Accessibility of self-service technology = Accessibility for anyone, anywhere, anytime

Phil Day At Accessibility Scotland 2018, Phil Day of NCR deep dives the accessibility challenges and opportunities building self-service technology at NCR, one of the world’s largest providers of self-service technology.



Download Phils slides in Microsoft Powerpoint format.


Well, thank you very much. I’ll start with the apology.

I’m using Elina’s slides, so anything that’s good is probably her work. Anything that’s disjointed and wrong is probably me.

NCR, we’re one of those brands that you, probably, never really thought about, but we do … I’m trying to say it politely, but we do the things that you, probably, have to use, but might not want to use.

One of the things we do is automated teller machines, or ATMs, or as they’re known in the UK, cash machines.

These are the things that you have in the street, but you can get money, any time, anywhere. On this particular slide, we’ve got a photograph of a person using an ATM, holding a receipt, and just looking at the amount that they’ve just drawn out.

We also do my personal, not favourite, self-checkouts.

These are in many supermarkets around the UK.

They’re popular elsewhere in Europe, as well. The idea behind these is that you can scan your shopping without having to wait for a cashier to do it for you.

This photograph has a slightly older generation of self-checkout. On the left, is the area where you would put your bags. In the middle, is the actual guts of the thing, where there’s a scale, there’s a scanner, there’s a display, and there’s a card reader for entering your cash. Then, on the right, is where the basket is.

Then, we also do check-in kiosks for things like hotels and airlines. This particular photograph is a gentleman using an airline check-in kiosk at an airport.

The reason that’s important is that self-service, you can’t rely on somebody having read the manual. You can’t rely on somebody having used it before.

You can’t even rely on somebody reading the language, in some cases. So, it’s really quite a challenging area. You also can’t rely on them being able to bring their own assistive technology and bolt it on, because there are various security rules, particularly, when you’re handling money.

I’ll talk briefly about the team, just to give you a flavour of how we try and make things accessible. There’s a lot of people, in this room, who do similar jobs and, I think, it’s sometimes useful in gatherings, like this, just to talk about the practicalities, the nuts and bolts, of what we do. We have a user-centred design team.

I manage about a third of the team with particular responsibility for accessibility and usability. I also handle interaction design, as well.

Then we have a few usability and accessibility specialists. The majority of the team is made up of industrial designers, so that’s product designers if you like, and also interaction designers. What our team does is it provides support for all the hardware products that NCR provide. That’s the kind of things that I showed you earlier, so ATMs, kiosks, check-in, that kind of thing.

Because we’re based in Dundee, the majority of our team … That happens to be where most other people look at financial products, so that’s where we spend most of our time.

If you’ve had a problem with an ATM, I’m sorry.

We also, a number of years ago, we noticed that although accessibility was important, and we had a few key people in the company who knew about it, and cared about it.

Usually, these were in little pockets and the information wasn’t shared, wasn’t disseminated, and also new people in didn’t know about it.

What we did was we set up this task force. This is a fancy name for a quarterly call of a few people, and email addresses that we shared out.

There’s people from my team there. There’s people from the UX department, who design software, user interfaces. There’s people from the law department, who, obviously, care slightly about accessibility and government relations. They’re the guys who work with governments on forthcoming regulation and also other parts, generally, sales and customer engagement.

The reason I pull that out was before we did that, we were confident we knew what we were doing, but other parts of the business didn’t tend to know we existed, or that they could get help. That was particularly important because large corporations tend to go on spending sprees and buy new companies. What we found was when a new acquisition came into the fold, they didn’t know they could have any help at all.

And, unfortunately, that often meant that they were propagating bad practice.

What we have are regular calls. We have an internet site, where we share information. And we try and help the rest of the organization to know about accessibility, but not to have to read all the standards and guidelines. I know it’s heresy, but there we are.

How we do it, well, really it’s pretty basic. We have a specialist. They’re assigned to a project at the start, and they give input for the duration of the project.

That’s sounds great, but you probably all know that’s not the way it works always in practice. What should happen is, right from content phase, through requirements, through early planning, and then deployment … development, sorry.

Then deployment, and then being out in the field. We should be involved in all those phases. In practice, we’re not always, but one key thing that I’ll share with you, that really made a difference, is that we got sign-off at two key points in the development progress.

When a hardware product comes out of engineering, and goes to our production facility, there’s a particular sign-off there. We sign off on usability and accessibility.

Then, again, we do the same after production out of habit. It does mean there’s a lot of boring paperwork for us to do, but the flip side of that is it means that, if people haven’t involved us, we have the right to say, “No, that product is not good enough.”

Of course, that causes issues, because then schedules go out the window, costs spiral, and everything else. But it does, very quickly, make the point that, if you’re not speaking to us, you’re going to have a problem, so you better speak to us soon, and safe issues later on.

The accessibility specialist evaluated products at specified stages, as I’ve said. We also decided to try and put a lot of the knowledge, that was in my head, when I started, because I was the only person there. Then a number of other people, as we developed the team.

We tried to put that knowledge down on paper. First of all, that was when we were training new people, it really helped. But, second, it means that you can actually offload some of the information. If somebody asks for help, rather than you going to see the product and evaluate it for them, you could say, “Well, what are you worried about?

Okay, if you look on this document, here’s some principles for you to think about. Then you can come back to me and we can have a more reasoned discussion.”

The bit that I get excited about, standards are great, please, everyone who works on standards, you’re great and we all love what you do, but this is the bit that I really like, and it’s user testing. In the field of accessibility, in a large corporation, like NCR, there’s often the danger that it’s seen as you have to meet standards. You’re there to reduce legal risk. You’re there to make sure that we meet ADA, or UK, I call it exact, or anything else. That’s true, but if that’s all we do, I don’t think we bring much value. One of the things we try and do is we try and innovate. We look for opportunities to innovate.

In the past, we’ve worked with groups like, RNIB, the Royal National Institute for Blind People.

That was, typically, on large-scale user tests with blind, partially-sighted people, but it was also to get input very early on in the concept phases. We also work with local disability groups.

Just a few months ago I worked with two, much smaller blind societies, who were just on our doorstep. That, actually, gave us access to people who really didn’t like using technology. They were very challenging users, but, for us, it was great because it meant that we got to speak to people who, normally, wouldn’t come through the door.

They didn’t use a computer. They didn’t use a smartphone. They didn’t use any online materials. They didn’t like using ATMs. Obviously, that made our job quite tricky, but it was good to get them involved.

We also work with … I’m based in Dundee, I should, probably, have said that. We have a local university and teaching hospital on our doorstep.

We’ve worked with them on a number of things, as well.

This slide has a couple of photographs, just of very early concept testing. This was RNIB, I think, in London, and these two gentlemen gave us very early feedback on a couple of concepts that we were thinking about developing, and then taking to a user test. As a result of this pilot test, by the way, we abandoned all but one of the concepts, and had to create some more. It was good to have that feedback early, before we wasted lots of time.

Having said, monitoring standards, it’s not all we do. It is an important part of it. ATMs are one of those boring areas that tends to end up in legislation.

We have things, like building codes, and they often refer to how you get in and out of the building, the access. Then some of the facilities within the building. And ATMs, often, fall into that. You have things, like public telephones. You have water fountains. You have accessible toilets. Then, sometimes, ATMs are in that.

Because of that, and because ATMs have been around a long time, they’re involved in the legislation. Other things that we do, like self-checkout, are beginning to fall into some of these regulations, as well.

There’s differences between the standards, some of them are voluntary, some of them are mandatory, some of them are law, some of them are industry guidelines, so on and so forth.

The big lesson I wanted to leave with you is that, we really struggle because they’re not harmonized. When you look at the detail, they don’t join up.

Although, they all say the same thing, “Make an ATM accessible to someone in a wheelchair. Make it accessible to somebody who can’t see.” These are all fine principles, but when they actually give you a functional requirement, those requirements differ.

My boss has been around a lot longer than I, reliably informs me that we’ve been involved in this stuff since the ’80s. I haven’t been there since the ’80s, I’d just like to say.

Rather than waffle on, in abstract terms, I thought it was, probably, more interesting to give you a wee case study. This was something that we did, probably … I’m trying to think, two, three years ago, maybe, maybe a bit longer than that. Looking at the challenge of making touchscreen pin-entry accessible.

Pin-entry is the four digit, or six digit, or whatever, sequence of numbers that you use to identify yourself at an ATM, or for an online, or for a chip reader at retail.

The reason it’s challenging is because it has to be secure. We have a method that works quite well.

This photograph is of a gentleman who’s using an ATM with headphones plugged in. Those headphones are also plugged into the ATM. He’s listening to the voice guidance and it’s telling him something along the lines of, “For cash, press one. For deposit, press two. For transfers, press three.” It’s a very well understood principle, and with the physical pin pad, as well, it works quite well. It’s quite an efficient way of working.

Unfortunately, we saw a great deal of a trend to touchscreens. We kicked off a number of projects that looked at how we could make touchscreens more accessible, particularly, in a public environment.

I’m not going to cover all of these. These are just up there to show you there’s a timeline of activity. This makes it look like we’re really systematic and thorough. In actual fact, it was just the way it happened.

We happened to be working on a project, on the left-hand side, the Universal Navigator, also called the Uninav. We were looking at a kiosk that had a very large touchscreen, 42-inch diagonal. Making that accessible to anyone with limited reach was, obviously, very challenging. Then how you make it accessible to somebody who can’t see very well is also very challenging.

After much iteration, we ended up with what is quite a simple device. Essentially, it’s up, down, left, right navigation device that you can navigate through on-screen options.

Then you push the central select button to activate them. It’s got embedded audio and audio volume.

The next one is the one I’m going to cover, so I’m not going to talk about that, accessible touch. We’ll come onto it, in a minute.

Then there’s a couple of projects, towards the right, that are, actually, still ongoing. We’re still working away on them. One is the use of gestures for secure pin entry. That’s particularly relevant to small devices, and I’ll tell you why later, if you’re interested over a cup of coffee. Then the final one we’re looking at is just starting to look at upper limb prostheses and how people, who use prostheses, can use a touchscreen.

A bit of technical background to that.

Most touchscreens that work well outside are projective capacitive, because they work well with gloves, with rain, with dust, with all the environmental effects that you’d expect.

Unfortunately, they don’t work well with a prosthetic device that doesn’t have a path to ground.

Essentially, if there’s not a good skin contact, wherever the device marries with your body, then they don’t work.

We’ve been doing some very early exploration on understanding the problem, understanding the huge variability there are in these devices, and also trying to work out how we can make things better.

If anyone is working in this area, I’d love to hear from you, because I’ve not really been able to find much in the literature, and it seems a dreadful pity.

Because there’s a lot of upper limb prostheses users out there, who do use smartphones. But, right now, they’re having to use knitted pads that go over the fingers, that they then attach with electrodes to another part of the body. They’re really ugly workarounds, and it would be nice to do something that was more appropriate.

Going back to where we were before I took us on a wee diversion to give you the context. The challenge is how we enter the pin on a touchscreen without any audio guidance.

There are accessible devices that allow you to enter numbers. Most of you are, probably, got a smartphone or a tablet.

Both, Apple and Android, Google, have solutions for that. Unfortunately, they all involve telling you what the number is, or telling you what number that you’ve selected before you enter it.

On an ATM that’s a real big no-no, because then, somebody else could come along, find out what your pin number is, clone your card, and steal all your money. We don’t think that’s a good idea, so we had to find an alternative.

These four concepts I’m going to talk to, in a little more detail, but essentially they were early concepts looking at different ways of entering numbers on a touch screen that didn’t require you to have fine manual dexterity and didn’t require you to have high levels of visual acuity either.

One of the things we saw, pretty early on, is that it’s very difficult with modern touchscreens.

In the old days you tended to have a plastic bezel around your display. Then the active area of the screen was just the bit in the middle, and you could feel the difference.

With the trend to edge-to-edge glass, that’s been popularized by all the consumer electronics, it looks great, if you can see, but if you can’t see, you can’t find the area of the screen that’s active. It all just feels like a piece of glass.

In this photograph, you’ll see, in one of the tests we did, we actually had to stick sticky tape around the edge to give people a tactile indicator of where the active area of the screen was. This was on a concept, we then abandoned later.

The photo also shows you a person using two hands to orient themselves on the touchscreen.

Again, this is quite common behaviour, so you find a reference anchor, with your non-dominate hand, and then anchor off that, to find things on the screen.

But that can cause serious issues with the touchscreen if you’ve not designed for it. We also found something that we really should have known, but we’ve forgotten.

I’ve been doing this for a while, and I have various degrees about this subject, and I put my hand on my heart and say, “We missed this one.” In a user test we had people saying,

“What does it mean to swipe left? What’s a double-tap? How does that work? Why does a touchscreen need to be touched?”

All these really basic things that we had taken for granted. It’s a good reminder, to me, not to think that all this technology is mature, and everyone knows how to use it.

The first, particular, concept that we had was derived around the idea of having the numbers in a line along the bottom, similar to a PC keypad, along the top. You have the numbers one to nine, then zero on the end, and what you would do is, you would find the numbers on a strip, then you’d slide your finger up, onto the screen, to activate it.

That’s the basic principle. This shows an early conceptual illustration of it, where we were playing with the idea of braille. I’ll show you some other representations, here they are, where we quickly moved away from braille and started to use abstract tactile features instead.

So, instead, we looked at using circles. We looked at using lines. We looked at using grooves. Essentially, these are all ways of telling people where the numbers were, and the five had a dot, or a pip, on it to differentiate it, as it does on a telephone keypad.

We also included cancel and enter, because, obviously, if you’re entering something, you need to be able to either, clear it or enter it. In terms of the … I think this is the first test that we did, which was … What was that, 40 people? Sorry, I’m not very good at math on the spot, but there we go.

In terms of the participants that we had, we worked with RNIB in London and Peterborough, and we had 21 participants who reported themselves as not having any useful residual vision.

Then we had 15 who were blind with some useful residual vision. Then we had three reported as partially sighted. We also looked at how often they used an ATM. There was a good spread of that. So we had some, they never used an ATM, we had some that used it more than once a week, and then we had a spread throughout the population.

Of those that used an ATM, only 17 used it independently, 22 had somebody to help them. That was quite disappointing, to me, because by this stage, ATMs have been in the field a long time, and these were folks who travelled across London to do this test, so they were very independent individuals, but they still weren’t using an ATM independently.

Then, similarly, how many of them were using voice guidance, only 13 after the sample size. We also asked about their experience with smartphones and tablets. As you’d expect, quite a number had them; 18 had a smartphone, nine had a smartphone and a tablet, and the rest, 12, I think that says, had neither. And most people used the speech output.

The concepts that we ended up testing. This was in a later test. We actually abandoned, we had, I think, it was 10 concepts in the first round of testing. We tested with approximately 50 people. After the first three hours we realized that out of the 10 concepts, nine had to be thrown away, and one had to be refined quite significantly.

The two, that you see on the left, are the refined versions of that one concept. In each case, it’s a tactile strip with grooves or holes in it. The numbers are aligned along the bottom of the screen, from one to nine, and then zero, cancel is on the far left, enter is on the far right with a raised circle. Then we also had a touchscreen that had haptic feedback on it, which we were quite excited about. Then we also had a traditional ATM keypad.

The tactile strip with grooves, I think, I’ve already explained how this works. You find the number on the strip, and then you move your finger up. You then hear a beep. You know it’s been selected, and the refinement we made was, you then double-tapped and you had to activate it.

There’s a photograph showing the actual strip that we made and we used for the testing. How am I doing for time? I realized I forgot the timer. Nevermind. Okay, I’ll maybe not play the videos then, because, otherwise, we’ll be here all day. Yeah.

In general, this test required … Well, we had the majority of people able to enter their pins successfully. That meant they were able to enter all four digits, and enter without making any errors, at all. Because that’s what you have to do at an ATM, otherwise, you end up with no card, and no money, and it’s quite distressing. People also felt quite confident about the method, but we had some mixed feedback about the security of the method.

The next one was a variation on a theme. Instead of the strip being at the bottom of the screen, and then you roll your finger up, and touch the touchscreen, this time we had holes in the strip, so you touched the touchscreen through the holes. A part from that, it was more or less the same concept. Again, I’ll skip the video. Again, this tested quite well, although, less people managed on the first attempt.

If we look at the two strips that we had, we had a few errors that people often made, such as forgetting that the cancel and enter are on the ends, and thinking those are numbers. We also had people touching the holes to count the numbers, and then entering things accidentally, which, obviously, is an issue, as well. Then we had some people trying to press the strip, itself.

The touchscreen with haptic feedback, I thought this was a very elegant solution. We were all quite excited about it. The participants for the test, when they heard that it was included, were quite excited.

But, unfortunately, it didn’t go very well, at all.

This was a metal box, essentially, with a touchscreen on the top, horizontal, surface. That has a 12-key keypad on it, with cancel, clear, and enter on the right-hand side.

Essentially, what you had was a pulse for each number. A stronger pulse for the five, and then clear, cancel, and enter were vocalized.

So you moved your finger around to find the number, and then you pushed harder on it to activate it. It all sounded terribly elegant. I’ll not show you the video. But, suffice it to say, it didn’t go very well. Only 19 of the participants managed to enter their pin on this, and 18 of our very polite, very generous, participants said it wasn’t acceptable.

This told us we really had a problem with this method. The ease of entering the pin was rated as difficult, as well. People just didn’t like it. People couldn’t find where things were. They got confused very easily. They didn’t like it shaking. They either felt the vibrations were too strong, or they couldn’t feel them. We had various people swearing either, at the device or ourselves, because they got so frustrated.

Then the control condition, as you’d expect, is a physical ATM keypad. For those who’ve not used an ATM recently, or didn’t remember, we have a raised pip on the five. Everything is arranged in the same layout, so one, two, three, is always at the top, as a telephone keypad is. Then cancel has a raised X. Clear has a raised bar or a raised left arrow, if you’re in the States. And enter has a raised circle. Those also have colour coding, red, amber, and green, respectively.

These have been around a long time and, as you’d expect, most people preferred this, in terms of rankings. It has the highest success rate. People liked the colour contrast.

They found it quite easy to use and, generally, they were happy. I’m afraid some of these charts were stolen from an academic presentation, so they look pretty grim, so I’m sorry about that. Essentially, what we’re trying to show here, is that, success, in terms of entering a pin, really, you only have three attempts. You have your first attempt, second attempt, and then, on the third attempt, if you get it wrong, your card is toast. So, most people only take two attempts, so that’s what we’re showing here.

On the ATM pin pad, 37 out of the sample managed on the first attempt, two on the second, and only one failed. The tactile strip, 28 on the first attempt, with three on the second. The tactile strip with holes was 17 and 14.

Then the haptic, as you can see, only 12 people managed first time, seven on the second attempt. That told us, quite clearly, that the haptic solution was not appropriate. Again, the ratings, this is self-reported ratings on a five-point Likert scale.

The nice thing about that, without going into the details, is that participants rated the new solution, of grooves, as equally as easy to use as a conventional pin pad, which we found to be quite encouraging.

As a result of that, we chose the groove solution as the winning solution. We did actually put it into production.

This is a photo of a real touchscreen ATM that went out into the field. There’s a touchscreen a the top, 10-inch diagonal, the tactile strip beneath it, beneath that is the headphone jack, on the right is a card reader, beneath that a fingerprint reader, and then the cash comes out the bottom. This was real, but, unfortunately, it was also very innovative and it’s been withdrawn.

It’s a good example of innovation, but it’s also a bad example of innovation, because we tied the innovation in terms of the touchscreen accessibility to other innovations, like cloud-based computing on an Android operating system, which was what it ran.

I’m going to rattle through this. I said earlier that the standards that we need to follow don’t match up. This illustration demonstrates just the height requirement that’s specified for an ATM. This is all about making it reachable for everyone, for tall standing people, for short standing people, for people in wheelchairs, anyone, at all.

There is very little overlap between the heights. This is just a small set of countries, we track over 30. And, unfortunately, there’s very little area of common overlap. In fact, we only have 200 millimetres, if we followed all the standards around the world. That’s not big enough to put a touchscreen in, unfortunately. So it means that we can’t really do our job.

Rather than slag off the standards, I thought it was probably useful just to pull out a few key features that we found helpful in accessibility standards. The first one is, if you’re writing a new standard, try and make it harmonize with what’s out there, rather than reinventing the wheel. On the left is an illustration of unobstructed side reach that’s taken from the US American’s with Disabilities Act standard that was re-released in 2010.

That’s freely available on the web, by the way. On the right is a European standard that I should remember the number of, because I help write it, but I can’t. 301549, that’s about the accessibility of public procurement, or public procurement of accessible ICT, I suppose, it would be. But, anyway. You can see that, rather than create something new, they looked at what was out there and harmonized quite closely with the US standard.

Another key feature I wanted to touch on is that … ATMs, you can’t just bolt on a screen reader, because then fraudsters can use that in order to exploit a backdoor and take all the money. They have close functionality, because of the very tight security. If the standard can take that into account, it really helps. Again, EN301549 differentiates between an open system, like your desktop PC, and a closed system, like an ATM, where you can’t easily bolt on third-party apps.

This one, I think, I’m probably preaching to the choir. You should detail functional requirements, not prescriptive features. So, don’t tell us how to do something, just tell us what you want us to do. I’m going to skip through the actual illustrations with that, because I think most people know about that one.

Then, one final one, is forbidding certain technologies outright, is not helpful. When some of these standards are written, the technology, at the time, means that some people have very strong opinions about a certain device. Things can change in the future, and if you exclude things in a standard, that can have long-term effects.

What makes a good standard? Well, we think, involving end users is very helpful.

We think involving disability advocacy groups is very helpful. And we also think involving manufacturers is helpful, as you’d expect.

I realize I’ve galloped through that. I apologize for being a little rough around the corners. I also apologize … If you’ve got difficult questions, you’ll see I cunningly left Elina’s contact details on that. But if we do have any questions, I’d love to hear from you.