Data-Driven Design

Hand drawn concept whiteboard drawing - A/B Testing

This is the second part of the series whereby I am turning the wonderful talks into actionable takeaways you can implement in your business. Yesterday’s article was about how to get Big links. This article is looking at Data Driven design by Oli Gardner and how it can lead to more sales and conversions.

Only a few years ago it was fairly simple to rank, you created a page bought some links and ranked highly on Google and got lots of traffic to your site so it didn’t matter too much the conversion rate as the traffic levels were high.

However that has all changed over the last few years, firstly ranking is so much harder than it was 10 -15 years ago, buying links is a dangerous game and even if you do rank nicely – there is a lot more noise now on Google. You have more ads, news, featured snippets plus a lot more meaning you can lose traffic quite easily, no longer are you just competing against the 10 blue links, you are now competing against other Google properties.

So with traffic so much harder to come by, webmasters are now having to make the sure the traffic they do get converts and this is a good thing from a users point of view.

I am not saying User Experience (UX) is a new concept it has been around for a lot of years, it’s only in the last few years it has become more important and discussed a lot more frequently.


Let me give you some quick maths to highlight the point.

Let’s say your site gets 100 visitors a day and get 1 conversion (1% conversion, this would be pretty good for mosts sites). If you wanted to get 2 conversions a day in the old method you would need to double your traffic to 200 visitors a day every day.

Whereas if you could improve conversion rate by 0.5% to 1.5% you would only need to get an extra 50 visitors a day to double your conversions. Getting 50 visitors a day is so much easier than getting 100. These numbers are only small but for some clients that get well over 5,000 sessions a day from Google alone, a small change in conversion can have a huge impact on performance.

The good thing about improving the conversion rate, it improve it for all channels so makes your Paid campaigns performance much better as well as direct and email.

Also, most sites are that poor at conversion rate we can usually double performance without increasing traffic. So we don’t need to do any SEO improvements to get an increase in performance.

As some who came from a data back ground, it’s so much easier to analyse and report especially to clients about the benefit.

It’s all about the numbers

Before we get into testing we need to identify what to test, but rather than going by what’s trendy in the industry at the moment, an explainer video or parallax scrolling or whatever is cool now, you need to use to determine these tests.

Being trendy doesn't make it right


Oli talks passionately about three key roles needed to get pages to perform better and how they need to interact better:

  • Designer
  • Marketer
  • Copywriter


He argues that these all need to work closely together to get the perfect user experience to get the best results. You shouldn’t expect a designer to start building or designing a page without content or data. Otherwise, they are just having a stab at this in the dark.

Hand drawn concept whiteboard drawing - A/B Testing

87% of marketers who he interviewed (over 700 respondents so wasn’t a small sample) said they felt they had the right to give design feedback back to designers. He calls Bullshit on this and I would have to agree with him, I am not a designer, I am more of a data analysis first and far from designer so I don’t give feedback, I am not qualified – I employ designers because they are specialists I don’t ask a designer to get me higher rankings etc as this isn’t their speciality – hence why I don’t give them feedback, I employ people who are good at what they do and I just let them do it. Where possible I try and get clients to give the designers trust (it’s ok to feed back if it’s off brand etc) but the best results we get is where we let designers do what they are good at designing the pages, but they need to work with the marketers and copywriters to get the data and content to make the page perform. Give them all the info required and let them get on with it. In the same light, marketers need to realise that copywriters in 2017 do understand Google, actually, I would argue they don’t need to understand Google – if they write for the user and not Google, this usually leads to better copy. Who remembers the days of:

Who remembers the days of:


Plus size prom dress with sash in back

If you’re a plus size girl and want a plus size prom dress, we have a lot of plus size prom dresses. We have blue plus size prom dresses, black plus size prom dresses, yellow plus size prom dresses, green plus size prom dresses, white plus size prom dresses, aqua plus size prom dresses, mint green plus size prom dresses and many more plus size prom dresses for plus size girls. This particular plus size prom dress is sleeveless and has a pretty sash in the back.


Yes we have all done it, it worked for a long time so why wouldn’t you do it, but now in 2017 it doesn’t work – actually it hasn’t worked for a few years. You need your content to sound natural, but copywriters of 2017 understand the importance of Google. Just give them the topic, theme etc you want them to write about and let them get on with it. Source Web Practices.

Only a few weeks ago, one of our clients wanted to improve the rankings of a page. They were 7th which wasn’t bad but like all clients, they wanted to be number 1. After going through the 10 results and looking at all the different metrics, there wasn’t much separating the 10. Actually two things we could see links and content. The content on their page was very weak. So I got our copywriter to look over the content on the page and to re-write it.

He took the page from about 300 words to just over 1,500 words and they key metric here was Organic Traffic from Google. We ran the test for 4 weeks and saw an 18% improvement in organic traffic to the page. Simply by letting the copywriter do what he does best and write.

But let’s get back to the numbers

Firstly everything should be tested and the winning test rolled out, even if you don’t like it. How many sites do you work on where you’re not the target audience, so why let your decision rule what the buyers want. They are they key here – they are the ones buying your (your client’s) products or services and not you. We have a client who sells Adult toys and games – I am not a  buyer of these products but I know what the data tells me.

If you’re unsure about what test to run, start on underperforming pages, where do users leave your site and try and work out why.

Once you think you know why – test it and see what impact it has.

A good place to start is in the checkout process – if someone starts this its usual frustration that stops them converting – remove the pain points and make the conversion as easy as possible. It isn’t by luck that Amazon offers one click purchase. It’s because they removed all the pain points to improve conversion rate.

Another way to get test ideas is to actually ask the users. I know asking people what they like / dislike can lead to some very good tests. It still surprises me how web based businesses don’t interact with their customers. Most customers are willing to offer feedback whether you’re online or bricks and mortar so take advantage. Just think if someone gives you that little bit of feedback that has been putting them off converting, how many others could it be affecting.

The final method I am going to suggest, but there are others is looking at what others are doing. I said in the previous article about stealing ideas from competitors and this can definitely be used here. It doesn’t always have to be in the same niche, customers don’t just shop in your niche they shop on multiple websites and once they find a good user experience this becomes their standard. By simply just looking in your niche you could be missing some great experiments to run – but also remember just because it works for them, doesn’t mean it will work for you – but that’s the beauty here, you’re letting the data tell you what works.

Key things to remember, let the data run for a reasonable amount of time, too soon and you might get the wrong results.

business man eliminate problem and find solution

Solving the problems

So you have identified potential problems/issues on your site and things you want to test, but leaving the last method aside it’s great to know you have a problem, but what’s the solution.

I don’t know if our method is perfect and it’s being tweaked all the time, it seems to work for us and me kind of stole it from Google (if it works for them) and Oli talks about it in his presentations, once we know the problem we all break away separately and try and find the solutions. It doesn’t have to be a long time. Oli says he allows his team around 10 minutes – I tend to allow a bit longer about 30 minutes to an hour but then we regroup and present our ideas. Again similar to what I discussed in the previous article about link building – there are no stupid or daft ideas. What usually happens is that not one idea is correct but we take the key elements from everyone’s ideas and create a test we think should win and then we run it.

Sometimes we get the clients involved at this stage, but sometimes it helps to look at it without the client’s prejudices. It all depends on the client and how involved they want to be.

A A B Testing

I was a big fan of A/B testing letting the data decide what changes were made until I started analysing after we started rolling out the winners weren’t seeing the results we expected.

We did everything by the A/B testing 101 rule book, but that’s the problem – customers and users don’t interact how you expect them too. It was then I started to realise maybe the control group data wasn’t as reliable as possible and it wasn’t till I was introduced to the AAB rule it started to make sense.

Basically, you have two control groups return the same or very similar results to make sure the statics of the experiment page is more accurate. If the results of control pages don’t match, we discount the data and either try and identify why and fix the issue or sometimes it just consumers going about their normal business, so we rerun the experiment. We still monitor the control pages to make sure the data is almost identical before making recommendations to the clients.

Micro Moments

While the end goal is to improve the “conversion rate” this is far easier said than done. There are so many factors which goes into the conversion, but if you start looking at the micro factors you can take the British Cycle mentality of continuous improvement you will win. They don’t tend to look to make one huge improvement but rather small mirco changes which combined together makes a huge difference.

If you’re not already doing A/B testing to improve the conversions on your site, I would recommend you start now, by logging into your analytics platform and looking at your highest exit pages and seeing what changes you can make to reduce the number of people existing your site.



Please remember these are my takings from the conference and not those of Oli, while I have used his presentation as inspiration and quoted him a few times the words are my own.

Please check back tomorrow to see How to Write customer driven copy that converts by Joel Klettke or check out the full series here.

In the previous article, we talked about How to get Big Links.

If you are struggling to A/B testing, why not check out our series of articles on A/B testing to help you improve your onsite conversions.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top