A/B Split Testing and Optimization?

[McLovin]

Hi Guys,

I was wondering if there is a way to test different variations of a landing page (for example the homepage) to test the most 'conversion friendly' variation of a page (for example to reduce the homepage bounce rate)? I would really like to test different variations of landing pages (but also header, and footer, etc.) but I don't know how feasible this would be? I've been looking at Google Analytics Content Experiment, but that requires two versions of a page. That's not a problem with a static HTML page, but is that also possible with dynamic pages?

Any guidance would be much appreciated!

[Design75]

Hi Guys,

I was wondering if there is a way to test different variations of a landing page (for example the homepage) to test the most 'conversion friendly' variation of a page (for example to reduce the homepage bounce rate)? I would really like to test different variations of landing pages (but also header, and footer, etc.) but I don't know how feasible this would be? I've been looking at Google Analytics Content Experiment, but that requires two versions of a page. That's not a problem with a static HTML page, but is that also possible with dynamic pages?

Any guidance would be much appreciated!

You could set up multiple templates, but they do not show at the same time.

[McLovin]

So if I understand correctly, there isn't really a way to 'load' two (or more) templates simultaneously (or two variations of the same template) and use an A/B testing tool (for example Google Analytics Content Experiment) to direct half the visitors to templateVariation1 and the other half to templateVariation2?

[Design75]

So if I understand correctly, there isn't really a way to 'load' two (or more) templates simultaneously (or two variations of the same template) and use an A/B testing tool (for example Google Analytics Content Experiment) to direct half the visitors to templateVariation1 and the other half to templateVariation2?

There is a module called template switch. May be that can be used, and then tell doodle the different uri for the templates
You can get these uri's by switching between the templates

[mc12345678]

There is a module called template switch. May be that can be used, and then tell doodle the different uri for the templates
You can get these uri's by switching between the templates

I was thinking the same template add-on could be used, that a value could be stored in the database as a sort of on/off switch and then anytime a new session was started to pull the template that matched that switch and flip the switch for the next visitor. Thus providing the A/B switch desired. Don't know how that all plays in with the "experiment" software, but at least would give a 50/50 split basically. Though other aspect of that would be addressing what a logged in user would get, would they stay on the then current template or be able to pick a "favorite" template?

Of course the options could be more than just for two templates, could cycle through any number of templates. Just remember the modifications made to work with one plugin may have a negative impact in an alternate template if that plugin is not also incorporated... (Shouldn't be the case, but is a possibility that could drive one batty to debug.)

[McLovin]

I was thinking the same template add-on could be used, that a value could be stored in the database as a sort of on/off switch and then anytime a new session was started to pull the template that matched that switch and flip the switch for the next visitor. Thus providing the A/B switch desired. Don't know how that all plays in with the "experiment" software, but at least would give a 50/50 split basically. Though other aspect of that would be addressing what a logged in user would get, would they stay on the then current template or be able to pick a "favorite" template?

Of course the options could be more than just for two templates, could cycle through any number of templates. Just remember the modifications made to work with one plugin may have a negative impact in an alternate template if that plugin is not also incorporated... (Shouldn't be the case, but is a possibility that could drive one batty to debug.)

Then would it be possible to 'split' the template files (for example copy/paste 'index.php' and rename it 'index_variant1.php') and have both of them function normally? Because if that is possible, then you could simply set up Google Analytics Content Experiment to compare index.php and index_variant.php. GACE would take care of the rest, that way you wouldn't have to mess with the Zen Cart core files or database.

[mc12345678]

Then would it be possible to 'split' the template files (for example copy/paste 'index.php' and rename it 'index_variant1.php') and have both of them function normally? Because if that is possible, then you could simply set up Google Analytics Content Experiment to compare index.php and index_variant.php. GACE would take care of the rest, that way you wouldn't have to mess with the Zen Cart core files or database.

Possible? Yes.
Practical, a good idea, easy to implement, etc... NO

What does GACE need to "test" these two different templates? If it is some sort of uri difference, then the question would be if the uri contained a parameter such as template=1 would that work for them?

In essence that is what the plugin should offer/allow. Designation of the template to be used through a uri get parameter:
Your host dot com/index.php?template=1 as a generic example would be for the one page, where as = 2 for the other. Otherwise, would suggest some sort of htaccess rewrite to append the template parameter based on the input uri which would also rewrite to the zc site appropriate address. It's really a lot about what does GACE require and what are standard setup methods for ZC that could support the expectation. But renaming/cloning/duplicating the index.php file is definitely something requiring a lot of additional work and relatively breaks the overall processing/operation.

[McLovin]

Possible? Yes.
Practical, a good idea, easy to implement, etc... NO

What does GACE need to "test" these two different templates? If it is some sort of uri difference, then the question would be if the uri contained a parameter such as template=1 would that work for them?

In essence that is what the plugin should offer/allow. Designation of the template to be used through a uri get parameter:
Your host dot com/index.php?template=1 as a generic example would be for the one page, where as = 2 for the other. Otherwise, would suggest some sort of htaccess rewrite to append the template parameter based on the input uri which would also rewrite to the zc site appropriate address. It's really a lot about what does GACE require and what are standard setup methods for ZC that could support the expectation. But renaming/cloning/duplicating the index.php file is definitely something requiring a lot of additional work and relatively breaks the overall processing/operation.

I've added an example screenshot of how Google Analytics performs a content experiment. Basically you need 2 (or more) landing pages. One original page and one variant of the original page. GACE will then send half (or whatever percentage you've set) of your visitors to one page (the original) and the rest to the other page (the variant). Depending on your goals (for example minimizing bounce rate) GACE will then analyze which page performed the best (had the lowest bounce rate in this case) and - if you want - compare it to another variant to see which of those is the most successful page.

This is very easy when it comes to static HTML pages, but I'm starting to think that it might not be that easy with ZenCart. Which is really a shame, because you don't really need a fully functioning template, you only need two or more variants of a landing page that you can compare.

[mc12345678]

So looking at the image that was provided, there is a statement that url parameters are supported. As a result, any number of methods to create a landing page could be used: ez-page, defined page, multiple template style home page, category x vs category y, etc....
Whatever uri would be needed to get the guest to the desired location is what GACE is looking for. So again kind of depends on what it is you want to test as to "test" uunderstanding that changing many things at once may not identify a particular aspect that is "better" but a true this or that. A properly formatted test of multiple differences would be able to be compared over a large enough population to identify particular attributes of the differences.

[McLovin]

So looking at the image that was provided, there is a statement that url parameters are supported. As a result, any number of methods to create a landing page could be used: ez-page, defined page, multiple template style home page, category x vs category y, etc....
Whatever uri would be needed to get the guest to the desired location is what GACE is looking for. So again kind of depends on what it is you want to test as to "test" uunderstanding that changing many things at once may not identify a particular aspect that is "better" but a true this or that. A properly formatted test of multiple differences would be able to be compared over a large enough population to identify particular attributes of the differences.

Thank you for your answers. So could I copy the original template files, rename them to their variant and use htaccess to get the guests to the right landing page?