Andy Black
Help people. Get paid. Help more people.
Staff member
FASTLANE INSIDER
EPIC CONTRIBUTOR
Read Fastlane!
Speedway Pass
Breakeven to £1,000/day overnight by NOT split-testing
(Originally posted on the inside here.)
We were running a pay-per-call campaign.
Here's a mockup of a page we were sending traffic to (substantially changed, but you get the idea):
People were searching for the marketing department.
Our ad said they'd be able to contact that marketing department in the headline.
The headline of the page said it was the number for the marketing department.
Yet not many people rang the number on the page.
We were running break-even.
Everyone in our team wanted to come up with different variations of the page including adding images, changing the colour and size of the phone icon, changing the headline, etc.
They wanted loads of landing page variations to split-test, and this would have resulted in loads of work for the design team and the development team.
Everyone talked about being a "data driven" company.
Except they weren't looking at the data.
They just wanted to throw traffic at each page, and "split-test".
Let the numbers decide, and we'll be "data driven".
Split-testing?
Data driven?
Meh...
Henry Ford once said that thinking was the hardest work there is, and that was why so few people do it.
It's easier to create many variations of a page and "let the traffic decide for us", than to look at the data we've already bought, come up with a theory of what is happening, and then come up with a test to prove or disprove the theory.
I didn't want to split-test. Instead, I spent the morning analysing the data we had.
I spotted that more people were ringing a number on the page that you got to when you clicked on this link:
When you clicked that link, you ended up on a page with three different numbers for three different departments within marketing.
Everyone seemed to be ringing the second number on that page.
We weren't driving traffic to that page, so people were ignoring the number on the page above, clicking the link on the right-hand menu, then scanning down and ringing the second number on the page.
My theory was that people didn't want to just ring "Marketing", but a specific section within Marketing.
...and that their eye was travelling over to the right hand menu, where they were scanning down.
So we swapped out the page and swapped in a different one to test my theory.
It looked like this (notice the different headline as well as swapping the columns):
We didn't "split-test" by creating a second page and splitting our traffic 50/50 between the pages to see which page "won".
We came up with a theory to explain what was happening, and created a new page that we believed would perform better.
So we just swapped it in, and sent ALL the traffic through the new page, and checked stats the next day.
Conversion rates shot up, and we were touching £1k a day profit ... overnight.
So how do I split-test?
I guess I don't really.
I analyse the data, come up with a theory about what is happening, then try and prove or disprove the theory.
I don't let Google's algorithm choose which of my ads is better, or which of my ads and landing pages has a better CPA.
If I did that then I wouldn't learn anything from the data that I'm buying.
The biggest benefit of paid search is the market intelligence that you buy. Don't let computer algorithms decide what is better for your business. Decide for yourself, and test to find out why.
Acting on one small insight can make a massive difference.
Computers don't get insights, you do.
(Originally posted on the inside here.)
We were running a pay-per-call campaign.
Here's a mockup of a page we were sending traffic to (substantially changed, but you get the idea):
People were searching for the marketing department.
Our ad said they'd be able to contact that marketing department in the headline.
The headline of the page said it was the number for the marketing department.
Yet not many people rang the number on the page.
We were running break-even.
Everyone in our team wanted to come up with different variations of the page including adding images, changing the colour and size of the phone icon, changing the headline, etc.
They wanted loads of landing page variations to split-test, and this would have resulted in loads of work for the design team and the development team.
Everyone talked about being a "data driven" company.
Except they weren't looking at the data.
They just wanted to throw traffic at each page, and "split-test".
Let the numbers decide, and we'll be "data driven".
Split-testing?
Data driven?
Meh...
Henry Ford once said that thinking was the hardest work there is, and that was why so few people do it.
It's easier to create many variations of a page and "let the traffic decide for us", than to look at the data we've already bought, come up with a theory of what is happening, and then come up with a test to prove or disprove the theory.
I didn't want to split-test. Instead, I spent the morning analysing the data we had.
I spotted that more people were ringing a number on the page that you got to when you clicked on this link:
When you clicked that link, you ended up on a page with three different numbers for three different departments within marketing.
Everyone seemed to be ringing the second number on that page.
We weren't driving traffic to that page, so people were ignoring the number on the page above, clicking the link on the right-hand menu, then scanning down and ringing the second number on the page.
My theory was that people didn't want to just ring "Marketing", but a specific section within Marketing.
...and that their eye was travelling over to the right hand menu, where they were scanning down.
So we swapped out the page and swapped in a different one to test my theory.
It looked like this (notice the different headline as well as swapping the columns):
We didn't "split-test" by creating a second page and splitting our traffic 50/50 between the pages to see which page "won".
We came up with a theory to explain what was happening, and created a new page that we believed would perform better.
So we just swapped it in, and sent ALL the traffic through the new page, and checked stats the next day.
Conversion rates shot up, and we were touching £1k a day profit ... overnight.
So how do I split-test?
I guess I don't really.
I analyse the data, come up with a theory about what is happening, then try and prove or disprove the theory.
I don't let Google's algorithm choose which of my ads is better, or which of my ads and landing pages has a better CPA.
If I did that then I wouldn't learn anything from the data that I'm buying.
The biggest benefit of paid search is the market intelligence that you buy. Don't let computer algorithms decide what is better for your business. Decide for yourself, and test to find out why.
Acting on one small insight can make a massive difference.
Computers don't get insights, you do.
Don't like ads? Remove them while supporting the forum:
Subscribe to Fastlane Insiders.
