Conversion Rate
Last updated
Last updated
Measure the percentage of users who perform an Event following the evaluation of a feature flag.
The denominator is Flag Impressions, while the numerator is the number of unique users who triggered the specified event.
Suppose there is a feature flag, register-button-copy
that serves three (3) variants for the Register button:
Register (baseline)
Sign Up
Join Now
In your application, you get the flag value, and also submit an event (register-event
) when the user registers a new account.
Since we have both Events and Flag Impressions, FlagSync can now begin aggregating conversion rates. See Correlating Flag Variants with Events.
In our example, let's say the variants are distributed equally to 100 users each, resulting in the following:
Variant | Flag Impressions |
---|---|
Out of these, the following number of register-events
were triggered.
Variant | Impressions | Events | Conversion Rate |
---|---|---|---|
In a real-world experiment, the conversion rate results may look like the following:
When the flag
function is called from the SDK, an impression is sent to FlagSync.
Additionally, when the track
function is called, an event is submitted.
If we've configured a conversion rate metric that ties together a given feature flag and event key, FlagSync is able to correlate conversion rates.
FlagSync registers a conversion event when it detects that a flag impression occurred prior to the submission of a related event, within the lifespan of a running experiment.
In simpler terms, if a user encounters a particular feature flag configuration and subsequently performs the associated action or event, FlagSync records it as a successful conversion.
Let's look back at the flag above, register-button-copy
. Say we want to test which variant converts better.
First, we'll obtain the flag value for the requesting user context.
We may pass the flag value to the UI to render the <Button />
text.
Once a user registers, we'll call track
with our event.
FlagSync is now aware that User ABC was served a particular flag variant, and subsequently provoked a
register-event
. A successful conversion is recorded.
By creating a conversion rate metric that aggregates register-event
events and tying it to the register-button-copy
flag within an Experiment, we can determine which variant yields the highest conversion rate.
Register
100
Sign Up
100
Join Now
100
Register
100
15 clicks
15%
Sign Up
100
25 clicks
25%
Join Now
100
35 clicks
35%