Conversion Order

In a typical funnel, events must happen one after another, with a distinct timestamp, in order for users to count as converted through the funnel. However, there are a number of situations where events will not have a distinctly different timestamp.

Consider the following example from PetBox. Here, PetBox is trying to measure the conversion success of their product. They are measuring a user journey beginning with the event Subscribe, and ending with the event Purchase Product. However, Download App and Start App could happen so close to one another that they are uploaded with the same timestamp.

To illustrate another common use case, consider that all events sent to Analytics via a data warehouse integration will often contain the same timestamp. Approximate Conversion Order allows these events to be analyzed using the Funnel tool.

Overview

The default Funnel Conversion Order is Sequential, meaning that events have to happen chronologically, with distinct timestamps for each user to count as converted through the Funnel steps. However, as illustrated above, often those two events occur with the same timestamp. In order to account for scenarios like this, toggle to Flexible conversion order. To switch between the two settings, select Conversion Settings in the menu bar, and then toggle between Sequential and Flexible in the Conversion Order line.

Sequential Order

Typically, each user in the funnel must complete each event in a chronological order, one after the other. In the example below, this means that Subscribe must happen before Create Profile, with a distinctly different timestamp. The same logic applies for each event that follows.

Flexible Order

If you select Flexible Conversion Order, then each user in the funnel may complete each event in a chronological order, or up to one day prior. The options for Conversion Order Limit start at one second, and are configurable to minutes, hours, and to one day. This would mean that users can perform subsequent events in the funnel with a timestamp prior to the first event in the funnel, and still be counted in the results. 

In the example below, Subscribe can happen with an earlier timestamp than Create Profile, and this user will count as having converted between the two steps. The same logic applies for each event that follows.

In the case of events with the same timestamps, timestamps are defined as the same if they occur within the same second. This means that if a user performs Subscribe and Create Profile at 00:00:00.99 and 00:00:01.00, then this user will not count as having converted. If a user performs Subscribe and Create Profile at 00:00:00.01 and 00:00:00.99, then this user will count as having converted.

Custom, joined and repeat events are supported in Flexible Conversion Order, but please note that there must be distinctly separate instances of each event for a user to count as converted.

Conversion Logic

Which event is prioritized in the conversion rate – for example in the case of the Average Conversion Time metric?

When the conversion rate is calculated, Analytics' logic prioritizes different events based on their timestamps, prioritizing the shortest amount of time, or absolute value, of their occurrence to event A.

Take a funnel with two events: Download App (Event A) and Start App (Event B). Conversion Order is set to Flexible, with a time limit of one hour.

In this example, the timestamps for events performed by a single user are as follows:

• Start App at 1:15 PM
• Download App at 2:00 PM
• Start App at 2:01 PM

Here, the absolute distance from Download App, or event A in the Funnel, is counted. Therefore, the instance of Start App that occurred one minute after Download App will be counted instead of the instance of Start App that occurred 45 minutes prior.

Some additional logic affects how these events are treated in the results:

• In the case of identical values (say, 30 minutes before and 30 minutes after) for such events, the event which occurred after event A will take precedence.
• In terms of Conversion Time results, any events occurring before event A will be counted as occurring with zero difference, as if they occurred at the same time as event A.