Our DNA is written in Swift

ABC Testing Banners

Recently we announced that we are starting pre-registrations for developer accounts for our Product Layer project. We bought a few thousand impressions on Ray Wenderlich’s well-known tutorial website. Different banner slogans produced different conversion results.


Product Layer is meant to become the most convenient way for mobile app developers to get some basic product metadata like product name or good thumbnail images. Basically what Amazon would be if it weren’t for their terms of service which forbid use of their Product Advertising API for mobile apps.

Product Layer’s approach will foot largely on a crowd-sourced approach which makers of mobile apps – you? – acting as a multiplicator of users. Essentially if one user adds a product to the Product Layer database – and others insure the quality – then all other users of all other apps will also have access to this. This approach necessitates that we start with a large enough crowd of app developers off the bat.

The whole thing can only work if we can find several hundred developers who have ideas for apps which we can enable. Because of this we started a teaser page where you can pre-register your developer account. We set up Google Analytics to track the number of sign-ups and here’s the chart since February 6th.


The first peak came when I announced the teaser site on Twitter, the second resulted from an article where I explained a bit further. I added a link to this article at the top of each page of my blog, hoping to see some additional traffic – and sign-ups – from that.

Incorporating the logo we had gotten from a 99 Designs competition late in 2013, we created 3 different banners to put into Ray’s ad rotation. The first banner was meant to just be a teaser.

ProductLayer Banner- 1

Some feedback I got on Twitter stated that more people would click on it if there was something concrete on the banner informing the user of a tangible benefit. So we varied the message two-fold.

I believe that a new breed of apps will become possible by the combination of being able to scan barcodes with any iPhone as well as getting product information over the ProductLayer API.

ProductLayer Banner 2

A variation on the theme tried to drive home this new app category “product-centric”. We believe that being able to create apps that you couldn’t do so before might be an enormous benefit to developers.

ProductLayer Banner 3

So how did these banners perform? You’d never guess!

If you only go by the click-through tracking performed by BuySellAds you would see this:

BSA Click-Through on Banners

The “something awesome” banner was active for 6000 more impression than the others because the idea to do some A/B testing did not hit us right away. Nevertheless there is a big difference in CTR between the “product-centric apps” #3 banner and the other two. Most people did click through on the teasing one without benefit #1, closely followed by the “new breed of apps” #2. Judging from this the “product-centric apps” message was not understood by the audience.

But the CTR is only the first half of the story since we don’t just care about getting the most traffic to the site, but also to sign up as many developers as possible. So let’s have a look at the Google Analytics for the goal conversions.

Google Analytics Results

The generic banner had ad content  “banner1” during the ad-rotation and “(not set)” before that time. The ad content “new breed” is “banner2”. “product-centric” is “banner3”. Here the story looks somewhat different.

Sorted by the conversion rate the original banner did the best in bringing us developers (29% conversion), but only after we started the ad rotation. Strangely, before that time it only converted 11% of visitors.

The second best converter was “product-centric” #3. Less people clicked on this banner when seeing it on Ray’s site, but of these a higher percentage signed up.

The “new breed” #2 which got about the same CTR as the teaser #1 did the worst. Visits through it where about the same as the teaser, but while the teaser #1 converted 29%, #2 only converted 9%.


I can only explain the difference in numbers for the teaser banner because for the interval when it did better it was part of a 3 banner ad rotation. Having multiple ads caused twice the number of visits on the same teaser banner. Then it still outperformed the other “more concrete” banners. Tickling curiosity seems to be paying off more than hinting at features.

On the other hand you cannot go by CTR alone, this is quite clear from comparing the BSA stats to Google’s. We should have removed the “new breed” #2 banner as soon as it got clear that it was underperforming. It did get the second highest number of visits but for some reason the message did not connect or rather did not convert.

I paid Ray Wenderlich $2.5 per 1000 views, $125 in total for 50,000 over the course of 2 weeks. At the conversion rates shown above this equates to $6 per sign-up. The jury is still out if this a reasonable price to pay, for what we are trying to achieve.

Categories: Advertising

1 Comment »


  1. AB testing for banner ads – CTR and conversion results | BannerSnack blog

Leave a Comment

%d bloggers like this: