Segmentation is the means by which we categorize users by a set of common characteristics in order to understand how their unique behavior can inform future product and marketing strategies. So how does segmentation work concretely? The answer: it depends on the types of questions you think are worth asking. In this article, we’ll teach you practical ways to segment users into groups using three elements: events, event properties and time.

We’ve covered a lot of content related to data collection and the importance of our KPIs. So far, we should have a solid point-of-view of how our product performance can be viewed on aggregate.

When we look at acquisition, activation, retention, referral or revenue, these metrics represent the performance of our entire user base. This can certainly be useful for understanding how well our business is performing on the whole, but from an operational and strategic point-of-view, this perspective is imprecise.

Enter segmentation. We’ve touched briefly on the idea of segmentation when we covered cohorts as a useful means of measuring long-term engagement or performance specifically related to retention and referral. While cohorts are most useful for these two KPIs, they can also be applied to the revenue KPI or any other metric where a long-term analysis could be useful or when we expect lower engagement or investment over time.

Although cohorts are a good example of how to segment your users by engagement, segmentation in product analytics is based on 3 main axes: events, event properties and time.

Segmentation by events

When we segment, we start by looking for people that did things. Because we’re focusing on products used by individuals, all things are done by users. As we’ve already seen, the ‘things’ we’re referring to are events. So, when we look at people that did things, we’re focusing on a subset of users that generated a specific event or a set of specific events.

As an example, we might want to know which users logged in, or which users submitted forms. By excluding all users that didn’t log in or didn’t submit forms, we’ve already executed a preliminary segmentation.

When segmenting by events, we can use a single event or multiple events. When referencing multiple events, we might want to analyze users that generated a first event and a second event, or we might want to analyze users that generated a first event or a second event. The ‘ands’ and ‘ors’ used here are significant because they can limit or expand the number of users we’re considering in our analysis.

As soon as you begin formulating questions about product performance, your first reflex should be to reference the event or set of events that are to be considered in your analysis.

Segmentation by event properties

Once you have a clear idea of the events you’ll be using in your segmentation, you can begin to narrow down your analysis by adding more precision. We’re now looking at people that did this or these things with these characteristics.

Precision in segmentation is wholly defined by referencing the properties of events. Remember that the main reason we collect data through event properties is to categorize interactions by specificity.

For example, if our website allows the submission of several forms, our data collection strategy should allow for an event property in each submitted_form event that specifies the type of form being submitted. In this case, we could narrow our analysis to users that generated the submitted_form event where the form_type property is equal to request for a quote. In this case, we would exclude any user that submitted a form for another reason where the form_type might include contact request or support ticket.

For any given event in an analysis, we can look at the event itself or the event and any of its properties in order to narrow down the scope of our analysis.

Segmentation by time

Time is one of the most important factors in any analysis. Without a representation of time, we’d be looking at the entire history of our users’ interactions. By adding in time, we’re considering people that did this or these things within this period of time (e.g. between a start date and an end date).

We might also want to analyze users’ interactions in a specific order or in no order at all. When we do analyze a set of events in a fixed order, we’re in effect looking at a funnel. By asking for users that did this and then this and then this, we’re laying out an ordered user path. We can compare the steps to each other to understand the drop-offs or we can look at the funnel in its totality in order to discover common characteristics of users that completed the funnel.

Every analysis whether it considers an order of events or a lack of event order requires a start time and end time. This not only lets us compare periods of usage between each other, but it also adds significance to the scope of users’ interactions. By grouping events together by time, we can identify modifications to our product that influenced user behavior in different periods. We can also group together events this way to categorize users by the frequency of their interactions with our product as is the case with a cohort analysis.

You’re now equipped to take on product analytics the right way. By mastering the concepts in this guide, you’ll be able to build a full understanding of your users and measure the effectiveness of your product strategies, marketing efforts and design.

Want to learn more about how to implement product analytics for your company? Download our paper, “Advanced Analytics for Products.”

Download