How to plan your integrated data layer

Developing your data plan is arguably the most important step in instrumenting effective measurement and analysis of your customer data; any analysis you perform will be entirely dependent on the dataset you collect. There are generally two approaches to data collection, top-down and bottom-up, that can be used in conjunction to flesh out your strategy.

Approaches to data collection

Top-down

The top-down approach focuses on using high-level business goals, like KPIs and use cases, to determine what user data you collect. At this stage, you’ll need to dig deeper into the understanding of your use cases to determine the specific data points needed to properly measure their performance and metrics.

Bottom-up

The bottom-up approach uses granular data to drive your strategy and should be used in parallel to the top-down approach. Bottom-up data management strategy planning requires fully exploring your digital properties and databases to understand all available data about the user and any actions that are possible for the user to perform. This low-level, granular view will allow you to uncover some areas of interest that you may not have considered when thinking about your data from the “top-down” perspective. From there, you can determine which granular data points are valuable for collection and analysis.

Both approaches require that you survey the full suite of digital properties because some types of customer data may only be available from an app or website, while other data may be stored server-side in your company’s databases, for example.

Types of data to collect

There are typically two main categories of data that you can collect on your users through your digital properties: user data and event data. User data refers to data about your customers while event data refers to data on customers’ actions; both of these types of data are important when creating a well-rounded data management strategy.

User data

User data describes identities and attributes of your users that should be maintained and persisted in their user profile. For most companies, the most basic user attributes to collect will include demographic data on the user’s age, gender, and location, but can be expanded to include more specific attributes like membership level, lifetime value to your company, or opt-in status for marketing communications. These attributes allow you to analyze your user base to gain insights and build segments to perform specific workflows.

If your business operates in the EU, you may also need to capture and manage your users’ consent status to adhere to the General Data Protection Regulation (GDPR) standards. Consent parameters for a user function very similarly to user attributes, in that they describe a persistent property of that user, based on their consent response. Robust customer data platforms, like mParticle, can capture and maintain consent status for individual users, allowing companies to action and filter based on a user’s consent status. Consent collection and management is something that requires thorough planning and consideration from your organization with input from legal and privacy teams to ensure compliance.

Event data

Event data describes behaviors performed by your users, typically within your digital properties. These events can describe explicit actions such as navigating through an app, watching a video, performing a search, or completing a purchase. They can also describe passive actions, such as loading a particular screen or completing a call to your servers from the client.

There are many ways to track event data, so it’s helpful to pick a methodology and stick to it through your instrumentation. For example, when tracking a user’s navigation through your app, you could log events for each individual tap/swipe that takes them to a new page, or you could log an event when a new page loads. You may even want to track both, but the important thing is to be consistent.

You’ll also need to decide how granular your event tracking should be. For example, if you have a sign-in page on your app or website, you likely want to track sign-in events as a key data point, but you could increase the granularity of event data to include sign-in attempts and failures prior to a successful sign-in. This type of granular tracking is particularly useful for funnel analysis and A/B testing when you want to understand precise user behaviors.

Taxonomy of events

Another important consideration when creating your data plan is the taxonomy that will be used for your events. There are an infinite number of ways to define the nomenclature of your data, but there are a few best practices that will help you keep your data scalable and usable.

Understand how the data will be used

Create a data model that your team will be able to use and understand. Keeping in mind who will be using the data after it’s captured is critical because their understanding of the data model will directly impact how well they are able to analyze the data. Ideally, every team member that will come into contact with your customer data should understand what each data point represents so that when the time comes to analyze or report on that data, they know how to find the right insights.

Naturally, most users will want to create event and attribute names that make sense in plain language, rather than coded names that will require translation. To improve legibility and organization, we recommend adding prefixes to event names so they can be easily grouped by module.

Keep event names abstract

Maintaining a user-friendly taxonomy means you need to build in scalability. Keep your data model scalable by keeping your event names abstract and adding more detailed information in the custom attributes of the event. As you add new content to your digital properties, the new values associated with that content will expand the set of metadata that you’ve captured, rather than inflate the set of unique event names.

One example of this type of abstraction can be applied to product views on an ecommerce site. Ideally, the event name could be abstracted to something like “View Product”, while the details of the product itself would be captured in custom attributes such as “Product ID,” “Category,” and “Product Description.” This taxonomy ensures that your data is easy to find and read.