This post was originally written for Alberon. I have reproduced it here because the original post is no longer live.
SEO is all about getting more visitors to your website – but just having more visitors isn’t the end goal. Ultimately you want those visitors to spend money with you, click on adverts, spread the word, etc. CRO is the process of increasing the percentage of visitors that take those desirable actions.
Identifying and measuring your outcomes
The first step of CRO is identifying what you want to achieve. Is your campaign about building brand awareness through social media, getting more enquiries from potential customers, or increasing online product sales?
Once you know this, you need a way to measure performance. This is important because making changes without a feedback loop is really just guesswork. For the above goals, the measurements might be number of social shares, number of enquiries submitted through an online form, or total value of online sales. Note that all of these can be automatically and accurately measured by software, so we can compare the results of different tests to see which convert better.
Review your site for potential issues
Before you can start making improvements, you need to work out what you’re going to change. To start, simply go through the website page-by-page, making a note of:
- What steps you want / expect the visitor to take
- Things that might distract them or prevent them taking those steps
- Anything that is unclear or confusing
- Any functionality that doesn’t work correctly
- Opportunities to improve your messaging
- Additional functionality you could add
- What best practices are you not currently following?
Ideally you should get several people to do this exercise, including some from your target audience.
Review your existing analytics data
Now you’re starting to get an idea of what may need to be changed, you can use any analytics software you have installed (e.g. Google Analytics) to corroborate these ideas as well as to find other potential areas for improvement. For example:
- If you identified a particular page that might put people off, look at where visitors went next after visiting that page – did they leave as you expect?
- Look at which pages have the highest Bounce Rate – especially key landing pages.
- Look at which pages have the highest Exit Rate – especially those further along the sales funnel (e.g. checkout pages).
- If you have any Goal Funnels set up, where do the most people drop out?
- What keywords do people search for on your internal Site Search? These could be topics that are difficult to find in your main navigation.
Plan the changes you will make
Now it’s time to take all those ideas you have come up with and organise them into a list or spreadsheet that includes:
- The change you would like to make
- How and why you think it will help
- What metrics you would use to measure the success of the change
- An estimate of how much time/money/effort it would require to make the change (relatively)
- How much of a benefit you would expect to see
- Any other changes in the list that would be affected (e.g. a full site redesign would make smaller design tweaks unnecessary)
The last three factors will help you to prioritise the list – if you plan to make major changes (e.g. a complete redesign), it may be worth doing those first; otherwise start with the easiest and most beneficial changes to get the biggest ROI.
Set up A/B tests
Now we get to the most important part of CRO – testing your hypotheses to see if they actually increase the conversion rate in practice.
To do this, set up A/B testing software such as Visual Website Optimizer on your website, and set up an experiment to run. This means you set up different variations of the page you’re testing and define the outcome to measure, then the software randomly shows your visitors either the original (control) or the modified version (variation). It then compares the conversion rate to see which is the most effective.
(Note: It is possible to have multiple variations and to run multiple tests at once, but it’s a bit more complicated and you have to wait longer for the results. For the rest of this article we’ll assume you are doing a simple A/B test with 1 control and 1 variation at a time, though the same principles apply.)
Wait for the results
This part is important! While you may be impatient to see the results, you need to observe a large number of visitors before the results will be statistically significant – meaning they are unlikely to have occurred by random chance – so you can be confident that your test was successful (or that it was not).
The exact sample size varies based on a number of factors including the old conversion rate, the percentage change in conversion rate, and how confident you want to be that the results are accurate – but, to give you a rough guide, at least 300-400 conversions per variation is recommended – if your conversion rate is around 5% that will require 6,000-8,000 visitors.
If you have a low volume of traffic it may take several weeks to reach the required sample size – but you will have to be patient, as the result may be invalid otherwise. If the traffic is very low, you may need to conduct further SEO work first to increase it, or use a slightly different approach to A/B testing.
You also need to account for variability caused by days of the week, bank holidays, seasonal changes and other external factors. This means you should generally wait between 2 weeks and 1 month before drawing any conclusions, you should measure for full weeks at a time, and you should be wary of any results generated in the lead up to Christmas.
Analyse the results
Assuming you are doing a simple A/B test, as described above, there are three possible outcomes:
- The new version outperformed the control – in which case you should make it permanent.
- The control outperformed the new version – in which case you should stick with the original.
- The result was inconclusive – meaning they performed about the same, or you couldn’t collect enough results to be statistically significant.
Do not be disheartened by negative or inconclusive results – it is all a part of learning what works for your target audience. But make sure you record the results so you can learn from them and refer back to them in the future.
It may also be useful to analyse the results in more detail – for example, the results may vary between visitors coming from natural search and those from an advertising campaign. You can learn from that information, even if your overall test result was inconclusive.
Rinse and repeat
When the first experiment is complete, you can either move onto another experiment from your list, or return to the brainstorming phase and think of new experiments to run based on what you learned. Either way, keep testing, because even small improvements compound over time to create big improvements in your conversion rate.
If you’re getting a lot of visitors from your SEO or marketing campaigns, but not converting enough of them to customers, now is the time to start thinking about Conversion Rate Optimisation and A/B testing.
Next time we will look at other ways to track your visitors and learn from their behaviour.