Rough and, for the most part, unedited notes taken during a mind-sizzling and extraordinarily useful masterclass with testing, content, IA, and usability whizz, @gerrymcgovern. The workshop was organised and hosted by the wonderful folk at Together London.
It must be said that these notes are not word-for-word and cannot encapsulate all the knowledge shared during the masterclass. Instead I have tried, as much as possible, to summarise the wealth of information that was fired my way. If you find what you read interesting, I suggest you find out more about Gerry and pester him for another masterclass near you. Or pick up his book, which I still have to read too – The Stranger’s Long Neck.
Some inspiring audio from Gerry
Gerry talks about the sell-in value of testing, research and evidence, the actual value of all those, and how we might speak to our clients.
Gerry doesn’t talk to clients about content, he talks about people.
Great websites never tell you about themselves, they help you to do what you want to do.
All websites are by nature self-service. Self-service websites should be run by service professionals.
Content and technology are the tools and means; owning the customer is where the future is.
Websites with the best customer relationships will be the most successful.
Understand how the customer behaves and thinks. Understand the words they respond to.
Behaviour-based metrics: understand how people behave by gathering metrics for behaviour, not just metrics of usage.
Audience-based navigation: 80% of the time, audience navigation doesn’t work. For instance, professors at a university were clicking on ‘Students’ instead of ‘Faculty’ in order to submit content/find out about students. Also, very often (with research) you will find that you don’t actually have that much unique content per audience group. In creating audience-based navigation, you simply give the impression you have unique content. Gerry used to be pro audience-navigation but he has in practice found it very problematic.
Rigorous testing shows that a single well-designed architecture/navigation path is optimal. This design should be based on studies of audiences and their ‘top tasks’. I’ll expand on top tasks later in this post.
Humour (Gerry has it in bucket loads): There’s a strong connection between drugs and the web – you’ve got users, traffic, and hits!
It’s a dehumanising process to call people users. As communication professionals we need to care about the people we are aiming to serve – our ‘users’ care about themselves – if we don’t care about them we will be on the losing end. You’ve got to develop empathy for the people who use your services. Touch points and human interaction (even if digital) are vital.
The worst way to design a website is to have 5 smart people in a room drinking lattes. The second worst way is to have 15 people in a room playing focus group. Design from real life experience and observation. We need to understand what people do rather than what they say they do.
The focus turned to tasks
Gerry talks about top tasks and tiny tasks.
Top tasks are those tasks that are most important to a person.
Tiny tasks are those that are least important.
Stakeholders are often scared to prioritise tasks and/or de-prioritise or completely let go of tasks that might be tiny arguing, ‘well, someone might need them?’
This question is often asked:
Q: Will focusing on top tasks hurt the tiny tasks?
A: Not getting the top tasks right means that tiny tasks will definitely be hurt. If you can’t get the fundamentals right no one’s going to stick around for the tiny tasks. Making the fundamentals or top tasks easy and quick to complete is primary.
Rigorous research, analysis, and testing underpins everything Gerry does (or at least everything he shared with us).
As a simplistic overview:
In finding out how to create a website experience that is satisfying and fast for the audience, Gerry focuses on finding out:
- What tasks people want to perform,
- and out of those, what the most important tasks (top tasks) are.
Gerry presents this data in spreadsheets, graphs, and pie charts so that it may be easily assimilated by everyone involved (stakeholders and team members) and so that changes might be agreed upon and implemented.
This no-assumptions/data-based methodology allows for unequivocal tracking of results achieved and, where success has been achieved, the ability to present a clear business case for rolling the process out to other areas of the organisation.
Top tasks study: whatever you do, do it well
If you don’t do a top task study well, don’t do it at all – or take a tiny segment and do it properly. For instance, study just Products or only Support – a single section to prove that you can get results. If you can attain positive results in one small area of the organisation, you’ll find it easier to sell the process into other departments of the organisation. Doing a tasks study badly means that top stakeholders will have a field day in finding gaps and inconsistencies and you’ll have lost any hope of selling a research based (and initially more costly, though in the long term potentially more profitable) way forward.
Be meticulous, methodical, and take your time.
A top tasks table, if done right, can produce the level 1 and level 2 information architecture of the site.
How to do a top tasks study
I think I’ve got this down but it was a lot to take in over the course of one day so there are gaps; this will however get the brain-ball rolling.
A list of top tasks needs to be created, this list might consist of 700 or more tasks. It might be very long, to say the least. Gerry calls this list the longlist.
To create the longlist, gather and analyse information from the following sources:
Find out whether different demographics have different top tasks by creating tabs within your spreadsheet for each. (I took pictures of these tables but they’re so blurred they’re not worth sharing. Sorry about that. Does iPad 3 have a better camera?)
You might have tabs for:
– General (you will always have this tab)
(These should change depending on circumstances.)
You will often find that there’s not much difference in results between demographics – another reason why audience-based navigation does not work.
A demographics study is however worth doing because:
- You don’t know that there won’t be important differences unless you look.
- If there are indeed no differences between demographics, you will have wonderful ammunition to sell the single navigation path to key project stakeholders.
Getting from a longlist (of tasks) to a shortlist
Once your long top task list is complete, the list needs to be refined down. Drastically.
To do this, hold 7 to 10 sessions with chosen stakeholders and experts. People involved in these sessions must be experienced people – people who can make educated decisions and who have genuine experience and knowledge about the specific service environment.
Gerry elucidated on these points in length. In short however, these are tasks that should not make it into the shortlist:
- Do not present brands – every phrase must be immediately obvious to the audience, brands are not immediately obvious.
- No goals – ‘I’d like to be happier’, these are limitless pie-in-the-sky statements.
- No organisational words – not ‘dealer locator’ but ‘shop locator’ can make a big difference to sales. What are the words your customers use?
- No tools or formats – e.g. calculators, interactive health tutorials, podcasts, videos. The question is, what does the tool do? This is the task.
- Not too high, not too low – the task shouldn’t be too specific nor too all embracing.
- No overlaps – delete tasks that are saying the same thing.
- Avoid verbs: if you can get rid of verbs, get rid of them. If you can get away with just the noun, do that – e.g. instead of ‘find training’ use ‘training’.
- Don’t have two tasks made as one e.g. ‘Choose a doctor, find a clinic’. Instead have ‘Doctors’ and ‘Clinics’ for instance.
- Avoid acronyms.
- Max 7 words/55 characters or less. It needs to be easily scannable.
With the shortlist complete, it’s time to survey
Gerry is a fan of Survey Monkey saying that it is cheap and brilliant.
Use surveys to find out what people think the most important tasks are to them. The survey should start with a few questions to gather information about the person completing it, followed by a list of tasks from which they should choose their most important tasks and rank them (I have assumed the ranking part).
Creating surveys: a few tips
- Surveys often start with category question such as gender, age, and locations etc. Have no more than 5 category questions.
- Make sure these questions are aimed at answering your questions about the importance of the differences between audiences/demographics.
- The length of your survey list (list of tasks) is important. Contrary to easy assumption, a list of 80 or more tasks is more effective than a list of 20. Longer lists force people to scan for and choose only those tasks that truly bounce out at them. It forces them into a natural and spontaneous response – similar to that experienced when browsing a website.
User testing: things to think about
User testing needn’t be an expensive or enormously extensive affair, especially to start out with.
- Gerry uses GoToMeeting but any similarly good online meeting service will do.
- Determine 6 to 12 frequently used tasks, which should be completed in an hour.
- 5 to 8 people is enough to determine if there’s significant problems in an environment. For initial testing, these people do not need to come from specific demographics. At this point, any group of people will reveal the sites biggest problems.
- Always press record then ask permission, it’s good to have it on record.
- Have both your script and their shared screen open on your desktop so that you can refer to both.
- It’s best to read from a script initially; when you’re more experienced you won’t need a script as much.
- You should be as un-interrupting and unobtrusive as possible. Like a good waiter, they should hardly know you’re there.
- Your communication style should be unemotional and dry. Gerry says, for the most part, just shut up.
- Use a timer to track how long each task takes – but don’t start the timer until the cursor moves. Sometimes a person will take 15 to 20 seconds to actually start the task.
- Gerry typically reads the task to the person first. They may want to read along. Give them the option to ask you not to read it out.
- Doing quick and cheap testing every 2 weeks offers the potential of significant improvements to an organisation.
Testing: task question checklist
Make sure questions are :
- Universal – everyone can do it. Tasks must be universal to the audience you are testing.
- Emotionally neutral.
- Not confidential.
- Clearly different from one another.
- Independent from other tasks. Do not bundle tasks together.
- Short – 30 words or less (if possible).They should:
- Describe one task and have one unique answer.
- Not contain hidden clues.There should:
- Be no change to the site within the testing period (so that all subjects experience the same site).
A sample of introduction text for a remote testing session:
As mentioned, I’m still getting to grips with this information so there are certainly gaps but I hope at least some of it makes sense and sets thought bubbles bubbling. A good deal of this workshop was also tweeted (by me) under the hashtag #tlmasterclass.
Many thanks again to Together London and of course to Gerry, for an interesting and inspiring day.