Always Be Testing

[Flex]

Figured I would share some inspiration here. I am continually testing my online assets, and I believe others should be too. If you are not, at least be collecting historical data for future comparisons sake.

You should actively be doing the following:

• Surveying your visitors
• A/B/n test (I like to include NPS scores on control & variation to compare as well)
• Heatmapping, scroll mapping, visitor recording
• Behavior event tracking

Tools I use for these things are:

• Hotjar - free in the lowest plan
• Optimizely - free in the lowest plan (if not resort to testing with GTM)
• Google Tag Manager - free
• Google Analytics - free

If you have elements that are interactive in any way, you should be triggering behavior events on them for Google Analytics. I personally am also a big fan of Google Tag Manager and the Javascript data layer. It's powerful if used to the full extent. If you also write these as well as pageviews to a database you can develop personalization and also user type development. Lets say you know that someone who clicks the same element 5+ times within a 2 second span is a rage clicker, you can serve them an automated chat bot. Things like this are powerful to understand.

I like to use Hotjar for heatmapping, movement mapping, scroll mapping, visitor recordings, funnel analytics (by page grouping), and their poll tool. Hotjar is extremely powerful, and their Javascript tagging & triggers are EVEN more powerful, especially when powered by something like Google Tag Manager. This allows you to set qualifications on server side, then trigger a GTM event, and based on that execute a trigger or tag. This allows you to see full on visualizations & recordings of your control/A/B/n tests and compare. It is absolutely huge.

I use Optimizely for AB testing, however this is pretty easy to do with Google Tag Manager as well if you are familiar with Javascript. This coupled with Hotjars NPS polling across variations is very handy.

I use Google Analytics for segmentation analysis and benchmarking. I also extract the Hotjar user ID and pass that into Google Analytics via Google Tag Manager. This is useful for analyzing a certain segments screen recordings down the road.

Here is the result of a month long testing and observation process:

• Decreased bounce rate (very slightly - Good thing)​
• Decreased pages per session (Good thing - I'll explain why below)​
• Increase Avg Session Duration (Good thing)​
• Increase Affiliate link CTR (Good thing)​

Note: Decrease in pages per session is due to less people interacting with site search. This is good because
they are finding the information they want on the post they land on more often.​

This test was removing individual product review links from my list posts, and instead integrating them directly into my list posts. That is why you see the avg session duration increase and pages per session drop. My hunch is also that my affiliate link CTR increase due to a few things:

1. Bringing my affiliate links into a higher % of scroll visibility
2. Providing people what they want without having to go anywhere, and still giving them the info they need to take the next step in the purchasing cycle.

Happy to help and answer questions about my processes. I am not a guru, but I absolutely believe in data collection, personalization, and the ability to optimize based off intent.

 

[Nat]

Hey @Flex I almost started a thread and then I remembered this one. I'm not testing at the level you are, but I have enough Analytics data on enough articles with enough traffic to start really benefitting from it. Until this point, Analytics have mostly been fun

I think I'm doing this right, but I'm trying to make sure since I've never really interpreted this stuff.

I've got a handful of pages that are getting organic search traffic for lists/review type articles. They're intent is very specific, they want to read a comparison of some things, or they want to know the best, etc. Since its organic, hopefully they then pull out the CC and buy. (I haven't been leaking or posting to social media, so I don't even have to filter the referrals for the past week or so.)

My analytics data is showing my decent ranking "review-ish" pages with this data range:

Avg time on page: 7:59-11:25
Bounce rate: 82-92%
% Exit: 81-92%​

Am I correct in saying that
1) If the Avg time on page was not so high, the bounce rate would be a very bad thing.
2) Because time on page is so high, it means the visitors can exit the site without needing to do any more digging on my site (because they got everything they needed).
3) Because almost all of these visitors are falling into product curiosity within my extremely broad vertical, but not my niche, I shouldn't focus much time worrying about the high exit rate & trying to force out an email submit.

 

[Flex]

@Nat - here are my opinion based responses:

For #1: No you are not correct, because you have to understand how time on page is calculated within Google Analytics. It takes the timestamp of the pageview and then the timestamp of the next pageview (or interaction event) and takes the difference to get the time on page. If you have no interaction events then a bounce receives a time on page of zero. This means that your bounced sessions to that particular page are not being factored into the avg time on page calculation with a default Google Analytics setup. Meaning your bounce rate is still an issue assuming a default GA setup.

For #2: Would need to see survey results to understand if people are finding everything they need. They could be spending so much time because they're not finding what they need and having to sift through quite a bit before realizing that.

For #3: I don't quite understand this one. You mean because your site isn't directly related to how your organic visits are arriving via query, that you shouldn't worry about a high exit rate?

As another note to this thread - I am about half way through a test on one of my websites comparison tables. The variation has images removed and is instead using HTML to highlight the various options (best, cheapest, etc). The variation is currently performing better on click ratio, it also has the side effect of reducing requests on the page since I've removed those images, which equals to quicker load times. This test has concluded yet, but food for thought in your own testing world.

 

[Vince]

This test was removing individual product review links from my list posts, and instead integrating them directly into my list posts.

I didn't follow your test exactly. Could you help expand on this?

Great post and I'll have to check out Tag Manager in more depth.

 

[Ryuzaki]

Great post and I'll have to check out Tag Manager in more depth.

Just a friendly heads up here:

• Digital Strategy Crash Course - Day 25 - Google Tag Manager

 

[Nat]

@Nat - here are my opinion based responses:

For #1: No you are not correct, because you have to understand how time on page is calculated within Google Analytics. It takes the timestamp of the pageview and then the timestamp of the next pageview (or interaction event) and takes the difference to get the time on page. If you have no interaction events then a bounce receives a time on page of zero. This means that your bounced sessions to that particular page are not being factored into the avg time on page calculation with a default Google Analytics setup. Meaning your bounce rate is still an issue assuming a default GA setup.

For #2: Would need to see survey results to understand if people are finding everything they need. They could be spending so much time because they're not finding what they need and having to sift through quite a bit before realizing that.

For #3: I don't quite understand this one. You mean because your site isn't directly related to how your organic visits are arriving via query, that you shouldn't worry about a high exit rate?

As another note to this thread - I am about half way through a test on one of my websites comparison tables. The variation has images removed and is instead using HTML to highlight the various options (best, cheapest, etc). The variation is currently performing better on click ratio, it also has the side effect of reducing requests on the page since I've removed those images, which equals to quicker load times. This test has concluded yet, but food for thought in your own testing world.

@Flex Thanks for such a detailed response! For some reason, I didn't notice the @ notification/quote and so I'm just now seeing this.

1) You mention "default" set-up, I guess there is a way to customize this. I started Google'ing it, but is there a way you'd recommend? Found one page that recommended adding a timer function to the GA with page visibility.

2) Very interesting. Does one of the tools you listed in the OP have a survey feature? It would be cool to have a floating/flashing survey box appear in the sidebar area when a visitor has been on page > 2 minutes || makes exit attempt.

3) You're right, that's pretty much what I'm saying. For example, its sort of like being in the weight-loss vertical with a niche focus for "women over 50 weight loss" and I do an analysis of the "top 5 fish oil supplements." 90+% of my organic visitors for "top fish oil" aren't actually women over 50. They like/want the review, but aren't interested in the rest of the site.