We’ve already written about why a solid database is critical for your business success. But we keep getting questions and follow up about one of the reasons we gave for why you want to have a great database: customer attribution modeling. Attribution modeling is one of the most complicated and challenging areas of web data and analytics, so before we think about how you may want to think about building out a model, let’s make sure we are all on the same page about what it is.
What is attribution modeling?
Attribution modeling takes into account all your acquisition paths and helps you understand where people are coming from, and how, to convert into a customer. At first, this sounds somewhat straightforward. But it probably isn’t. Chances are, customers have numerous touch points with your product. Take, for example, an app that has a website. Customers can sign up via the app or the website. The owners of the app want to track downloads and signups. However, since customers can sign up via the app or website, the owners have to determine if the signup happened on the app or the website. In addition, they may have other questions such as whether those who downloaded the app then signed up, or had already signed up before downloading the app. These are questions that are not straightforward or easy to answer. A quick whip around your analytics software will probably not give you adequate information to answer these questions. Building an attribution model might.
A quick whip around your analytics software will probably not give you adequate information to answer these questions. Building an attribution model might.
Step 1: Figuring out if you even have an attribution modeling problem
If most of your conversions come after one or two visits, and are pretty straightforward – i.e., you post an ad, and you get a conversion from the ad, you probably don’t need to spend time building an attribution model. If you have a long sales cycle (i.e., many visits to your site), engage in multi-channel marketing or have a site that isn’t particularly e-commerce optimised, you will probably want to spend the time building out a model.
Step 2: What do you want to find out from your attribution model?
This is often the hardest thing to address: what you want to discover with your attribution model. It might be helpful to review strategic and ROI questions you have, like in the scenario above. It could also be helpful to analyze one marketing channel at a time and determine if one underperforms the rest by far, or if one is the main source of sales. Common attribution models include:
- online-offline models that endeavour to determine which digital marketing channels are impacting offline marketing, and vice versa.
- multi-channel attribution models that try to separate out digital marketing channels and their impact on conversions, even if they can’t be easily extricated and therefore measured.
- multi-device attribution, when multiple devices such as desktop, tablet and mobile have an impact on your conversions.
Step 3: Audit your data collection capabilities
As we’re written before, your marketing is only as good as the data you have. Make sure you have all the data collection software you need, as well as the data points necessary to get the information.
Step 4: Understand customers’ purchase journey
You need to know quite a bit about your customers’ decision-making process, objections and buying triggers to help you interpret the data you get out of your model, and eventually to make choices from that data. The best way to understand the journey is to interview clients, sales and support staff, do market research and understand the customer decision-making process.
Step 5: Develop and test a thesis
The best way to then dig in and get started is to make a thesis that you can test. For instance, going back to the above example, we could hypothesize that a most people download and sign up on the app rather than the website. This could have an effect on where and how to advertise sign up; for instance, advertising on other mobile apps more heavily than on search or display. A good attribution model would then compare sign ups from the website and the app.
Notice that the one hypothesis is very clean and simple; you can then make things more complex by comparing various attribution models to make decisions, such as, in this case, assigning each user a tag that is then used to track whether they open the app on web or app after signing up.
Attribution modeling is both a science and an art. If you have a complex buying cycle and questions you want answered, with the help of the tips above, you can start finding solutions with attribution modeling.