Download the PHP package matfish/craft-optimum without Composer

Craft Optimum

IMPORTANT: Version 2.1.0 (Craft 5)/ 1.5.0 (Craft 4) introduces a breaking change. If you upgrade to this version, you will need to call optimumFireEvent explicitly in your twig template.

This plugin allows the user to conduct server-side A/B testing in CraftCMS. As opposed to client-side testing (e.g using Google Optimize), the test variant is rendered on the server-side, resulting in better UX, performance and enhanced flexibility.

Once an experiment is set you can send the data to your tracking platform and start tracking. By default, the event is sent for Google Analytics 4 as a custom dimension. You then have the full power of analytics to compare the test groups over different metrics (e.g conversion, engagement etc.)

If you are using a different analytics platform, you can still use the plugin to conduct the test, by passing a trackingPlatform or a custom fireEvent closure (the latter is useful for platforms that are not supported by Optimum).

Table of Contents

1. Requirements
2. Installation
3. Usage
   • 3.1. Create an experiment in the control panel
   • 3.2. Create the variants in twig
     • Polymorphism
     • Explicit Declaration
     • Variant value only
   • 3.3. Fire the event
   • 3.4. Test your variants
4. GA4 Setup (If applicable)
5. Troubleshooting
6. Caveats
7. Local Development
8. License

Requirements

Craft CMS 4.x or later.

Installation

1. Include the package:

2. Install the plugin:

Usage

1. Create an experiment in the control panel:

Click on the new "Experiments" menu item and then click on "New Experiment". An experiment consists of the following fields:

• Name E.g 'Banner Types'
• Handle - E.g bannerTypes. This is a unique identifier for the experiment.
• Enabled? - Whether the experiment is currently active. This can be used to pause or permanently stop the experiment
• Population Segment - Optional field to limit the experiment to a specific population segment by percentage. Defaults to 100% of users. The weights of the variants are distributed only across the included users. No event will be fired for excluded users (even when the fireEvent method is explicitly called).
• Variants - The different variants for the experiment. E.g if you are testing different hero banners you would set their reference here. An "original" variant is pre-set and refers to the control group, represented by your current code. Its handle cannot be modified, and it cannot be deleted. Each variant comprises three fields:
  • Name: Human readable name. This will be sent to GA4 as the value (E.g "Wide Banner")
  • Handle: Used for naming the variant template in twig (E.g "wideBanner")
  • Weight: Relative weight (probability) in percents (e.g 40). The sum of all variants' weight must add up to 100%
• Starts at: Optional field to defer the experiment. If left empty experiment starts immediately (assuming that "Enabled?" is on).
• Ends at: Set to 30 days in the future by default.

  2. Create the variants in twig

  Method A: Polymorphism

Wrap the part of code you wish to test (your "original" variant) with the {% optimum 'experimentHandle' %} tag like so:

Then create templates corresponding to each variant (except for "original") using the following naming convention: _optimum/{experimentHandle}/{variantHandle}.twig

E.g:

• _optimum/bannerType/wideBanner.twig
• _optimum/bannerType/narrowBanner.twig

Inside each template paste the code for the variation you wish to test. E.g:

NOTE: Don't include the variant templates in your main template code. Optimum will automagically load the correct template at runtime.

Method B: Explicit Variant Declaration

While method A is useful when you want to switch components, sometimes you may wish to switch the location of the component on the page (e.g Test different CTAs positions). With method B, you can declare multiple optimum blocks with the second parameter being the variant:

E.g:

The plugin will only compile the relevant variant.

Method C: Get only variant value

In some cases there is no need to create multiple templates, as the value of the random variant can replace a constant in the code. E.g, Suppose you have a blog and want to test different per-page values. Here is an example implementation for a hypothetical (or is it?) recordsPerPage experiment:

3. Fire the event

Send experiment and variant to your tracking platform:

By default, this will send the event to GA4. e.g:

You can modify the tracking code by specifying a trackingPlatform or by overriding the fireEvent setting:

1. Create a new file in your config folder called optimum.php
2. Add the following code:

Or, if the platform is not supported by Optimum, you can specify a custom function:

NOTE: If you are using a tracking platform that is not currently supported by Optimum, we encourage you to share your custom tracking code implementation. This will help us expand our support for additional platforms in future updates, benefiting the entire community. Please consider submitting your custom tracking code example to the project repository or reaching out to the maintainer.

4. Test your variants

Now that everything is set up, the plugin will randomize a variant and persist it in a cookie, to keep the experience consistent per-user. You can test your variants (and the original) by adding a ?optimum={variant} query parameter to your URL. E.g ?optimum=wideBanner or ?optimum=original. The plugin will disregard the parameter if the value does not correspond to one of the variants.

Appendix: How to Set a Custom Dimension in GA4

This section assumes you are using the default ga4 tracking platform. For other platforms, please refer to the instructions for your specific tracking platform.

1. Open GA for your property and go to Configure->Custom Definitions
2. Click on the Create custom dimensions button
3. In the modal fill in the following details:
   • Dimension Name : Descriptive name. Can be anything you want.
   • Scope : Event
   • Event parameter : Experiment handle (e.g bannerType).

4. Click "Save"

   

Et voila:

All Done! Once GA has collected enough data, you can start comparing the performance of the different cohorts/test groups:

If using a different tracking platform, you will need to set up the tracking accordingly.

Troubleshooting

Before opening an issue please make sure that:

1. Cookies are enabled
2. Caching is disabled on the testable page (e.g Blitz), as plugin decides in real-time which variant to serve.
3. If using GA4 for tracking: GTM is installed on the page (type gtag in the console to verify).

   Caveats

• Code inside the optimum tag is scoped. Variables defined inside the block containing the original variation (or in the variant templates) will not be available externally.

Local Development

When developing locally, if using GA4 tacking code, you are likely not going to have gtag installed, which will result in the following console error:

While there is no issue with ignoring this for development, for the sake of completion, and to see what arguments are being sent to GA, you may want to add a dummy gtag function in your <head> section:

You can use the same method to mock other tracking functions.

License

You can try Optimum in a development environment for as long as you like. Once your site goes live, you are required to purchase a license for the plugin. License is purchasable through the Craft Plugin Store.

For more information, see Craft's Commercial Plugin Licensing.