Here’s how to implement an attribution model

In order to make your products and services known, you spend a great deal of money creating marketing campaigns that must be effective and relevant. But how do you actually accurately measure the ROI of your actions? How do you know if your investment in Google Adwords is more effective than that on Facebook, for example? Are you sure that your marketing email campaigns or affiliate marketing campaigns have been the only conversion triggers? Faced with this complex problem, attribution models attempt to unlock the mysteries of the customer path, with a view to identifying and optimizing the progression that drives customers towards a purchase decision.

Implementing an attribution model: Why?

Attribution is a strategic method that measures the respective effectiveness of your various channels and your points of contact within the conversion paths taken by your customers. Yet, within a hyper-fragmented environment, it is never easy to know which steps have ultimately guided the customer decision. This can be very gradual or very sudden, depending on the needs and expectations, whether expressed or not.

Attribution models allow you to optimize your marketing strategy and better allocate your expenses by channel, based on their effectiveness. The goal is not to be everywhere, at all times, but rather to invest where it is really relevant, based on your products, your offers, and the customer profiles. To this end, an attribution model allows you to rate your channels based on the impact that they have had on the conversion. In plain language: what is it that made a customer decide to purchase a product? This is a decision that is not linear and one that depends on a wide variety of factors. And this is where everything gets complicated, as accurately retracing a decision is always to a certain extent random.

The different attribution models

There exist several different attribution models. Here we will present to you the most well known and the most widely used. Spoiler: none of them is actually 100% effective.

The last-click model

This is the most traditional model and one of the most widely used, especially in affiliate marketing. This model involves attributing the conversion to the last recorded click. It is a very basic model, as a conversion is always preceded by several interactions with a brand. To rely solely on the last click will never give you a global view of the customer path.

The first-click model

This model attributes full credit for the conversion to the first point of contact. It is simple, yet highly flawed, as it is only the start of a long path that may include numerous other areas of influence. The fact that a visitor clicks on a Facebook ad does not make this the sole trigger for the sale.

Linear attribution

Since each conversion utilizes several channels, linear attribution smooths out all of these channels in identical fashion. The conversion is thus distributed in equal parts between the various channels that were involved in the conversion process. This model has in its favor the fact that it integrates the multi-channel factor into the analysis, however, its distribution remains arbitrary. For linear attribution, each channel has the same value.

Attribution based on positioning within the path

This model takes into account all of the points of contact, however, it attributes greater value to those that are close to the conversion. This is a useful model in the sense that it considers the last steps to be the most important, to the detriment of the discovery phase. This is an assumption that holds a certain logic, even if it attributes practically zero value to the initial steps.

Parabolic attribution

This is a model that emphasizes the importance of the first and last channel on the conversion path. The first, because this is the discovery of the brand, and the last because this is the one that occurs immediately prior to the conversion and that has managed to influence the purchase act. All of the intermediate channels are smoothed out. This model is a hybrid of the first-click, last-click and linear models.

Can contribution analysis respond to attribution bias?

Theoretically, contribution analysis enables the analysis of the performance of all levers across the entire purchase path. It is possible to know which have held the most weight and had the greatest impact. This is a model based on fact, which avoids the majority of classic attribution errors by smoothing out the path. Contribution analysis renders attribution less arbitrary, with improved segmentation of the participants in the conversion chain. There are four main types of participant:

• Initiator: this is the first visit whereby exposure to a brand results in its discovery.
• Passer: an important role as it makes visitors return to the site, without necessarily resulting in a conversion.
• Autonomous: this is the one used when a conversion is only made possible on account of a single lever.
• Striker: he participates during the final sprint, immediately prior to conversion.

The attribution models are as useful as they are complex and assume a very good knowledge of the user experience and the customer path. Making decisions based on an imperfect or arbitrary model can prove dangerous in the long-term. This is why it is important to take time to refine the implementation of your model alongside data and web marketing professionals. Would you like to know more? Check out our intelligent attribution tool, SMARTattribution!