By Rich Chapman
Since being in digital I have seen a standard timeline for businesses developing their websites. Make a new one every four or five years to experience an evolutionary leap forwards.
Is this still the right thing to do given the technology and options available to us today?
Basically no, it was never a great option anyway. Building websites, in general, is a difficult task and these days, websites are key revenue drivers for businesses, making it increasingly risky if it goes wrong. I still see cases where organic rankings plummet and conversion rates drop after so much hope has been pinned on a new site launch. It’s an emotional rollercoaster of stress, a sense of achievement on launch day and then panic.
Nowadays there is so much more available to us to mitigate the risk of launching a new website. Yet it is still untapped and companies are reticent to make the additional investment which is a small percentage of the overall cost. We all need to feel we are getting a good deal right so its an element regularly dropped from proposals.
So how do we improve this gambling situation? We need to be able to see into the future and find out how a new site will perform on launch. Good news! We can! Well, sort of…
No, we don’t have a time machine… but we can pre-test a website to see how it performs before exposing it to our entire user base and business to the new unknown. In my experience a lot of stakeholders want to have input on designs and battle for site real estate, this then defines how the new website is designed, from internal opinion alone and HIPPOs. To avoid this trap there are two ways which can give unbiased insight:
User testing
User testing outside of your own web environment can give you a level of feedback and information you simply can’t get from internal stakeholders and outside help. Even as an experienced CRO I can’t tell you for sure which new design is going to be better than your current one. We have to ask user testers what they think.
There are various techniques such as preference tests where user testers will vote for their preferred version, this type of feedback is great at the design stage of a website build.
Another is a click test, this involves finding out what a user would click on first upon landing on the new design. This ensures users are engaging and clicking the call to action most relevant for the business.
One of my favourites is the five second flash test. Users are shown the new version for five seconds and then asked some non-leading open questions: “What does the company do?”, “What would you click on first?”, “Which page element stood out the most?”. The answers from this type of test tell us how scan readers interpret the new design. Businesses can also run this test on the current version and see how the answers compare.
Any of the above can settle design debates and give real information on what users will respond best to. Designs can be updated and retested until 90% of user testers prefer a version. Not so much a shot in the dark now.
A/B testing
The other option is to start testing new designs and website experiences on the live website through A/B testing software. The software enables us to send a percentage of live traffic (usually 50/50) to a new version which is measured against the original. So let’s say designers have followed an internal brief, come up with a new homepage design and some stakeholders like it and some don’t, that’s normal. To find out if the new design really is better (and who is right) it can be tested against the original.
These rounds of testing can be done piece by piece on different layouts, images, fonts, branding, journeys and more. Gradually this gives valuable information on how users respond to the new design and importantly, to change.
Top tip
If you have a large user base and a high amount of returning visitors you can let them know that you will be launching a new website. Send them emails with a launch date combined with a promotion maybe.
One step further is to create a beta site and get feedback from users before the big switch is done. Companies like the BBC and Facebook regularly use this technique. It is a staple in the gaming industry, gamers are invited to use a beta version knowing it might break. Their reward for giving feedback is early access and feeling like a VIP, the game producers get free insight and debugging, win win.
Round up
Adding user testing and a/b testing does make a web build a more lengthy and expensive process. However, from experience, it is worth it. Web site changes can be vanity driven and a “need” to be done at a fast pace leading to errors. Going with a user led approach may be longer but it will help safeguard the business.
It’s also a mindset change, moving from completely changing a site every three to five years to constant tested small changes and evolution. An iterative tested approach removes stress, big lump sum costs and keeps websites up to date.
By Rich Chapman
CRO consultant at Impression.
Ah, but correlation is not causation, a sceptic might argue. The chart fits other explanations too. For example, both ecommerce and online advertising rely on the same technologies, so of course they grow together. Or perhaps ecommerce and online advertising are both superior to their offline versions and people have simply begun to use them both more over time.
Some of this change in the media mix is driven by lockdowns and reduced available reach from channels like out-of-home and events, but some is also because more and more people are shopping online, and that makes the second task more important.
In the chart, the proportion of total sales driven by search ads is around three times bigger in Google’s attribution tool than it is in our econometric modelling. The reason is that in two thirds of customer journeys that involve a search click, the ad didn’t actually generate the sale, it acted as a signpost, helping someone who had already made their decision to complete their purchase.
There is still a lot to learn in this area, but the above chart is typical of our limited experience. The online ads that are most often an investment into future sales are those that target new rather than existing customers and reach rather than engagement. They typically have richer creative, particularly video, and they highlight newer or less well-known products.
In other cases, the signposting job is revealed to be critical. In the right panel, the switch-off revealed that without a presence in generic search for these keywords, even a customer who had already decided to buy could be diverted and fail to arrive safely.



