The one aspect about history is that technology will always change, sometimes in ways unexpected. That tenet holds true for attribution analysis in marketing.
Marketers have historically attributed the last customer online action as the contributor to a conversion. But customer experience influences, along with the ability to measure those influences, has expanded well beyond a computer screen at home. Analytics solution and digital ad providers are looking for ways to let marketers make their adjustments to the expansion and gain a better sense of customer engagement and a superior campaign ROI.
In July, for example, Google eliminated the last click attribution as a default setting from Google Ads. Prior to that, Google removed the last click attribution default in Google Analytics reports. The new reporting setting represents an acknowledgement that last click is no longer an accurate choice for understanding the customer journey. The alterations appear subtle from an analyst's perspective, but the changes represent Google's appreciation of the significant evolution customer journey measurement has undergone.
Looking at the Basics of Attribution
Let's cover the basics on attribution. Attribution is the assigning of a conversion to different interactions a customer has with digital marketing media. The interactions are meant to encourage the customer to completing a conversion, which can be an online purchase, downloading an app or registering for an event. An attribution model is developed from these interactions — the model can be a set of rules to support the aforementioned assumptions or an algorithm that determines conversion credit based on data related to the touchpoints on conversion paths.
There are several attribution models. Marketers turn to one of the following in making analytic assumptions to describe campaign performance.
Last Interaction attribution assigns 100% of the credit for the conversion to the channel from which the last interaction before the conversion occurred.
The credit is assigned this way. Let's say marketing for a new product used campaigns with digital ads, paid Facebook posts, Twitter ads, YouTube videos and email as channels, each with content tagged in analytics. A review of the analytic reports shows that email received the most conversions — email seemed important, so email gets the conversion credit. But it's possible that email was a facilitator for some conversions, not the entire cause of it. Assigning email with all the conversion credit completely ignores touchpoints in the other channels that were as likely an influence on the recipient's decision to click on the conversion link. The ads or YouTube videos could have been seen as a reminder of the offer.
The other side of the last interaction coin is first interaction attribution. First interaction attribution assigns 100% of the conversion credit to the channel in which the first interaction before conversion occurred. Both last and first interactions limit the measurement dynamics of customer behavior to surface-level media.
More channels create the need for more model choices that can get beyond the surface-level media. One of those models is linear attribution. Linear attribution assigns each touchpoint equal conversion credit. In our simple example’s context, the ads, YouTube videos and email channels would receive 33% of the conversion credit.
Another is time decay attribution. Time decay attribution credits touchpoints according to timing. Time decay attribution is based on exponential decay, a consistent reduction rate of value over a period of time, to favor touchpoints that occurred closest to the time of the sale. Touchpoints further away from the time of the conversion receive less credit. In our example, ads that trigger a conversion immediately after would be credited as influential touchpoints.
All analytic tools incorporate variations of these attribution models in their reporting features. Google Analytics, for example, has a beta feature that applies these models in its reports. Thus, users can select the type of attribution model that best matches their measurement needs. GA4 distributes attribution credit in two formats: rules-based models and a data-driven model. The rules-based model is a selection of the aforementioned models plus a position model that combines first click/last click, and linear in a 40/40/20 split. The data-driven model relies on an algorithmic review of the data to determine the attribution credit.
An interesting note in the GA4 developer pages is that direct visits are excluded from all its attribution models "unless the path to conversion consists entirely of direct visit(s)." That is notable since many analysts divide last click attribution as with or without direct visits. Direct visits are visits to your website domain direct — the customer either bookmarked the URL or entered the domain in their browser.
Related Article: The Evolution of Web Analytics
Attribution’s Online History
The general emphasis on last click comes from the digital era where people conducted online purchases from their desktop at home. When someone clicked on a website, marketers assumed that the metrics reflected customers at home. So, their web page views and sessions represented their intention to purchase. This viewpoint colored how marketers view a purchase button click and associated marketing media.
The current marketing era is clearly more expansive, with internet access more widespread. Smartphones and tablet devices rapidly evolved from roles as a secondary screen for engagement into the go-to means for connecting online, changing when and where customers connect to a digital presence. New access capabilities to go online, such as wireless internet in vehicles, created new functional opportunities for apps, such as ordering on the go. Over the years social media has broadened its media options to include social ads and live video among consumer influences of social-enabled commerce. This is not to mention email, apps and chatbots among digital media choices.
As a result, marketing research has revealed valuable customer activity within channels stir customer engagement before the click. The latest study by Twitter and Publicis, for example, identified brand conversation on social media as the new review of a product or service from the brand. Conversation has become touchpoints away from a conversion click.
Marketers must now craft messages realizing that the locations of message delivery can potentially blend into the customer experience. If a company is running ads on a plethora of channels such as CTV ads, podcast sponsorships, Google Display Network, and several social media platforms, then the company has multiple touchpoints that influences a customer’s purchase decision. Those locations can trigger millions of touchpoints to analyze. As a result, attribution analysis has become necessary to gain a true sense of customer intention, and as a result, manage the customer experience through a number of advanced analysis techniques.
So, what can marketers do to better balance attribution, especially as the marketplace shifts in technology? There are a number of approaches, but here are three, each of which marketers can use to start an attribution analysis.
1. Increase audits of potential customer influences on the sales cycle.
Periodically reviewing the sales cycles events against campaign schedules can reveal customer behavior patterns against touchpoints where a company interacts with a customer. Don't look initially for changes in trends — in this case, you are looking at some indications of patterns just to begin developing a framework. The end result is coordinating complimentary display and digital campaign decisions based on when and how customers interact.
2. Refine a Marketing Mix Model analysis to understand touchpoints better.
Marketing Mix Model is where the framework gets developed. You start with a question: With messages across multiple screens more common, which engagement touchpoint influenced sales the most? The answer lies in a Marketing Mix Model analysis.
In the analysis marketers review channel spend, then build a regression model that represents sales or market share. The end benefit is a better view of touchpoints and a better managed marketing budget that is aligned to meaningful attribution. R programming or Python can be used to develop the model.
3. Analyze lift against the time of your content and events touchpoints.
With opportunities to talk to customers, especially with social chat, campaign lift or return on ad spend (ROAS) trends can emerge when compared against the timing of campaign and event communication. Conduct a predictive analysis where possible — advance techniques for time series are an example. Advanced analysis structures a channel comparison to determine if each engagement instance along the customer journey brought the best return on spend.
Related Article: Eliminating Vanity Metrics From the Analytics Portfolio
Final Thoughts on Attribution Analysis and the Customer Journey
Attribution analysis is about making every step in the customer journey count. The marketing community has graduated away from analytics mantras that hailed websites as central strategy tactics to touchpoints that compliment customer experiences. The marketing teams that understand the nuances of attribution will be the ones who will set a historic pace of success for their organizations.