Marketers recognise that accurately attributing revenue to marketing efforts is key to better decision-making, budgeting, and strategy. However, implementing marketing attribution effectively is challenging. In marketing attribution, we assign credit to various marketing channels or touchpoints to determine which contribute most to conversion. It is an important step in marketing, as we want to make sure our spendings on, for instance, an ad campaign are justified. With the number of available marketing attribution models growing, it can be difficult to choose the right one.
This blog provides a brief explanation of two attribution model categories. Once this foundation is laid out, it shares key insights from The Data Story’s research on attribution model performance to help you navigate the options and find the best fit for your needs.
Marketing attribution models can be divided into two main categories: heuristic and data-driven. Heuristic models, the more rigid ones, like first-touch and linear attribution, are widely used in practice for their simplicity. Heuristic means we use a practical method of reaching a goal, even though it might be suboptimal. We accept this suboptimality because finding the optimal solution might be too costly, or even impossible. It is as if we follow a predefined hiking trail. The path is marked and familiar, so we know we will reach our destination. However, it might not be the shortest or most scenic route. In the context of attribution models, it means applying predefined rules to assign credit to touchpoints regardless of data structure. In the first-touch attribution model, this translates to assigning all credit to a customer’s first interaction.
Data-driven models adapt to specific situations, or data. They use algorithms to dynamically assign credit based on data patterns, making them more accurate but also more complex. As a result, data-driven models often outperform heuristic models in accurately assigning credit to touchpoints. To stay on the path of our hiking metaphor, you can imagine we adapt the hike using satellite data to find the best path. Based on traffic or weather data, we can reroute with greater precision. This does, however, require advanced tools and understanding, as is the case with attribution models.
Data-driven marketing attribution models range in flexibility, which refers to how well a model adjusts to unique patterns in your data. For example, a linear regression (or ‘line of best fit’ as you might remember from school) is considered an inflexible model as it can only fit to the data in a straight line. Spline regressions, however, are considered flexible models as they are not limited by linearity. Instead, they can take many shapes and forms in finding the best fit to the data.
Our research on marketing attribution models focuses on the performance of Markov Chain attribution models. These data-driven models share a common underlying mechanism but vary in flexibility. We evaluated their performance across hundreds of simulated marketing datasets, systematically varying four key factors that shape marketing data: average customer journey length, conversion rate, number of marketing channels, and total journeys in the dataset.
In general, more flexible models are better suited for complex data. Such as when customer journeys are longer or when the number of marketing channels is high. However, there are situations where the most flexible models were outperformed, such as when the conversion rate is lower than 0.03%.
These results highlight the significant impact of data structure on marketing attribution model performance. It goes to show how important data understanding is in finding the right model. Relying on an arbitrary model can lead to inaccurate attribution and suboptimal allocation of marketing resources. To avoid this, we recommend selecting an attribution model that aligns with the specific characteristics and complexity of your data. Ensuring more accurate attribution and better resource optimization. Curious about what attribution models align with your data? Feel free to reach out1
Feel like reading more data stories? Then take a look at our blog page. Always want to stay up to date? Be sure to follow us on LinkedIn!
English 
Nederlands