Sometimes called split testing, sometimes called A/B testing however they all mean exactly the same thing.
Testing is crucial to not only the success of an account, but also the constant improvement. I always say, “the day we reach a 100% conversion rate, we can stop testing on the landing page” but even this statement isn’t 100% true as we would then be testing to help improve the CPC and bring that down.
Questions we often hear are:
- Why do we need to split test?
- Should we be split testing?
- Are we split testing?
- When do we stop testing?
- What do we need to test?
In this article I will try to answer all of these questions and more.
Why should you be testing?
To explain why you should look at testing below are some numbers to help motivate you. To keep things simple I have kept the numbers quite low.
You can see just by improving the CPC by 5p you can have an extra 3 conversions and by improving the conversion rate by just 5% gives you an extra 5 conversions in the example above. In the real world this could mean an extra 100 conversions.
The 2 key areas of testing from a PPC point of view are:
- Ad copy testing
- Landing page testing
I will start off with ad copy testing.
The number of times I have come across accounts with only 1 ad (which also happens to be exactly the same) in every ad group is quite scary. The fact that this happens as often as it does goes to show how lazy people can be when building out an account. I will refer you to one of our previous articles about writing ad copy: http://circusppc.com/making-your-ads-f-a-b/
On average we recommend testing 3-4 ads per ad group however this will vary by industry and by promotions you are running at the time.
What to test in the ad copy?
- Different Headlines
- Different headline 2’s
- Different messages in the description
- Try price vs no price – applicable to retail only.
- Offer vs no offer (if you are running an offer)
- Seasonality vs generic
- Display URL
- Try generic vs specific
- Different extensions
- Different sitelinks
- Different call out extensions
- Different structured snippets
- Different price extensions – applicable to retail only
Moving on to landing page testing:
Over the years I have learnt it isn’t what you think is better which will actually perform better. Even if you know your products and services inside out, you need to let your users decide what they want to see.
Some examples are:
A law firm which had very content heavy landing pages with no call to action near the top tried a landing page with no information or very minimal information with a lot of calls to action. The new landing page looked much slicker and looked like a definite winner. The stats proved otherwise and the old content heavy landing page performed better. In this industry we found people didn’t like to be pushed to convert into lead but like to feel like they are making their own decision. Later we tried some calls to action on all of the content pages which worked well.
In iGaming a few years ago everyone was using the 1, 2, 3 step process. One of our clients had a homepage which was slick enough to be a landing page but also listed the games they had. They tested the homepage vs a landing page using the same 1-2-3 step process. The homepage won. This again proved a good crossover between an easy funnel to conversion but with information worked well.
So what can be tested on a landing page?
The answer is anything and everything on the page.
To give you some ideas below is a list of things to get you started:
Remember: This list is only to get you started and by no means the full list of things to test.
How to measure the results?
One crucial point I would like to point out here is to ensure you have conversion tracking. I have come across a few accounts in my time which didn’t have conversion tracking. If you cannot measure what you are doing accurately then you cannot optimise accurately.
What will determine the success or failure of the test depends on your goal to start with. What did you want to improve?
- Conversion rate
- Cost of sale
The main thing you will want to improve is you’re CPA or if you are in retail, your cost of sale or ROI. Therefore if the CTR is down but you are getting more sales at a cheaper CPA that will trump a stronger CPC and even a stronger conversion rate sometimes (Although a stronger conversion rate will usually mean more sales and a stronger CPA).
It is important to decide on the main goal however before you start the test so you can monitor this closely.
How long shall run the test?
It depends on when you feel like you have enough data. Generally I like to run tests for 2 weeks, have 1 week to analyse the data and decide on the new test to start the following week.
This however doesn’t work if you do not have enough data. If you are testing on only 1 ad group, you may sometimes have to run a test for a month. On the flip side however if you are running a test across a few campaigns you may have enough data within a week.
I have finished my test, what now?
Are you at 100% conversion rate?
In that case why don’t you start a new test to help to continue improve the performance?
- Try to test one portion of your account rather than your whole account.
- Try to run only 1 test at a time – too many tests will skew the data
- If you are running any special promotions this will also skew your figures.
- There are landing page testing platforms which exist which you can use to test the small changes on the landing pages.
- Always keep testing
For further information, or for more help and tips please do get in touch.
Written by Ahmed Chopdat PPC Director at Circus PPC Agency