Understanding the user knowledge path in an e-commerce business
Perhaps the most critical analytical task in an e-commerce business is understanding the user journey. How do users arrive at your product? How do they behave once they are there? Do they buy something or do they disappear before the purchase stage? If they disappear, where do they do so? And why do they leave? Every step of the analytical process throws up new questions that must be answered, and effectively answering these questions means being able to tease insights out from your data and then communicating them clearly to your colleagues in product, marketing, and beyond.
Rebuilding The User Journey
Reconstructing the user journey means, above all, reconstructing users; who are these people that are visiting your site and/or app? A typical e-commerce business will have three stages of knowledge about a customer: total anonymity upon the initial arrival, contact information (email address at least, perhaps name as well) following a registration, and then full details (name and address for sure, possibly other pieces of information) once they make a purchase. At each step of this process, users are also performing actions that can be mapped against their emerging profiles as tracked events.
From an analytical standpoint, the challenge is to connect all of the actions that an individual user has performed as they move along the path of knowledge from total anonymity to a completed purchase. This is crucial to do, because if you don’t, you cannot properly attribute individual purchases to different digital marketing campaigns, for example. This therefore requires you to aggregate all of a user’s visits into a coherent user profile, through using a mix of tools to connect user visits across various devices, platforms, browsers, and statuses (or at least to the greatest degree of accuracy possible).
Multi-device tracking is a complicated subject that I don’t really want to go into in a big way here, but my opinion is that it makes sense to use a mixture of deterministic and probabilistic tools to help you in your efforts to identify users across multiple devices. Here’s a good summary of the topic for you to dive into.
Understanding the User Journey
There are four categories of data that are useful for understanding the conversion cycle in an e-commerce business:
- User acquisition attribution: Organic and paid sources
- User interactions/behavior: How users interact with the site/apps
- Purchases: What do users actually buy (and then what do they keep and what do they return)
- User profiles: Age, gender, location, technology, interests and beyond
What is crucial here is to understand how these different categories of data interact in order to effectively segment users and better understand user behavior at each step of the sales funnel. For example, it could be that the conversion rate from the landing page collapses from one week to the next — if you are looking at just the user behavior data then you could assume it was a problem with the page when the real culprit might be a paid marketing campaign that has dumped a large number of very low-quality users on the service, none of whom are interested in making a purchase.
Let’s take a quick look at each of these data categories:
Attribution: Which marketing channels led to how many sales? In order to know this, all visits need to be properly attributed to the relevant traffic source, medium, and campaign (if relevant); this can be done by parsing the referrer information. In conjunction with the user ID information, this data can be used to build up a map of the traffic sources that users take on their journey from new users to actual paying customers.
User interactions: High-quality event tracking is essential to conversion optimization. Therefore I recommend capturing all user interactions, and making sure that these events are tracked with a rich set of variables to enable deeper analysis (in a data warehouse environment, these would be additional columns, whereas in Google Analytics these would be additional custom dimensions and metrics). Also, the information captured must be consistent across all platforms in terms of trigger points and variable values, and tracking should be done in the primary language of the business (which would tend to be English. It is especially important that all events connected to the sales funnel are captured, including interactions on product pages through to adding and removing items from the shopping cart, and passing through the five steps of the payment process.
Purchases: OK, to be fair, this is a sub category of user interactions, but in the context of an e-commerce business, purchases are important enough to warrant a category of their own. Why? Well, simply put, customer purchases keep the lights on and keep you all in your jobs! With your purchase event tracking you will be trying to capture details like item name, item ID, quantity, price, revenue, item category — information that can help you calculate average order value, average purchase value per item, profit per order, and so on.
User profiles: These are built iteratively over time. Initially, when users are anonymous, you will generally just have a cookie or device ID to work with (or maybe a browser fingerprint if your tracking is more advanced), but as time goes on you begin to add more information to the user profile, such as their contact details (email address being by far the most important piece of information), and then concrete details like name, address, postcode, etc once they make a purchase. The value of information like this is that you can use this to begin doing some deeper analysis of your paying customers, i.e. segmenting customers by city to see where your strongest markets are.
Putting it all together
With a combination of user, attribution and event data you can build a better understanding of the user journey, on both an individual and on an aggregated basis. This information can help you to begin to answer business-critical questions such as:
- What marketing campaigns have the highest conversion rates and/or the highest customer lifetime value?
- Which user activities correlate most closely with eventual sales? And can you use this to predict early in the user journey if a user will buy?
- Which product categories are associated with the highest conversion rates and/or customer lifetime value? Should the service be reorganized to promote them to a greater degree?
- How often do users buy recommended items? Can this algorithm be tweaked for better results?
- Where in the user journey are users most likely to leave without returning?
- What is the effectiveness of using email to remind users to buy items in their cart?
- Are there any demographic factors that can explain differential conversion rates?
- What is the difference between conversion rates on mobile and web, and does the gap represent a product problem (i.e. that one platform is less effective) or a user behavior one (that users brows on one platform and buy on another)?