Customer support


It’s possibly irrational, but I still like the idea of being able to walk into a physical branch of a bank. Maybe it’s because I like knowing that if things got really bad, I could just turn up and complain in-person.
We’ve probably all felt the alternative: trying to complain to an online-only company, when all they have is a phone line, which nobody answers. You feel helpless.
The challenger banks—who don’t have any branches—are battling against this perception, trying to convince the world that you don’t need them at all.
Meanwhile, the incumbent banks are closing branches to reduce costs, and trying to optimise how many they keep open.
The trend is clear: in the future we will do less of our banking in-branch, and more online, likely through a mobile app.
Summary: There are some clear winners and losers here. But overall, the user experience of customer support is desperate for some innovation.
In this chapter I discuss:
🚀
🎠
😤
📱
How responsive is their customer support?
To answer this, I tracked two realistic customer service scenarios over a few weeks, whilst controlling as many of the variables as I could.
The first was to see how quickly I can speak with a human over the phone. This was measured from the moment I made the call, so it includes all the time I spent listening to their automated messages.
This is because it’s misleading for a bank to claim that they answer all their calls within 5 minutes, if the user has to sit through 15 minutes of announcements first.
Notes on chart above: Mon-Fri, 9-5 • Mean avg. of 5 calls • See footnotes for full results & methodology.
But this doesn’t paint the full picture, you also need to understand how consistent the banks were.
But what about the growing popularity of giving support over a live chat?
I wanted to mimic the following plausible scenario: you’re trying to send a friend some money late into the evening, perhaps for a taxi, but the app keeps crashing.
Note: Co-op, Metro, First Direct and Nationwide do not have in-app live support, at least, not out of hours.
Notes on chart above: Outside working hours • Median avg. of 5 chats • See footnotes for full results & methodology.
And when you look at how consistent the banks were with their replies, you can clearly see which ones offer suitable ‘out of hours’ support.
In fact, Revolut do something really strange: if they don’t reply within a few hours they just close the ticket.

Whilst the results of this test are clearly anecdotal, and from a relatively small sample size, it does suggest which companies can consistently provide responsive customer support.
The experience of customer support
But what about the broader user experience? That’s what the rest of this chapter explores.
1. "Press 1, followed by a hash..."
Press 1, followed by a hash, for problems with your card. Press 2, followed by a hash if you’re having problems with your mobile app. Press 3, followed by hash, if you’re having problems logging into your account. Press 4…
Wait, what was 1 again?
This is a terrible user experience. Not only is it an unnecessary memory challenge, but often none of the choices feel suitable.
Call routing like this was introduced in the 1980s, when a company would have one phone number, and it’d be printed in the local paper.
But, 40 years later, it’s nothing more than an annoyance for most users. And, it’s totally unnecessary now. This is what should happen:
👆
1. Click on ‘help’ in the app
It’s what they’ll do first anyway.
🤔
2. In-app questions
This is the digital version of ‘press 1 for…‘, with the bonus value of being able to solve many of their issues within the app.
☎️
3. Show a call button linked to a variable number
This routes directly to the right team, based on that specific problem.
So which banks still ask you to make selections while on the phone?
Can click a direct link from the app: | Still required to use call ‘options’: | Does not have a human-operated phone line: |
First Direct Lloyds Nationwide Natwest Starling | Barclays Co-op HSBC Metro Monzo Santander | Revolut |
2. The problem with queues
Imagine queueing for a ride at Disneyland, but the queue wraps around a corner and it’s neither moving, nor can you see how long it is. It’d be frustrating, right?
But although people love to moan about long queues, there are two important subconscious things influencing their behaviour, while they’re in one.
⏳
1. You’re able to constantly re-estimate a wait time
Visually, you can guess if the queue is 5 minutes, or 5 hours.
🚶♂️
2. The feeling of progression is rewarding
The physical action of moving forward creates a feeling of progression.
When either of these elements are missing, the experience is considerably more frustrating. In particular, without being able to track your progress, you don’t have as good a reference for the time you’ve already invested.
Or rather: someone who knows that they’ve completed 60% of a process is more likely to continue than someone who is 6 minutes into a process of unknown length.
Imagine how much harder it’d be to climb a mountain if you didn’t know how high the peak was.

In the physical world, these factors come with no effort—it’s just how queues work. But in the digital world its value needs to be recognised and implemented.
For clarity, saying “your wait time is approximately 20 minutes” at the beginning of the process is not the same. The real value is in the subconscious ability to constantly re-estimate your waiting time based on your progress and position.
Disneyland know this, and they’ve doubled down on it by anchoring different points of the queue with new approximate waiting times—this helps keep your sense of progress up, while maintaining realistic expectations.

You may be thinking
“Sure, but this would be really difficult to do for a call centre queue”.
Nope, Twilio do it—if you implement their API correctly. You can set it to tell the user every 30 seconds where they are in the queue, and how long the estimated wait is.
Or even better, give them push notifications via the app to show them where they are in the queue.
3. Holding music
To reuse the Disney analogy, imagine you’re in that queue—the one where you’re standing still—but this time there’s a speaker right next to your head.
That speaker is playing the same 20 second jingle over and over again, and for some reason it is massively distorted. Like, it’s twice as loud as it needs to be, and more distorted than you’ve heard from any speaker made in the last decade.
Well, this is precisely what hold music is. Just have a listen for yourself.
Something I need to make clear here is that the quality is not terrible because I’ve recorded it poorly, I promise you, that’s how bad they sound.
Seriously, listen to Santander’s—why it is so loud?
So, which banks play the same jingle on repeat?
Plays the same jingle on repeat: | Seems to have multiple songs: | Does not play any hold music: |
Barclays Co-op First Direct HSBC Santander Starling | Metro Bank Monzo | Lloyds Nationwide Natwest |
4. Persistence of live chat
It’s common for apps that require authentication—like banks—to automatically log you out after a period of inactivity.
This is actually a good thing, because if you leave your phone on a table in Starbucks, and somebody finds it, they can’t access your online banking.
But the vast majority of ‘auto-locks’ happen while you’re at no risk, like when you’ve got a live chat window open and your phone goes to sleep.
It’s obvious really, but if this happens, you’d expect to reauthorise yourself, and land straight back onto the live chat. But most of the time, your journey looks more like this:
💬
1. You’re on the live chat.
Waiting for a reply.
👋
2. App logs you out due to inactivity.
Perhaps you’re on another app, or your phone went into sleep mode.
🏠
3. Takes you to the homepage on reboot.
When you reopen the app it just loads the normal homepage.
So which banks kept persistence of the live chat, even after having to reauthenticate yourself?
App page was persistent: | Just puts you on the homepage: |
Monzo Revolut | Barclays HSBC Lloyds Natwest Santander Starling |
I should add that I tested this so many times, and neither group was absolutely consistent. Very occasionally Starling would also redirect you to the chat page, but most of the time it didn’t.
Conclusion
Every now and then with software you see something brilliant, and it immediately feels like the obvious thing to do. It’s a rare moment, and as someone who obsesses over UX, it feels like a glimpse into the future.
Well, I had one of those moments, while I was on the phone to Monzo, navigating my way through their automated decision tree.
“Press 1 for help with your bank statements…”
I actually pressed that option by accident—which nullified the speed run—but then something amazing happened: I instantly received a notification.

This is revolutionary—and I mean that literally, not hyperbolically— because it demonstrates a new way to interact with a customer.
The potential for this concept is huge. It could remove a lot of the friction from customer support:
🥈
1. To make queueing better
Push notifications to track your progress in the queue.
🤯
2. To replace hold music
Be told when you’re next—no awful music required.
🙋♂️
3. To prove your identity with biometrics
And get rid of telephone banking passwords.
👩⚕️
4. To help diagnose your problems
“We’ve sent you a notification, can you click that and tell us what happens”.
And it’s not just limited to banking. Any company that has both an app, and telephone customer support should be implementing this.
Your phone is already your wallet, your car keys, your TV and your light switch.
Now, I appreciate that some people—particularly older customers—will probably always be reliant on the traditional methods. And that’s fine, innovation doesn’t have to immediately replace its predecessor.
In fact, it rarely does.
Calling methodology
Here are a full list of the rules I kept to when making these calls.
1. 5 calls in total, on different days — I mostly did 1 call per day, but for a few of the banks due to time constraints I had to do 2 in one day (one in the morning, one in the afternoon).
2. All calls made in batches — I made all the calls in batches, one after another. I also randomised the sequence each time.
3. Avoided peak times — I avoided first thing in the morning, lunchtime and the very end of the day.
4. Timer started when I placed the call — I started the timer when I placed the call, so the times you see include all the automated messages / decision trees. The timer was stopped as soon as an operator spoke.
5. Always the same options — I would always let all the options play out (i.e, if they listed 11 options I would listen to all 11), and then select the same ones with each bank. I went for “Help with my mobile app” where ever possible.
Live chat
Here are a full list of the rules I kept to when making these live chats.
1. 5 chats in total, all on different days — All the chats were made between 9pm-11pm, on different days.
2. Random order — I randomised the sequence each time. So some messages were sent at 9pm, sometimes they were more like 10.30pm.
3. Message content — I sent the same message each time, which was a simple “Hey?”, or if there was a chatbot, i’d say “Speak to a human”. If I lied about fraud I thought they might expedite my message.
4. The timer starting — The timer started when I’d actually initiated a live chat. For example, it doesn’t include the time I spent chatting to a bot. When the bot says “Okay, we’ll find you an agent”, that’s when I ‘d start.
5. The timer ending — The timer ends when the operator sends their first message. Some banks say “connected to ____”, but then minutes go by before they respond.
My attempt to limit wasting valuable time
This was going to be an earlier chapter, but then the situation with Coronavirus worsened and I felt like it’d be irresponsible to do this in the peak of the panic.
However, by late June, I felt like as long as I could do the experiment without wasting much of the phone operators time, then it’d be reasonable to do so.
So I adopted the following procedures:
1. I hung up as soon as the operator picked up the phone. This wasted a lot of my time on hold, but barely any of their time on a call. Hopefully they just moved onto another call fairly swiftly.
2. I ended the live chat once I had a reply. In other words, I didn’t leave outstanding support messages that people would keep coming back to.
3. I avoided peak times. I wouldn’t call first thing in the morning, or at the end of the day. Live chat messages were all sent late into the evening too.
By doing this, I feel comfortable enough that I was wasting a very small amount of each bank’s time. I estimate that in total it was less than 5 minutes for each bank.
Why I used the median average, not the mean for the live chat
Typically, I use the mean average for the charts, however this time I decided to use the median.
The rationale here was that there were clear outliers. For example, here’s are Lloyds live chat response times:
#1: 65 seconds
#2: 23 seconds
#3: 4 minutes, 32 seconds
#4: 12 hours, 2 minutes
#5: 9 minutes
There’s an obvious outlier there, and a mean average would give them a score of 147 minutes, which feels unfair.
Likewise, look at Monzo’s live chat response times:
#1: 6 minutes, 10 seconds
#2: 10 hours, 45 minutes
#3: 10 hours, 38 minutes
#4: 11 hours, 20 minutes
#5: 10 hours, 36 minutes
Clearly, their overnight support did not respond to the vast majority of my messages. Their average score should be in the 10/11 hour mark, whereas a mean average would give them an average time of about 8 hours.
Median averages are much better at handling data with clear outliers, and I felt like they were more representational here.