I see advice on a constant basis about how to increase your newsletter subscriptions. Most of the time, they are part of a generic list with claims about doubling or tripling conversion rates, yet I very rarely see them accompanied by actual analytical data confirming the recommendations. So a few months ago, I set out to begin testing these ideas.

The claim

Using a static bar (think “Hello Bar”) at the top of your website with a CTA for your newsletter sign up increases opt-in conversion rates.

I've seen this claim so many times now that it would be inaccurate to cite any single source as being the inspiration behind this test. There are also multiple services that provide different “styles” of static bars. You can choose to put your signup form directly in the box or to link to your newsletter subscription CTA page, etc. You can make it stand out or blend in. The options for a seemingly simple site feature are surprisingly numerous.

The background data for the test site

1. This is an affiliate site I own that does about 250K page views per month (so we are not talking about a site with 500 visits a month).
2. The site already features a “subscribe bribe” (stay tuned for an upcoming test on that) and has for a long time.
3. There are two other newsletter subscription options built into the site navigation. One in the upper half of the sidebar (but below the fold on most browsers) and one in the footer. Both of these mention the subscribe bribe.
4. This data is for confirmed opt-ins (vs. people who sign up but never click the confirmation email) – meaning they fully completed the double opt-in process.
5. I made no other front end changes to the site within the time frame of the data I'm sharing.
6. This site (like all of my sites) uses Aweber (aff link) as its mailing list service.
7. Typically I would A/B test the concept simultaneously using Simple Page Tester (aff link). However, in this particular case I simply made the change site-wide as I was pretty confident the effect would be positive. And since I own this site, I can make changes based on being “pretty confident” vs. statistically sure. Had I foreseen writing the results of this change up as a post, I would have used A/B testing.

The specifics of this test

• I used a plugin called the Peanut Butter Plugin (PBB). It was cheap ($15 as of the time of this post, though they do have a more limited free version) and already had Google Analytics event tracking integrated into it. There are (countless) other options to implement a static bar such as the Hello Bar which is subscription based (no thanks). You can also code that sucker in yourself. But PBB is what I used.
• The PBBs I used was coded to match the site colors.
• I used two separate PBBs during this test.
  • One was a written CTA for the subscribe bribe that linked to the newsletter sign up page. This bar appeared sitewide except for in the “ideas” category of the site. We'll call this PBB E-book CTA.
  • The second was a written CTA telling people they could expect more ideas on X by signing up for the newsletter (with no mention of the subscribe bribe) in the PBB. This PBB included a link to the newsletter sign up page (that page does mention the subscribe bribe though). This second PBB only appeared within the “ideas” category on the site, which is the most trafficked category. We'll call this PBB Ideas CTA
• I tracked both PBB bars used with separate Google event tracking codes so I could isolate the data for each in Google Analytics.
• The results listed below are based on a six-week window beginning on January 8, 2015 and running through February 22, 2015. It compares the data for this period with the data for the six week period prior (November 23, 2015 through January 7, 2015).

An important thing to note for beginners

In the results, we're segmenting the data by conversion rate. So, why aren't we simply using the increase or decrease of subscribers to determine the success or failure of a method? We can't do that (with any reliability) because lots of outside variables can affect the fluctuation of the overall number of subscribers. Traffic can fluctuate in regard to seasonality of the site, holidays, being featured on a large site, etc. To truly measure the success or failure of a campaign like this, we need to look at the conversion rate. “Out of every 100 people who visit my site, what percentage of them sign up for the newsletter?” That number has a much higher likelihood of accurately interpreting the results of a test like this.

The results

This test resulted in an increase of 56.35% in the overall site double opt-in conversion rate.

When comparing the conversion rates of the PBB E-book CTA (21.95%) vs. the PBB Ideas CTA (27.91%), it appears that the PBB Ideas CTA won out. People seemed to respond more to the promise of future information that was similar to the specific information on the page they were on vs. the immediate gratification of the subscribe bribe.

So what now?

Now that I know the bar has a positive – and significant – impact on opt-in conversion rate, in general, it's time to start running much more granular tests.

• What happens if I make the bar stand out vs. blend in with the design?
• What if I use bright red?
• What if I use bright orange?
• What if I add the ability to subscribe (AKA the subscribe form) into the bar directly?
• What happens when I add both a name and email request?
• What happens if I only ask for their email address?
• What happens when I try different fonts?
• What happens if I try different font colors?
• What happens if I use the word “free?”
• What happens if I swap the word “great” for the word “amazing?”
• What if I make the bar taller?

You get the idea. Once you start going granular with the changes, you'll want to run these as A/B tests as it will be faster at obtaining statistically reliable results. As I mentioned above, I use Simple Page Tester to run split tests (aff link).

Future tests

I've got several more tests planned, but if you have something specific you'd like to see me test out, please let me know in the comments below.