Attribution Modeling: “The Channel Analyzer”
Previously I shared an overview of various components involved in Marketing Analytics. In the 2nd week, let’s take a deeper dive into this ocean and see how the magic works!
What is Attribution Modeling
Attribution modeling is a way to understand and analyze the various marketing channels or touchpoints that contribute to a desired action, such as a product sale, sign-up, or any other conversion. It helps determine which marketing efforts or channels impact the customer’s decision-making process most.
It’s like solving a mystery, figuring out which clues (like ads, friends’ suggestions) convince people. It uses data to show which clues are most important in making people buy or sign up. Businesses use this to know where to put their marketing efforts for better results.
I suppose many out there consider starting with Market Mix Modelling before Attribution Models, but I have a different take on that. Based on my opinion, Attribution Modelling provides a detailed and granular analysis by attributing value to each touchpoint within a customer’s journey. On top of that, it often provides near-real-time insights as it tracks customer interactions in current or recent time-frame. This real-time analysis empowers businesses to adapt their marketing strategies swiftly, steering them away from missed opportunities and towards timely adjustments.
Now that we have gained a hold on attribution modeling, let’s explore the family.
Family of Attribution Models
1. Last Click or Last Touch Attribution Models: This is the default option for most digital marketing campaigns. It attributes all credit for conversion to the last interaction before a conversion ignoring previous touchpoints. It focuses solely on the final action that led to the conversion. This model is good for measuring initial impressions.
When to use:
- Just getting started with the attribution
- Campaigns mostly rely on paid ads for new customers
- The product has a short purchase cycle
When NOT to use:
- Spending more than $50,000 per month on ads(as using multiple channels now)
- Established brand with lots of organic traffic
- The product has a long purchase cycle
2. First Click or First Touch Attribution Models: As the name suggests, it credits the first interaction of a customer with the brand for the conversion. Being the second most popular option for digital marketing, this model emphasizes the initial touchpoint that introduces the customer to the brand.
When to use:
- First step up from the last click
- Using top-of-funnel marketing channels eg. Social Media ads
- Product has a longer purchase cycle
When NOT to use:
- Running lots of activation campaigns
- Need a quick optimization feedback loop
- The product has a shorter purchase cycle
3. Time Decay Attribution Models: It assigns more credit to touchpoints closer in time to the conversion event. This model recognizes that interactions closer to the conversion might have more impact than earlier ones.
When to use:
- First step up from Last Click Attribution
- Using multiple marketing channels
- Product has a longer purchase cycle
When NOT to use:
- Spending less than $50,000 per month on ads
- Can’t easily track user behavior over time
- The product has a shorter purchase cycle
4. Linear Attribution Models: It shares credit evenly across all touchpoints throughout the customer journey. This model makes no value judgment about touches and considers all interactions to be equally important in influencing the conversion.
When to use:
- Using lots of smaller channels
- The product has a complex purchase cycle
- Don’t have a strong opinion on relative value
When NOT to use:
- One or two channels are dominant over others
- Have a limited ability to optimize budgets
- Can’t easily track user behavior over time
5. Position-Based or U-shaped Attribution Models: It is the most flexible model giving more weight to the first and last touchpoints, with less credit to the interactions in between. This model also allows manual weighting of value.
When to use:
- In need of a more sophisticated model
- Using multiple marketing channels
- Spending more than $100,000 per month on ads
When NOT to use:
- One or two channels are dominant over others
- Don’t have a strong opinion on relative value
- Spending less than $100,000 per month on ads
6. Data-Driven Attribution Models: It utilizes advanced algorithms or machine learning to assign credit based on data analysis limiting human bias. This model considers various factors and patterns in customer behavior to determine the value of each touchpoint.
When to use:
- Don’t have a strong opinion on relative value
- Have a complex marketing channel mix
- Spending more than $100,000 per month on ads
When NOT to use:
- One or two channels are dominant over others
- No analyst team to test and validate the model
- Spending less than $100,000 per month on ads
Attribution Model Implementation
Now that we have covered the concept, it’s time to get our hands dirty with the implementation. There are several ways we can create an attribution model.
Google Analytics 4 (GA4)
Google Analytics 4 is a cross-platform analytics tool that helps businesses and marketers track site visitors and user behavior metrics. GA4 uses machine learning and AI to provide insights into how users interact with a website and app. Google Analytics caters more to the marketing aspects rather than the technical data science aspects. If your goal is to measure metrics and enhance performance, Google Analytics suits your needs.
To access attribution reports in GA4, select “Advertising” in the left-hand menu. This presents two options. Let’s start by exploring the Model Comparison report.
This report showcases marketing channels on the left, alongside Conversions and Revenue figures on the right. You might wonder why there are dual sets of these metrics. Well, it’s all about comparing various attribution models applied to marketing data, empowering you to derive insights tailored to your business strategy.
Among the six models previously discussed, let’s consider a comparison between the Last-Click Attribution Model and the Time-Decay Attribution Model. Here, you’ll find Conversion and Revenue data for both models. By scrolling to the far right, you’ll discover the percentage increase or decrease when comparing these models. A positive percent denotes the potential undervaluation of a marketing channel when using the last-click model, while a negative percent indicates potential overvaluation. These insights help us understand how different models influence our perception of marketing channels.
Now, let’s move on to the next report: the Conversion Paths report. This report presents a visual map illustrating the various marketing touchpoints that users engage with before converting.
To all our dedicated readers, thank you for embarking on this insightful journey with me into implementing models for enhanced analysis. I trust the shared information has ignited your curiosity and motivated further exploration in this dynamic field!
In the upcoming segment, we’ll explore practical implementation, providing clear steps for seamless integration into your strategies. Take your time to absorb the knowledge gained and return rejuvenated for the next installment.
Be sure to stay tuned for the next blog, where we’ll delve deeper into the practicalities and nuances of model implementation.
Excited to share more insights with you soon!
