Strategy
6 steps to building a successful and flexible global website
December 21, 2018
Steve Peters
And most importantly, how to do it cost-effectively and with the flexibility that ensures you won't need to re-platform ever again
A guide for digital marketing managers, marketing directors and IT teams that are about to embark on a website redesign or re-platform.
In 2014, Marks & Spencer relaunched their site as what we'd refer to here at Code as a ‚big bang release'. The launch was two years in the making, cost a lot of money and came with a comprehensive feature list. The release certainly got a lot of attention, but mainly because it was blamed for Marks & Spencer's subsequent 8.1% drop in sales.
So what lessons can we learn from this? Here are our six key tips on how to avoid a simlar fate; follow them to create a site that you can not only launch quickly (avoiding the two years of scoping and build time that M&S went through), but that's also guaranteed to deliver ROI.
This sounds like the most basic step, but it is one that's often missed: what do you actually want your website to make your customers and potential customers see, feel and do?
To define your vision, start with a comprehensive market review. Map all your competitors onto a grid to establish how they're positioning themselves and their sites.
Make sure you use a broad range of examples, from direct competitors to smaller players, content publishers and other sites that may compete for your customers' attention.
We did this exercise with Amnesty International looking at a range of large charities, small single issues charities such as One.org, news websites like the BBC and The Guardian, BuzzFeed, petition sites such as Avaaz and even Amnesty's local websites such as Amnesty USA and Amnesty France.
All these competitors were mapped along an Expertise and an Ownership axis, to see if there was a gap that Amnesty.org could own. It is also worth recognising that your website doesn't have to (and shouldn't have to) play all roles to all people.
By completing a channel analysis you can see which messages might fit better into each medium, from social media to publishers. Again, this is something we did for Amnesty. We found that their channels were spending quite a lot of time in the 'mobilisation' space and identified a gap for their website to play more of an inspiration role.
The end goal is for you to have a vision for your site that is both unique and ownable to your organisation. Ask yourself, what can your website do that no other organisation can do? Use these questions to create a list of three or four things that your site must do. This is your end point, your guiding star to refer back to throughout your site's lifespan.
Get as much information about your customers as possible. You can do this using a number of different tactics, from online surveys to customer interviews and user testing. The more information you can gather the better.
You can then use this data to create a segmentation for your audience. A good way to do this is to split the users into groups based on their level of involvement with your organisation. Once you have the segments, map these audience interactions with your organisation and the customer journey they might take. It's important to validate these journeys with real users, so you can be sure that your assumptions are correct.
Your segmentation should allow you to create personas; a personal identity that you can relate to when considering what this customer might think, feel and do. You can then use these personas to inform your design by imagining that you're creating content specifically for that person. To take your thought process up a notch, you can try applying persuasion strategy to think about what persuasive nudges might encourage the person outlined in your persona to take action.
This step seems obvious, but 'KPI' as a term is banded around a lot. With so much data available to us as marketeers, it is really easy to have a huge list of performance indicators that give you a lot of information, but not necessarily a snapshot into the health of your platform and how it is performing in relation to your goals.
When it comes to key performance indicators, you should focus in on those that will really show the success of your website; we wouldn't recommend having more than four KPIs in mind. It's also important to have one KPI that links back specifically to platform health, which can encompass site speed and performance.
Once you have your KPIs, it is a great idea to put all the information into a dashboard that visualises how the site is performing.
We create HTML dashboards for our clients, so they can monitor how they are tracking against their performance indicators in real time. Having this dashboard visible‚ for example, at your workstation ‚ allows your stakeholders to see at a glance how the site is performing.
There are many tools and techniques that can be used to prove your ideas before you have to spend a penny on development time. Tree testing is a great methodology you can use to measure how easy it is for a user to find content and complete actions on your site. They allow you to test the live site and create hypotheses for change, before offering a new experience to the same group of users to see which one they prefer.
We have clients who have observed a six-fold improvement in the effectiveness of the customer experience by applying tree testing methods. This gives us absolute confidence that building the experience will be a complete success.
When looking at forms, or booking or buying processes that require multiple steps, a similar technique can be deployed. A group of users are set tasks on the 'old' site, for example 'search and find a yellow table size 100x100, add it to your basket and checkout'.
You should test and time users to observe the challenges they face in this process, then take the learnings to create new ways to overcome those problems. You can visualise this in a simple prototype that doesn't require lots of investment. This method is really powerful. We've seen users complete tasks in half the time based on refining existing processes.
The Lean Startup Circle have done some research that demonstrates that 60% of software features are waste. Think about this when looking at the long feature list you are giving to your development team.
The best way to approach any features is to focus on the most valuable thing to your business, and deliver that first. That said, it's a task that is easier said than done, especially with a large number of stakeholders with different expectations and requirements. However, this approach can be much more risk averse and can prove more cost effective.
By releasing in stages you can unlock return on investment much faster from the parts of the site that you have released first. We recently did this for our client Transform, where the most important priority for them was to update the homepage to tie into the TV advertising that they were running at the time. By focusing on this change and releasing as soon as it was ready, we were able to increase conversion by 15% and unlock further investment for the next stage in the process.
We recently launched a new search feature on the Amnesty website. As part of the delivery, we specified the entirety of the features that were requested, then defined what the initial minimum viable product (MVP) was.
MVP doesn't have to literally mean 'minimum'; it means the features that will have the most impact immediately. Once we had this MVP, we broke the rest of the features into other release packages, so we had five packages of work in total. We set acceptance criteria for each package and once launched, we measured how each performed.
After the third package, our tests showed that the performance of the search function was at its optimum level, and as a result, chose not to release packages four and five, which would have been classed as waste. It is really important to measure the success of your product, and to learn and develop from each iteration.
For more information about our approach, contact Steve at hello@codecomputerlove.com
