Top Task Analysis Lite

[00:00:00]
>> Now, I highly recommend top task analysis if you could get to do it, but I'm gonna make the presumption that maybe you aren't going to have the opportunity to do it that thoroughly. So what I'd like to do is introduce you to what I've called top task analysis lite, right?

[00:00:18]
[LAUGH] Which is where I've gone through and I've kind of hacked it around a bit. Probably Jerry would freak out if he saw what I've done to his lovely technique. And try and make it a little bit faster to turn around. It's still not the fastest and easiest thing to do, but I think now we're in a position where the return on investment of putting that effort in.

[00:00:41]
In terms of cutting dead debate and arguments and politics and all the rest of it, means that in many cases it is worthwhile doing. And obviously, on smaller sites, it's easier to do anyway because there's not as many tasks, there's not as many questions. So, here are my steps and what I recommend you do on each step in order to establish your list of top tasks.

[00:01:03]
Starts off the same, right? Start off by gathering a list of potential tasks, right? Everything that people might wanna know or might wanna do on your website or your app. And there's a whole lot of places you can look, and we've kind of touched on these already in some of the earlier stuff where I was talking about looking at existing material.

[00:01:24]
You can look at existing user research, testing, customer-facing stuff, analytics, etc, right? So you look at all these sources, and then what you do is you write a big-ass list of everything that you think someone might want to do on the website. Now, that's quite our work, right?

[00:01:43]
That bit takes a little while to do. So, if you're really need to cut things back to the point where you're on the edge of being pretty dodgy in terms of quality of data, but let's go with it, right? If you really need to cut back, your other option is to ask ChatGPT, right?

[00:02:01]
It's remarkably good at brainstorming this stuff, right? So, and you can kind of get away with skipping this stage, because later on we're gonna give users the option to add in anything we've missed, right? So our list doesn't need to be perfect. It just needs to be a starting point, right?

[00:02:23]
So yeah, it's great to base it on a load of research and that kind of stuff to get our initial list, but it's not the end of the world if we don't, right? We just need something to start people off with. So, yeah, I'll just do a prompt like this.

[00:02:39]
Somebody considering signing up for front end masters or whatever the thing is, what tasks or questions you might have when visiting the website? Write these as a list that we could use as part of the top task analysis. Sure enough, it'll come up with a load of perfectly reasonable sounding questions.

[00:02:57]
Some of them you'll look through and go, well, that's a bit rubbish. And so you'll get rid of it. It might inspire you to adds some others install it yourself, but it's actually not a bad way of starting if you're really are strapped for time, right? So we've got pretty, let's say poor quality initial list, but it is a starting point.

[00:03:20]
So the next step in the equation is to create a survey, but it's a particular type of survey we're gonna create. Well, what we're gonna do, is we're going to use a tool called poll unit. Now, there are other tools available, but that happens to be the one that I use.

[00:03:37]
And poll unit, allows you to create a list of options that people can either vote on, right? Vote up, or they can add their own additional options on it. So all we're doing is pre populating it with some initial ideas, but people could always suggest their own and then other people can come in and vote for those things.

[00:04:04]
Does that make sense? Am I making sense there, right? Kind of, yeah, okay? So hopefully I'll be able to show you it and make a bit more sense. So, what we're gonna do is, we're gonna create a new survey on pole unit and then we're gonna populate that survey with some initial tasks.

[00:04:21]
So you can see on the right hand side of the screen there. This is one I created and is running at the moment. What information do you need from the network rail website? That's the question that we're asking. And then I pre-filled it with a load of ideas that ChatGPT came out with, and my use of research came up with, just to get people thinking.

[00:04:42]
So sustainability information, information for the media, what it's like working for Network Rail. Getting involved in community rail and various other things, right? So I just dumped a load of ideas in there. Doesn't need to be comprehensive, because remember, if someone find something that isn't in the list, then they can add it themselves, right?

[00:05:03]
So, then what we're gonna do is we're gonna ask users, we're gonna send this out to users and we're gonna say, have a look through that list, right? If you find the thing that you want to know, or the thing that you care about, click the heart and it'll vote for it.

[00:05:19]
She will vote and that will be your yes I want that. If you find it's not there, you can click the big add idea button and put in your own idea and then other people can vote on that, right? So, it's basically like, you get those vote for a feature on an app sometimes, so you can suggest a feature and the very top down, it's exactly the same.

[00:05:42]
So you'd leave that running for a while, you send it out to some users, or use a tool like askable to recruit users to complete it if you need to. You send that out and you'll get back a load of results, right? Wonderful, so now we've got a load of results of people that have voted for and suggested various things.

[00:06:06]
And our job then, is very simple. We just got to clean up the results a bit, because it will be a bit messy, right? What you'll do is you'll have, for example, duplicates. There'll be somebody who has said I want us to go back to the University of Florida example from earlier, somebody might have put it in as health care.

[00:06:29]
And somebody might have put health care packages, or health care plans, and it's all health care basically. So you're gonna kind of group those ones together. And then probably you're gonna simplify the wording a little bit as well, because sometimes they put in long things like I want to know about healthcare packages where you just need it to sell healthcare, right?

[00:06:50]
Yeah, so it's doing a bit of a tidy up. Now, depending on how popular this has been, that can be quite a lot of work. Going through and tidying up those results. So once again, let's get ChatGPT to do some of the hard work for us. ChatGPT turns out to be a really useful tool for user research and testing, so I will be leveraging it a bit.

[00:07:13]
So in this particular case, what I basically say, I'm gonna give you a prompt, right? You'll get the slides, so you can see this yourself later. I want you to act as a data analyst. The attach file is the results of a top task analysis. So basically I've downloaded the results off of pol unit, opened them up and I've narrowed it down to two colons, right?

[00:07:35]
It's got a load of other junk in there. I delete all the other junk and it ends up being two columns, ideas, which are the ideas that have been submitted or I've added to the list and votes. So the ideas columns contain suggestions for tasks users might want to complete in this case on an HR website.

[00:07:54]
And the votes columns is the number of times the user has voted for the task. Please complete the following steps in order. This is important, the wording of this because I've done this many times and Chat GPT needs to be told very specifically what to do, otherwise it gets upset.

[00:08:11]
First of all, simplify each entry into the ideas column to less than three words while attempting to maintain meaning. So that's where it gets rid of health care packages and just puts healthcare. Step two, identify any ideas that conceptually similar. So healthcare and dental care, for example. You could argue those are conceptually similar.

[00:08:35]
Combined similar ideas into a single row with a clear and concise name. Also add together the votes for any idea that has been combined. So it'll add dental care and healthcare together into a single thing combining the votes. Output the new list of tasks with the associated combined votes as a new CSV file with the same columns.

[00:08:58]
The goal is to end up with a simple simplified list of less than 70 ideas with the associated votes, right? So we don't want hundreds and hundreds of ideas, otherwise it's gonna be too big for what we're gonna do next, right? But about 70. So what you will see when you've done that exercise is basically a list of 70 ideas and the, the number of votes will be very much stacked towards about a third of that list.

[00:09:27]
It'll be like, imagine a curve where most of the votes and then it tails off to this long kind of tail, yeah.
>> I'm missing one little part here.
>> Only one? It's quite a complicated thing to explain to you-
>> So you've taken all your data, you've put it into chat GTP first and then you put this-

[00:09:48]
>> Yes, so basically-
>> How do you do that and sequentially in Chat GPT?
>> So, in chat GPT, you need the pro account events, you need to pay a bit money, right? But you only need to do it for one month. And then it'll give you a little paper clip next to the search text box.

[00:10:07]
It gives you where you can upload a file, right? So it knows the file is attached, does that make sense?
>> I didn't know that part about it.
>> Yeah, well, if you haven't used the pro account, you wouldn't know. I think Claude, which basically is ChatGPT, very similar.

[00:10:23]
I think that lets you upload a file on the free account, in which case, use Claude. It doesn't really matter, any large language model will do the trick, right? So the great thing is that's gonna now churn out this list of tasks, and you're gonna be able to see very clearly what the top ones are bonus, right?

[00:10:43]
So immediately you now know what it is that the website's got to be built around, right? Yay, anyway, what you'll end up with is a list that looks something like this, right? So the top third of the list are gonna be your top tasks. Notice how it's slowly tailing off and it goes down further until nothing.

[00:11:01]
So most of the top third of your votes are gonna be, really stacked towards a certain number of tasks at the top. And those are your top tasks. Now, that might be enough for you, right? If time is tight, money's tight, you might go, great, I know the top tasks.

[00:11:24]
Now let me create an information architecture around it, right? However, if you think there's gonna be a lot of arguments about the information architecture, if you want to do something that's a little bit more user-orientated. We can actually use those top tasks to define our top-level navigation, right?

[00:11:47]
And this is where it gets really, really quite powerful, and you end up with an information architecture that is rock-solid. A, it'll work well for users, and B, it'll be really difficult for anybody to argue against. Not that I'm obsessed with shutting stakeholders up, but sometimes it's needed, right?

[00:12:08]
So we're gonna use top tasks to establish our top level navigation. So why focus on the top task? Well, it's crucial that users can complete their primary tasks and those tasks are readily accessible, because they're the main actions people want to do on the website, right? And it's imperative that the smaller tasks don't get in the way and stop you doing the top tasks.

[00:12:31]
It's the old needle in the haystack thing, you can't find the one thing you really want to do, because of all the other clutter. So that's why we're focusing on top tasks, but why top level navigation, right? We're gonna start by looking at the top level navigation, because there's been some really interesting research done around this, right?

[00:12:51]
If a user gets their first click correct when they're trying to complete a task, they've got an 87% chance of completing the action correctly. If they get the first click wrong, if they start off by going in the wrong direction, that drops to just 46% chance. So that first click, in other words the top level navigation is absolutely critical.

[00:13:21]
So if we can build our top level navigation around the main tasks that somebody wants to complete in your app or your website, then the chance of them successfully complete finding that top task is gonna be much, much higher, all right? So that's all logic. So how do we build a top level navigation around top tasks?

[00:13:47]
Well, we do that with card sorting.