Forums (Obsolete)

Member

Boolean - 11/28/2013 9:35:14 AM

A/B Testing pages in repeaters

Hi

We have a repeater that lists our articles and we would like to A/B test some of those documents.

- Articles
----Article #1
|______Article #1 - B
----Article #2
----Article #3

The problem is that as soon as we make a variant page of "Article #1 - B" it shows up in our repeaters. This seems like a pretty obvious oversight so I'm curious how your supposed to solve this (without custom coding properties into all our doctypes and repeaters).

Thanks!


Member	kentico_sandroj - 11/30/2013 5:06:28 PM RE:A/B Testing pages in repeaters Hello, Have you followed the documentation for configuring the variants; specifically the Exclude from navigation? Which Web parts are showing the variant - what is the repeater type? Also, what is the hotfix version for the instance in question? Best Regards, Sandro


Member	Boolean - 12/2/2013 10:36:19 AM RE:A/B Testing pages in repeaters Hi Yes we have excluded the variant from navigation, but the item is not being used in navigation - it's being used in a repeater. We have a standard repeater that lists all our articles, but there's no way to exclude variant pages from that repeater. I setup a test using the latest hotfix (7.0.64) in the Kentico sample site:

Member

kentico_sandroj - 12/2/2013 2:35:46 PM

RE:A/B Testing pages in repeaters

Hello,

Thank you for clarifying those points. It seems that you would have to filter our the variants using the WHERE condition, e.g. NodeAliasPath NOT LIKE '/home/%'or NodeAlias NOT LIKE '%Variant%'

Another option is to place all variants in a specific part of the content tree where they will be outside of the repeater scope.

Best Regards,
Sandro

Member

Boolean - 12/2/2013 3:53:24 PM

RE:A/B Testing pages in repeaters

kentico_sandroj wrote: Hello,

Thank you for clarifying those points. It seems that you would have to filter our the variants using the WHERE condition, e.g. NodeAliasPath NOT LIKE '/home/%'or NodeAlias NOT LIKE '%Variant%'

So what's the point of even making a variant page? The A/B testing module seems kind of useless without being able to filter them out. We can use the WHERE statement on variants in some places, but what if we have an alias path that happens to have the word Variant in it and isn't part of the A/B test? It seems like a pretty hackish solution to something that's supposed to be a built in feature.

kentico_sandroj wrote:
Another option is to place all variants in a specific part of the content tree where they will be outside of the repeater scope.

That won't work because the repeater won't find the documents in those paths, the templates will not inherit correctly, back links wont work, any macros based off paths would break etc. etc. etc.

Member

kentico_sandroj - 12/2/2013 4:57:34 PM

RE:A/B Testing pages in repeaters

Hello,

The navigation Web parts were designed with the ability to filter out the documents based on settings under Properties > Navigation. If you are using a standard repeater you would have to either use the WHERE condition or a customize the control to also check the same property as the navigation Web parts.

It makes sense to use A/B testing when you are trying to optimize the structure of an entire page, in which case you would have completely separate documents in the content tree. Otherwise if you are optimizing multiple variants of one page, MVT tests may be more appropriate.

That won't work because the repeater won't find the documents in those paths - isn't that the idea?

The templates will not inherit correctly - set them manually under Properties > template. You can configure inheritance levels.

Back links wont work, any macros based off paths would break - This would be an issue with your custom code, not a limitation of the cms.

Configuring A/B or MVT testing requires additional work in order to get accurate results, it is up to you which approach to take.

Best Regards,
Sandro


Member	Boolean - 12/3/2013 1:12:31 PM RE:A/B Testing pages in repeaters I'll give those suggestions a try, thanks Sandro. It sounds like MTV testing might be better suited to what we are trying to achieve. Cheers.