29 May 2020
By Abi Cox
It wasn’t long ago that we launched our new website. We took a user-centred design approach and promised to keep testing and improving it. We knew we had to make sure people could get the information and advice they needed.
A global pandemic wasn’t part of our plans, but coronavirus has shown that our website is one of our most important assets. When it comes to our website, nothing is more important than getting the content and language right.
We’re still learning - by no means have we got everything right. But we wanted to share what we’ve done so far, for anyone who wants to get started doing the same thing. And we’d really like to hear from you if you find this helpful.
How do you know your advice is compassionate?
‘Be compassionate’ is one of our values as a charity. So we need to know that, pandemic or otherwise, we are showing compassion with our advice.
Our Ealing drug and alcohol service helped us test this with four volunteers who previously used the service. We asked them to read our relapse advice page and highlight everything that seems caring in one colour, and anything that doesn’t seem caring in another. This is called a highlighter test. (Thank you to Pete Gale for your helpful gov.uk blog about how to do this.)
We learned that the most caring phrases on the page were those talking about avoiding social isolation, so we moved this advice to the top of the ‘preventing relapse’ section.
I think it’s all pretty right. I’ve lapsed myself a few times. It’s very good… I think a lot of people feel guilt and they tend to isolate themselves… so the thing is to get out there.
We also made sure we mentioned keeping in contact with people over the phone and video call in our coronavirus advice for people who use drugs.
Can people understand your advice?
It’s tempting to rely on subject matter experts within your organisation when creating content. While they are the right people to give you the facts, you might find you’re getting the language wrong. That’s why it’s so important to work with the people you’re writing for.
To measure whether people could understand our content, we used the ‘cloze test’ technique. You take a passage of text, remove every sixth word (or whatever number you choose) and then ask the user to fill the gaps. It seems like it shouldn’t work but it really does.
If the user understands the text, they should guess at least 60% of the correct answers. We used the tips for cutting down or quitting alcohol page for this test.
3 out of the 4 participants passed the test, which was encouraging, but there was a 100% failure rate on 2 phrases:
1. “Write down how much you’re drinking, to help you keep ____.”
The correct answer is ‘track’, but everyone wrote ‘a record’. This is clearly more natural language for users.
2. “Have a look at the NHS’s 5 steps to ____ wellbeing.”
The correct answer is ‘mental’. Everyone wrote a different answer for this.
We have since made lots of small tweaks to the page to use language that seems to be more natural and therefore easier to understand for people. This test is very easy to do - we’d recommend giving it a go.
Do people know what action to take?
Our ‘I need help now’ page is one of our most important pages. We know that people are looking for urgent help. So the information must be clear, calm and helpful.
We asked people to read the page, then we took the screen away and asked some questions to test whether they knew what to do next. We asked them questions like ‘what are your options for getting help?’
Lots of people got the answers right, which was a great result. But it was important to recognise that we were testing with people who had used our services for some time, so they might have already known the answers to the questions. If you’re going to do this test, we’d recommend you use a page with new information people wouldn’t already know, and ask open (rather than yes/no) questions to really test their knowledge.
Making sure our coronavirus advice is helpful
Over the last few months, we’ve been publishing lots of advice to support people who use our services and people who use drugs and alcohol. The pandemic meant we had to move quickly, and the circumstances meant that we had to change our approach to testing, while still trying to validate our content as much as possible.
1. We asked people to read pages aloud
Soon after we started publishing coronavirus advice, we worked with two of our volunteer peer mentors to make sure we were getting it right. We asked them to read two of our pages aloud over the phone and comment as they were reading. We made it really clear we were testing the content, not them, and they should say whatever they were thinking without worrying about hurting our feelings!
It’s remarkable how much you can learn from this exercise. Firstly, you find out what you’ve missed, because people naturally have thought associations as they’re reading aloud.
For example, one of the peer mentors we spoke to mentioned that people were unsure what to do if the phone line for their service was busy and they couldn’t get through. So we added some advice around that.
Secondly, if you listen carefully as people read, you can hear when they stumble over a word or phrase, or take longer reading something, because it’s not simple enough. This helps you improve the language.
Luckily, the feedback was mostly positive:
The way you’ve worded it is absolutely brilliant. I’ve come through to stuff before and it’s like 'what'? Too many big words. Service users like the immediate relief - you see it, it tells you what’s what, it doesn’t cause panic.
2. We used social media to ask for feedback
Generally whenever we published a new page we would put a call-out on social media for people to let us know what they thought. We got some useful feedback this way. For example, when we published our benzos harm reduction advice, one of our Twitter followers suggested we should mention that a lot of benzos are varying in strength at the moment.
We were able to write a few sentences, get sign-off, and publish them on our site within a few hours of receiving the feedback.
As a team we recognise we don’t know everything, and the best people to help us get it right are the people we’re writing for. So we always try and ask ‘What do you think? What have we missed?’
3. We used data to understand what kind of advice people need
We have been running a campaign on Facebook for a few weeks to support people who have found their drinking is creeping up during lockdown. We wanted to try pointing people towards our webchat service so they could speak to our online team for advice.
Looking at the data we saw that over 2,000 people clicked through to the campaign page but not a single person used the webchat service. What went wrong? Looking at Google Analytics, people who clicked through viewed an average of 2.45 pages, so it seems this group of people perhaps aren’t ready to speak to someone about their drinking yet, but they are interested in reading some advice. This was a useful exercise in showing that webchat isn’t always the solution. The people that responded to our adverts didn’t want to chat, they wanted to read. This meant we could save our webchat resource for people who would most benefit from a human interaction.
What do people want to read?
Over the last few weeks, we have been collecting data using Hotjar, a website behaviour analysis tool, to find out how people are interacting with our site. This is a really quick and easy way to get a sense of what people are interested in.
A pandemic has the potential to create information overload, so heatmaps and anonymised visitor recordings help to show us what people want.
The next step for us will be to go through everything, watch all the recordings, analyse the click and heat maps, and come up with a list of improvements. This will be really helpful as we try and strip out any content that is no longer useful.
Some resources we used
- Nielsen Norman Group has some great information on how to test content.
- This article by Angela Colter on testing content is really good.
- We used this user research in government blog to help us with crafting the highlighter test.
- This Medium article by Beant Kaur Dhillon was helpful for learning about the best testing techniques.
- We’d recommend Hotjar for some quick insights into how people are using your site.
What did you think?
We’d love to hear what you thought of this article – did you find anything useful? Is there something we missed? If you have any feedback or you’d like to talk to us about how we’ve been developing our website, please feel free to tweet us at @changegrowlive. We’re always looking for opportunities to share ideas with anyone who’s supporting people to change their lives.