Do you do A/B testing on your affiliate sites?

[MichaelCorfman]

Affiliates often have choices to make when building their affiliate sites.

Do I feature Casino A or Casino B?
Do I run banner A or Banner B?
Do I feature news articles or reviews?
Do I use a banner for a site or a screen shot with a headline and description?

Oftentimes affiliates make these decisions based on their opinion of what will work best. Sometimes affiliates will conduct actual tests comparing alternative approaches and measuring which one performs better.

How about you? Do you ever perform actual tests comparing and measuring the performance of two or more alternative approaches?

Besides answering the poll question, be sure to share with us actual experiments you have run and what you found out in the process.

Michael

[Alain]

Hi,
I've been testing aff sites regularly with OpenX. This open code based ad server allows serving ads / banners using a wide range of filters as well as rotating between different banners of the same brand or different brands.
I mainly use the geo location filter and banner rotation for ctr.
In my opinion its also a very good ctr tool that can help you determine which brands get clicked the most. Later this info can be compared to sign ups / downloads and Real depositors in order to get the full stats.

Another good tool is a WP A/B plugin which can switch between different themes and give you a good feedback on what works better.

Hope it helps,
Alain

[jackten]

You aren't operating a real business if you don't do tests.

That's one of the main things separating a hobby (doing what you feel like) with a business (doing what makes money).

[grem]

It would also depend on what aspect of your site you are A/B testing however almost everything can be tested and studied for optimization

[TheGooner]

NO - I do NOT do A/B testing - and I have no plans to do so in the future.

We get 10,000 people a day, from different sources, via different keywords, for different promotions. To assume that everything is equal from one period to the next is simply wrong.

There are many, many variables involved in the flow of traffic, the makeup of that traffic and events going onsite, that to assume ANY relevance to different clickthrough rates is as a result of the banner is unjustifiable.

The effort to set-up specific campaigns and ring-fence the testing to isolate the different variables simply does not justify the time spent - especially when any conclusions are rendered void by the vast variability of the traffic.

A busy commercial website is not a lab.

[Shane]

We've received enough requests from affiliates who want to be able to do this that we've built an ad serving option into referback's media library.

Some use it to A/B test and other simply to serve multiple brands in one location - based on TheGooner's logic which says why only show one brand to various visitors as it might not appeal to everyone.

Serving multiple brands and/or variations of banners (game focused versus offer focused versus Unique Selling Point focused etc.) raises the chances of hitting more visitors 'sweet spot'.

It also allows for serving various language and currency banners based on visitor IP and browser language settings.

So whether purely your A/B testing or not, using some form of ad serving is very valuable.

[EugeneK]

it's all depends on the site scale and quality.
if the project has high traffic the testing is quick and adequate.

[casinogod]

We have done some A/B testing but it is something I think we should be doign MORE of. It does not even need to be done casino versus casino, I analysed some of our stats yesterday where we had links to the same casino but with different landing page from the same areas of the same site (and during the same time frame). It turned out one landing page was converting clicks to players 6 times better than the other.

I know there are always variables at play like position each link on the page and wording of the link that where indeed different and could make a difference but I have now set up a new trackign pointing to the better converting landing page usign the exact same link text and locations etc as the poorly performing landing page and I am fulyl expecting it to convert better (of course only time will tell but a 6X difference has to be more than just these few differences).

Also you can always geo-target banners and links so you can ensure that at least the casino you send to is one that accept players from the region you are sending.

The only thing stopping me doing more A/B testing is time, one of these days I will have a nice system made that lets me see all this kind of thing.

Wow Shane I never knew you guys had an ad server in you media library, I will have to take a little look at that.

Although as per our chat I think you shoud focus on getting campaign tracker reports sorted out first before making any other widgets as per our chat in the swimming pool the other week! (Yup thats right folks we actually talked about campaign trackers in a swimming pool, maybe we should start a new poll asking where's the strangest place you have talked shop? lol, thats gotta be a contender).

[casinogod]

A agree with what youa re saying, I have studied the click thru rates on banners in various positions on our key sites until I learned what our visitors seemed to prefer. But an ad server would make the entire process a lot easier. I did look at openx but as far as I could tell I would need to install it on each site I wanted to use it on IF I wanted the banners to be hosted on the local site (eg banner hosted on site a.com for site a.com and hosted on site b.com for site b.com) but that would mean having lots of logins rather than just controling all from one central point. Do you know a work around or plugin that caters for that need in openx?

Hi,
I've been testing aff sites regularly with OpenX. This open code based ad server allows serving ads / banners using a wide range of filters as well as rotating between different banners of the same brand or different brands.
I mainly use the geo location filter and banner rotation for ctr.
In my opinion its also a very good ctr tool that can help you determine which brands get clicked the most. Later this info can be compared to sign ups / downloads and Real depositors in order to get the full stats.

Another good tool is a WP A/B plugin which can switch between different themes and give you a good feedback on what works better.

Hope it helps,
Alain

[jackten]

There are many, many variables involved in the flow of traffic, the makeup of that traffic and events going onsite, that to assume ANY relevance to different clickthrough rates is as a result of the banner is unjustifiable.

I respect your opinion, but you're talking about a website that's already established and you most likely know the metrics of your promoted brands and conversions.

I also disagree with the assumption that different traffic makes CTR irrelevant. People are people. The title or headline A is going to attract more attention, traffic and conversion than option B regardless of having many sources of traffic.

Good old AIDA: Attention, interest, desire, action.

One of the worst advice you can give is not to test. The goal of my websites isn't to generate thousands of visitors per day. My goal is to convert visitors into sign ups and long term players. That's what everything on a website should be geared towards and you accomplish this with testing.

It's not rare for marketers to see a 200%+ increase in conversions by changing small things after running tests. Who here can realistically say they really don't care about making more money for the same number of visitors?

I recall Gooner saying his website is covered with banners because he figured out it makes more money. That's a test, even if no software was used.

[casinobonusguy]

I agree with Gooners Logic on this one. Also I work with partners I feel I can trust ,If I am earning good money I tend to give a program the benefit .One thing we find from time to time Only with RTG is if we have special landing pages sometimes these are deativated for no reason .So after a couple experiences we stopped using them.I use my bottom line to determine if we will continue working with Casino A over Casino B.
Casino B can use argument if we GAVE THEM Casino B positions they will earn more but it is getting tougher to mess with good money to take a chance on something different.
BTW Shane where have you been hiding ?lol

[casinobonusguy]

We always did redirect links and used static banner/links but last month we switch a couple sites to use Connextra system for bodog ,unibet and couple other programs.We have noticed a bit of improvement in stats.We only change our banners every 3-4 months at best so considering to use more direct links from the casino affiliate back end .I think they can do a better job than us and especially with programs like wagershare and c-planet by using their codes we see refering url , keywords etc which helps us with our on site seo .

[TheGooner]

I recall Gooner saying his website is covered with banners because he figured out it makes more money. That's a test, even if no software was used.

Now I did not say that I don't change, tweak and test things for conversions - you really shouldn't take large leaps to conclusions that have no basis on the information - mind you that's exactly what A/B testing does!


Of course we monitor performance of partners and pages .... but that's a huge leap and jump away from conducting an A/B test and then drawing very speculative conclusions about banner clickthrough or text effectiveness that are usually statistically incorrect because of a wide range of other variables.

I understand the concept of mass marketing - but that's a concept that is better aimed at broadcasting across TV, radio or press when you might reach a consistant userbase - and cannot possible identify your audience.

On the internet - people are not just a resource called "people" - and if you are thinking that way then your are probably not conducting your onsite advertising efficiently. They are all individuals and will react differently.

We've done a lot of demographic research - based on visitors from 170 countries with interests in four different verticals, and have drawn up a list of our top 10 targets for internet visitors that we try to reach.

If you stop and think about your visitors / readers, and the vast number of different triggers that they have, then you will realise just how diverse your traffic can be ...

Typically one of our pages will have 8-10 different advertisements scattered throughout the content - and depending on the users location, and the way that they entered our site- 40-50% of those ads are specially targetted at them.

The OP question and poll was about A/B testing. A/B testing simply does not work for us - as the significantly varying demographics of our visitors to each page or banner means that there is far more variance going on than the simple change that you want to test.

[jackten]

On the internet - people are not just a resource called "people" - and if you are thinking that way then your are probably not conducting your onsite advertising efficiently. They are all individuals and will react differently.

I wish you'd elaborate your view on that more clearly (no offense, asking legitimately).

How are people not a variable called people? A/B testing is nothing more than measuring statistically what most people do, regardless of if they come from Canada, the UK or China. People typically follow set patterns.

For example, it's been established that putting your opt-in form in the top right corner of the screen increases conversions significantly. That's where the average viewer's focus goes when a web page loads. This has been discovered through testing. We're not talking about 5% changes over a small sample. People who've redesigned their website to put focus on the top-right opt-in form have seen massive increase. I saw a case study where the website owner reported a 300% increase in his opt-in rate.

Of course, someone can argue that not everyone reacts the same way. Some people will opt-in even if the form is buried at the bottom of the page. But as a business owner, I go for the low-hanging fruit. Which, again, is found with tests.

What you seem to be talking about, different visitors wanting different things, is to personnalize your website based on country (geolocalisation) or tastes (offering multiple options). That's very wise, but in no way does it reduce the empirical value of A/B tests.

[craig96]

I have seen some terrible affiliate sites in the past and still today, these sites do look like a project that either failed or is making money so they never update their site or it's design.

With that being said if you design a site that you yourself would use why would no one else use it? As long as your eyes do not bleed looking at a site then you will use it.

[KasinoKing]

You aren't operating a real business if you don't do tests.

That's one of the main things separating a hobby (doing what you feel like) with a business (doing what makes money).

That's me then!

I'm a gambler - a player - it's my hobby.
I do what I feel like and it makes me money - what could be better?

I don't do A / B testing.

KK

[xecutable]

I test once on a while, and imply all sorts of other testings as well. Considering the length I have been in this industry I believe this is justified and it's part of the learning curve.

I believe everyone has their own views on this and there's no right or wrong answer, as all of our traffic comes from different sources, for different reasons at different times, therefore it's not really comparable.

I do not believe in common logic, or advices, although I can take them into consideration, I prefer to see something for myself. It's something I enjoy doing as well. I'd never test if I hated testing, and do it just cause somebody on the forums said so! Therefore all sorts of arguing around here is pointless...