So I was at Mozcon 2015 when Rand Fishkin did an experiment to test the impact on CTR, and did it work.
He got the audience of about 1,500 to click on position 1 and bounce back (negative signals) and then click on position 4 and interact with the site (positive signals).
Within about 30 minutes the two websites swapped, from 4th to 1st and 1st fell to 4th and stayed like that for about 2 days. Interesting the company that was in number 1 was Amazon so a huge site.
He then repeated this test several times over Twitter and at SearchLove London 2015 as well.
Google then claimed there was a bug or something and that it shouldn’t have worked and that CTR isn’t a ranking factor.
Well as a good SEO practitioner I’ve learnt to not believe Google or most people in our industry. I prefer to run our own test to determine what is working to improve rankings and sales for our clients (please note experiments are not done on clients sites for obvious reasons, some fail and actually cause long-term harm).
Anyway I was catching away to a friend of mine over a pint or two and said I wanted to run this experiment but wasn’t sure how I could get such a big audience to trail it out, I was sure it was still a ranking factor but I wanted the data – after all we are a data-driven agency.
He suggested I should try SerpClix – basically, it allows you to buy real humans to click perform and google search and click and interact on your site. The key here is REAL humans with real browsing history, I suspect Google is smart enough to know when bots are trying to do this, but essentially you are hiring real humans to perform your queries.
You have the option to pick the country and each country varies on price or you can choose worldwide for a lower cost.
Your results must be on within the top 3 pages to be approved for their users to perform the searches.
So I thought I would do an experiment earlier in the year (remember this is on a test site and not a client’s website).
The keywords I picked I already ranked on average in the UK at 4.7 so in the top 5 and had around 800-1000 searches per month.
After only a few days of sending around 15-25 users a day I jumped from 4.7 to 2.7 and kept rising to 1.7 after about a week.
The keyword is comparing to products so ‘product x’ vs ‘product y’.
I have also started ranking for keywords I haven’t ranked for before and get traffic from them. Including ‘Product x’ review and ‘Product y review’.
Also for the phrase ‘product y’ vs ‘product x’ – so basically the query in verse has jumped from position 3.9 to 2.
The great thing is because of its self-fulling because in my opinion it’s a great article it deserves to rank there and when people land its well read and people don’t bounce back – our rankings have stayed consistent and still rank now on average around 1.9.
What does this mean for you?
If you are stuck and want to improve your rankings you need to improve your CTR. There are plenty of things you can do to maximise this.
Page titles – make these appealing, remember this is your first chance to impress the user and encourage them to click on your site.
Meta descriptions – here you have two lines to really sell your page. Use this well, but don’t give the answer away. Here are some clever tricks I have used recently to boost CTR:
- Include a call to action (click here, view more, ……. – these are my favourite 3)
- Start to answer the query without answering the query (basically, write like a politician would speak)
- Use PPC data – you can clearly see which ads have the best CTR and take the winning elements and use this in your meta descriptions and titles
That test was done several months ago and like I mentioned the results have stayed pretty similar, I am guessing what was holding me back was my CTR as no additional links have been added and the on page content hasn’t changed.
However, around 2 months ago I did another experiment. This time the result was on page 2 average position 13.1.
Again I decided to send traffic to the term of around 15-25 users a day.
I did see an initial jump onto page 1 to around position 8.7 but then that was my ceiling. I couldn’t move any higher.
Also as soon as the traffic stop coming I dropped straight back down to 11.2. So while it was an improvement it soon disappeared after the traffic stopped. My theory is the page doesn’t have some of the ranking factors for page 1, maybe not enough domain authority, the content isn’t quite good (even though I personally think it’s in the top 5), to slow to load, there is over 200 factors so it could be one of a number.
So go away and identify which pages are ranking on page 1 and look at your titles and meta descriptions, no really look at them and decide if they could be better, do they really make the users want to click through.
I really want to test this theory in the map Pack in Google and see how it affects rankings. I am speaking with SerpClix to see if this is possible as would need local users in a certain region. I will update once the test is done.