The Guide to Marketing Attribution - How To Optimize For ROI

The Complete Guide to Marketing Attribution - How To Optimize For Efficiency And 10x Return On Your Marketing Budget

Oliver Pan
July 4, 2023
10 minutes read

If you work in the B2C space, you are no stranger to Marketing Attribution. It helps you identify the marketing channels & campaigns that are driving your customers’ decision to purchase. In this post, I aim to provide a crash course on Marketing Attribution and introduce the “Rocket Fuel” Model, which is an AI-driven model that’s extremely effective when it comes to discovering insights that are not as obvious via other Marketing Attribution methods.

If you want to jump right into implementation, feel free to follow along by using my Python Notebook.

Part 1: The Theory

At a high level, Marketing Attribution can be leveraged in the following ways:

Resource Allocation: Marketing attribution helps businesses determine where to allocate their marketing budget and efforts. By understanding which channels and touchpoints are driving the most conversions/orders, companies can invest more in those areas and optimize their marketing strategies accordingly. This minimizes wasteful spending on ineffective channels and optimizes Return On Ad Spend (ROAS) & overall marketing ROI.
Performance Evaluation: Although all paid ads platforms (Google, Meta, TikTok, etc.) have revenue & conversion metrics available, ROAS metrics are often inflated. This is because paid channels will claim a conversion as long as the customer was exposed to an ad a couple days prior to the conversion. It is critical for companies to collect & track behavioral and conversion metrics on their own through a first-party data platform like Google Analytics or Triple Whale and do their own due diligence to understand which paid & unpaid channels are the most critical in their marketing mix.
Customer Journey Mapping: Marketing attribution also provides valuable insights into the customer journey. By tracking and analyzing customer interactions across various touchpoints, businesses can gain a deeper understanding of how customers engage with their brand, which touchpoints are most influential in customers’ decision-making process. This is especially true for branding & educational marketing efforts, which warm customers up to the idea of buying but generally isn’t the stage of the customer journey where they pull out their credit cards.

Despite Marketing Attribution’s many benefits, it’s not easy to get it right, partially due to the many methodologies that are available.

Last-click Attribution: This model gives credit for a conversion to the last touchpoint a customer interacted with before making a purchase. This method completely ignores the influence of other touchpoints in the customer journey.
First-click Attribution: Unlike last-click attribution, this model attributes the conversion to the first touchpoint a customer encountered. It provides insights into what channels get customers through the door but ignores subsequent marketing touchpoints prior to a purchase.
Multi-touch Attribution: This model splits credit between all touchpoints throughout a customer’s conversion journey. Depending on the specific methodology used, conversion credit can be split equally between all marketing touchpoints or weighted more towards the touchpoints that are chronologically closer to a conversion. This approach is lightyears ahead of the methods mentioned above and will directionally guide your marketing team on which channels are working well.
AI-driven Attribution: As you might have guessed from the name, this approach leverages Artificial Intelligence & Machine Learning algorithms to examine the conversion journey of each and every one of your customers and assign a weight to each marketing channel & campaign. Among all the marketing attribution methods available, this is the Crème de la crème when it comes to accuracy. Out of all the AI-driven attribution methods available, the one that I’ve leveraged most frequently to help retail brands and B2C software companies nail marketing attribution is the Markov Chain method – Or as I’d like to call it, the “Rocket Fuel” method.

Get actionable business insights in your inbox

Bite-sized insights, right at your finger tips

Now, if you are anything like the many B2C professionals that I’ve had the pleasure of working with, you hate “Black Box” processes – which essentially means you get an output from a process after feeding it certain inputs, with no visibility on what transformation happened between the two steps.

So, please let me explain what happens during the “Rocket Fuel” marketing attribution process: let’s say Jane sees an organic TikTok video about a Yoga accessory brand, went to the website, signed up for the email newsletter but ended up buying one month later after seeing a retargeting ad on Youtube and finally bought a pair of Yoga pants. The AI behind the “Rocket Fuel” model will look at the purchasing journey of all buyers like Jane, identify patterns, and assign an importance to each marketing channel/campaign based on how much of a role it played in convincing a prospect to make a purchase.

Although the process of the Rocket Fuel Marketing Attribution Method seems quite complex, its output is simple. It just assigns a number of orders/conversions and revenue to each marketing channel/campaign. With this info, your team will then be able to reallocate marketing budget, optimize for ROI, and identify IMPORTANT marketing channels that are driving conversions/revenue in a more subtle way (hint: channels outside of paid ads!). In the Yoga accessory company example above, the Rocket Fuel Method might reveal that TikTok organic videos are responsible for 30% of conversion and revenue, but the management team was just about to cut the content creation budget by 50%. Disaster averted!

At this point of the post, the benefits of Rocket Fuel Marketing Attribution is evident. So let’s talk about the next steps you need to take to take full advantage of this. Assuming your organization has a functioning Customer Data Platform (CDP) that tracks most customer touchpoints, the next step is to leverage Python to run the data analysis. If you have an in-house data scientist, feel free to forward this article to him/her! If you have any questions, you can also reach out to me on LinkedIn.

Part 2: The Data Preparation & Transformation

All right! Without further ado, let’s jump into the data. Your Customer Data Platform should allow you to identify touchpoints along customers’ buying journey and summarize everything into a data table like this:

You will have a user’s unique ID (labeled as ‘cookie’ in the example), timestamp of an interaction, interaction type, whether it’s a conversion/purchase (boolean value), value of that conversion/purchase, and the marketing channel that directly drove that touchpoint.

Before any kind of AI can be done, the data needs to be aggregated at a customer level. This new data table will show each touchpoint along a customer’s journey, how many purchases they end up making, and the total revenue that each customer brings in. (Please note, since the average conversion rate of an eCommerce website is 1-4% (source: Adobe), most users in your database will have 0 for their conversion and converson_value columns. This is to be expected.

The final step before the magic can happen is yet another step of data transformation. But this time, you will need to group the dataset by unique conversion paths and add the total number of orders, total revenue, and the total number of customers that had this journey but didn’t end up placing an order (added as the ‘total_null’ here).

Part 3: Enter The “Rocket Fuel” Model

The “Rocket Fuel” or Markov Chain model produces the most insights when compared to the other more obvious attribution model mentioned in the Theory section. In the example dataset, you will quickly discover that Instagram plays a key role in customers’ purchasing decision while it might not be as obvious when you look at the last marketing touchpoint that drove the purchasing decision. On the contrary, Paid Search’s impact on customers’ purchasing decisions appears to be inflated in other attribution models. If you are the Marketing Director of this company, the insights that we just discovered should justify a repurpose of 5% of the Paid Search budget to creating engaging Instagram content that turn more prospects into customers.

At the risk of stating the obvious, some channels are more prone to bringing in low-value customers. That’s why I ALWAYS recommend looking at attribution of conversions as well as revenue! In this case, they seem to align. But it’s very possible that you find out that certain channels/campaigns are bringing in lots of low-value or one-off customers! In that case, divesting would be the natural next step.

Now, if you want to get REALLY nerdy, there’s actually another use case for the Rocket Fuel Model beyond Marketing Attribution. The exhibit below is a transition diagram, which essentially tells you what the next step in a customer’s journey will be if we know the channel they have been exposed to. Let’s pretend you are the company in the sample dataset – you will quickly realize that customers that are exposed to an online video ad have a 61% chance of seeing another video ad as their next touchpoint. With that in mind, you will want to create a cohesive story line among your online video ads! The same thing can be said about Facebook and Instagram, but this is most likely due to the parent company Meta actively showing your Instagram feed to folks that have seen your Facebook ads, and vice versa. Still, my point remains – a transition diagram can inform which marketing channels complement each other the most and allow you to align the messaging across said channels.

Congratulations! If you’ve made it this far in the article, you now have a high level standing of the Rocket Fuel Marketing Attribution Model and how it can be used to improve the return on your company’ marketing spend. While the example shown with the sample dataset might not 100% align with your company’s situation, the whole process can easily be replicated to enable data-driven growth. If you have questions, feel free to connect with me on LinkedIn or shoot me an email directly. Let’s drive this growth!