[quote cite=”~ John Wanamaker”]Half the money I spend on advertising is wasted; the trouble is, I don’t know which half.[/quote]
Most advertisers associate digital marketing success with the interaction that occurs just before conversion (i.e., the last click). For example, in the case of a keyword ad, the entire credit for a conversion is typically assigned to the ad that was clicked last and the corresponding keyword. In such processes, it is common practice to increase the budget for the source that generated the last click, while reducing budgets for other sources3. However, after budget reallocation, it may be observed that the last-click source no longer performs as effectively as before. This is because interactions with other ads along the customer journey are being overlooked. Through attribution models, we determine how much credit or importance to assign to each source for conversions, enabling performance comparisons between sources and facilitating budget optimization4.
Different Attribution Models
In the previous paragraph, I mentioned that most of the credit is typically assigned to the last click. One of our models and the most commonly used one is known as the last-click model.
Last Non-Direct Click Attribution Model
The last click model assigns all credit to the last interaction and the corresponding source. No evaluation is made for any interaction other than the last interaction, including any interaction prior to the last click. As previously mentioned, although this is the most commonly used model, it is also the least efficient. One reason for its widespread use is that it simplifies the measurement process and, as a result, is pre-defined in most analysis tools.
Last Non-Direct Click Attribution Model
In this model, all credit is assigned to the interaction immediately preceding the last interaction. While this may seem more logical than the last click model, if conversions are driven primarily by direct traffic or email promotions and conversions occur through these sources, then these interactions are effectively overlooked within the model. For example, suppose a user visits your website via Google Ads, adds items to their cart, and completes a purchase, but does not complete a conversion. If we send them an email with a coupon and a reminder of the items in their cart, and they then complete a conversion, the model will assign credit to the Google Ads rather than the email.
Google Ads Last Click Attribution Model
This model assigns all credit to the last ad interaction within the engagement flow, disregarding all other interactions (including the last non-ad interaction) that occurred prior to the conversion. For example, suppose a user visited your website through various time intervals via Ads > Social Media > Ads > Newsletter > Direct channels before the conversion, and then completed the conversion directly through direct traffic. In this case, all credit would be assigned to the last ad interaction that occurred before the newsletter.
First Click Attribution Model
It assigns all credit to the first interaction. For instance, if the first click came from a text ad, the entire conversion would be attributed to that ad and its corresponding keyword.
Linear Attribution Model
In the linear (or equal) attribution model, credit for the conversion is distributed equally among all sources the user interacted with prior to the conversion.
Time Decay Attribution Model
Interactions occurring closer to the conversion receive more credit. For clicks occurring more than seven days prior to conversion, half credit is assigned during distribution. That is, an ad interaction occurring eight days before conversion receives only half the credit of an ad interaction occurring one day before conversion. This model can be used during the initial stage, when interactions are low and need to be understood.
Position-Based Model
The first and last interactions receive 40% each of the credit, while the remaining 20% is distributed among the other interactions. In this model, the first interaction is prioritized because it brings the user to the site/webpage, and the last interaction is prioritized because it leads to conversion.
The position-based model can also be described as W-shaped, assigning credit not only to the first and last interactions, but also to the second interaction. 590% of the credit is allocated between the first, second, and last interactions, while the remaining 10% is distributed among the other interaction points5.
Customised/Personalized Attribution Model
It is recommended that before implementing a custom/personalized attribution model, you have previously tested it against a time-decay model and have answered the following fundamental questions:
- What interactions are most important to you? For example, do you believe there is a relationship between page depth, session duration, downloads, and conversion?
- What is your ideal conversion time? How long does your customer take to complete their conversion journey? What duration do you aim for?
- Can you identify a common pattern among conversion journeys that end with repeat purchases?
- What is the most valuable interaction from a financial standpoint?
- Do you send data from different sources or offline data to Google Analytics?
Custom model operations, including model comparisons and adjustments, are performed via the page Google Analytics > Conversions > Multi-Channel Funnels > Model Comparison Tool. You can also perform model comparisons among defined models via your Ads panel by following the steps: Tools > Attribution > Model Comparison.
Data-driven Attribution Model
Conversion credit is distributed based on historical data associated with a specific conversion event. Unlike other models, this model calculates the actual contribution of each interaction along the conversion path. However, this process is only available for accounts that have sufficient data.
- We can identify high-performing channels that reach the conversion goal and help manage resources accordingly. This enables us to evaluate advertising agencies and advertising campaigns as well.
- We can evaluate and measure each channel’s share of conversion on a channel-by-channel basis. For example, within the engagement flow that leads to conversion via ads, we can also see the keyword-level conversion data associated with ads6.
- We can evaluate the return on investment (ROI) for ad spend across channels.
- We can measure the performance of channels that support conversions (assisted conversions) and optimize the customer journey with new channels.
- We can compare online and offline marketing performance.
- We can access information regarding the time lag and path length associated with the conversion journey.
Footnotes
- Attribution (marketing) ↩
- About attribution models, Google Ads Help ↩
- What is Attribution Modeling and How Can it Help you Understand your Marketing Efforts Better? ↩
- Multi-Channel Attribution Modeling: The Good, Bad and Ugly Models ↩
- All You Need To Know About First Click, Last Click and Attribution Modelling ↩
- Become a champion in attribution modeling: a how to guide ↩