In the simpler days before the Internet, attribution was easy. Perhaps there was an extension after the phone number given in a television ad or a department number included with the address for mail-in orders that could tie the sale to a specific campaign. Back then, sales tended to progress in a linear fashion. Now, however, the sales journey is anything but linear— and so are your marketing efforts. The problem is that this multi-channel approach makes it difficult for you to attribute your ROI to the correct channel or channels.
There are many different types of attribution modeling. Each has its own set of benefits and disadvantages. Most of them rely on the concept of search sessions, which can provide you with insight into the steps followed along the path to a purchase. Not every model is ideal for every business, but one is surely right for you.
You are planning a vacation to a distant city. You start by searching for hotels in that city, and you receive a list of paid advertisements as well as ranked results. You click on one of the listings and visit the hotel's website. From there, you follow the link to the hotel's Facebook page to read comments from former guests. You realize that you are going to need to rent a car during your stay, so you search for car rental agencies. Selecting one, you go to their website and find a targeted ad for the hotel you are considering. You decide to go ahead and make your reservations, so you click on the ad and book your room.
- The last-click model will give 100 percent of the credit to the click immediately preceding the conversion. With the scenario given, that would be the targeted ad. The last-click model offers ease of set-up and can give a direct comparison of identical campaigns across different channels. However, it ignores the start of the funnel and is more focused on the end, which could lead to overlook channels and efforts that are instrumental in guiding customers to the conversion.
- The first-click model assigns all credit to the first click in the search session, which would be clicking on the link in the search results to visit the website. It is easy to implement and offers you data on how the customer located you. However, it is biased toward the early stages of the purchase journey and fails to provide you with any insights on the other touchpoints.
- The linear attribution model assigns equal credit to every touchpoint in a session. This model is most useful for efforts that are intended to foster awareness and maintain contact throughout the purchase journey. However, it does not identify which touchpoints had the greatest impact, increasing your risk of spending more on certain channels that provide little in the way of returns.
- The position-based model is something of a hybrid between the first- and last-click models. You assign a percentage to every touchpoint, typically with the first and last touchpoints receiving the greater credit. This model lets you see the original touchpoint that garnered the customer's interest as well as the touchpoint that resulted in the conversion. This model may not be ideal if you are plagued with "dirty data" or lack the experience to assign realistic values to the various touchpoints.
- The time-decay model assigns the most credit to the final touchpoint, and as you trace the journey back to the origin, each step decreases in value. Many consider this model to most closely mimic the customer's behavior along his purchase journey. The primary drawback is that you may be assigning too little credit to the effort that had the greatest impact on convincing the customer to make a purchase, such as a blog post or your rankings in the search engine results.
- Lastly, you can choose a custom attribution model. This allows you to build a model around your specific objectives, and you can use other attribution models to compare your results. This gives you the greatest insight into your business as the custom model is developed just for you. Having a custom attribution model is seldom recommended until you have familiarized yourself with using the other models, and you will also need to be certain which channels and variables you need to use.
The choice of an attribution model should be matched to your goals as well as your ability to gather meaningful results from the model. No model is perfect for every enterprise, so you may need to experiment a bit to determine which is most useful to you.