Conversational Analytics: Finding value from listening

0:01
So hi, everyone. I’m here with Frank Sherlock today, who’s the VP of international call miner. Frank, very, very welcome. Good. Good to chat with you today.

0:10
Thank you, Chris. So just a few words about what we do in call miner. We’re an intellectual interaction analytics company. So, in essence, what we do is we allow our customers to gain insight from all of the interactions across audio web chat, and that email and other channels to 10, which is data, we turn that unstructured data into structured data to help improve business performance, improve customer and employee outcomes. And it’s a pleasure to be here, Chris. And looking forward to our conversation, though,

0:47
I thought it was I thought it’d be good just to chat a little bit. First around, what are some of the changes that that you guys have seen through through COVID? I mean, it’s nearly a year now, certainly in the UK since we were locked down. And I mean, what’s what’s been the arc of change that you’ve seen in the industry over the last 12 months or so?

1:04
Wow. Big question. Yeah, who’d have ever thought we’d be here 12 months, but it’s an unprecedent unprecedented events. So it’s, it’s hard to say whether organization’s responses were right and appropriate. But I can certainly talk about what what we seen and what some of our, some of our customers seen. I think, in the early stages, particularly there was obviously the big push to there’s a lot of change, and change brings complexity brings brings challenges for organisations, in particular, that know the push to self serve the trying to get agents working from home in a hurry, and even some changes within within customer attitudes. So what we seen in we do, we do literally hundreds of 1000s of hours of conversations we analyse on a daily basis. So where we were able to see quite quickly what the what some of the changes were. So first thing we seen was there’s a lot more, there’s a lot more understanding, you know, we’re all in this together. There were there was no things like two way empathy, for example, you know, people would start the phone calls by saying, Oh, isn’t the situation terrible? No, are you working from home. So a lot more report, which we seen, maybe handle times were a bit longer in relation to that, but that we didn’t necessarily see a big downturn in sort of customer outcomes or KPIs that are much more handle time. There’s also a couple of technical challenges with with things like audio quality, and a lot of the self service channels were, some of them were put together in somewhat of a hurry. So we seen with the breakdown of inappropriate self service channels, a lot of customers arriving into the contact centre for agents assistants, in an already agitated state know that that sort of they were the sort of dynamics that we seen in general, I mean, as a company, we had let you know, we got ahead of this a little bit. We know we were we established a Coronavirus, think tank. So we got all of our all of our customers together to see what they needed, how we could help them. And through that thing tank, you know, there’s a couple of themes that we that we helped with. One was, I think, to understand the impact of COVID on the cut on the customer base, and we are able to produce some sort of some new analytics categories to help them to hone in on the customer voice of the agent. And then what were the changes and challenges and how do you prioritise that those challenges? And then finally, I guess, you know, how do you how do you how do you coach and performance manage agents in the environment? So, you know, that that that was that particular exercise when I won an award from Stevie Award, which is a speech analytics Council award. So,

4:23
attempted industries, I mean, just construct one industry in particular there that we do a lot of business where the insurance industry highly regulated. And it was really important to them that the pandemic was not seen as a reason not seen as a as an excuse for for maybe prompting customers to health insurance or life insurance. Now, sir, so helping them helping to ensure that they are health and life insurance customers, the agents were doing the right thing. And then on the the other side of that, who takes a car insurance, you know, the impact of the pandemic, there has been claims or claims are radically down in, in our insurance and some of the some of the more forward thinking insurers that responded to that with things like you know, mileage changes, etc, to dynamically change, you’ll change your premium, I think the impact of that this year or through the last 12 months, would be probably a downward trend on on the amount of claims. So So from insurance, car insurance, a car insurer, perhaps an improvement in profitability, because haven’t got the cost of clothes associated with claims this year. But then on the inverse side, next year, we’re going to see that downward pressure on premiums. So I think it will, how this all plays out, ultimately, I think will will, will depend a lot around some of the uncertainty on the on the economy. I guess it’s a long answer, Chris, to a short question. I guess my final point on this would be I think the real heroes of all of this, you know, besides obviously, they’re they’re the the the key workers and the health service, in terms of customer communication and contact channels, that the real heroes have been the agents, you know, thrust into working from home, isolated from colleagues, and still trying to provide the best possible service to the consumer. So I would say hats off to the agents be quite low point.

6:41
And I suppose in terms of some of the so in terms of like the call recording and the analytics, then you’re seeing use, so you saw greater empathy coming through? And then what do you see more vulnerability coming through into the light people starting to think about vulnerability or mentioning that on the calls? I mean, if you sort of track some of the data that comes through?

6:58
Oh, yeah, I mean, so I think it’s a huge topic. I mean, there’s been a lot of a lot of publications around vulnerability from the FCO, SCA, the ONS and various people, I’ve got a couple of couple of stats that I There are pre prepared on vulnerability, because I was hoping you can ask me about this. But so to put it into context, about the the, the environment that organisations are having to cope with, and vulnerability is is shifting dramatically. Now the NCAA reported, I think, last week, 27 point 7 million adults in the UK have a sub characteristics of vulnerability and more wealth, low financial resilience, or more negative recent negative life events. Employment to the UK unemployment figures came out today. And that’s likely to reach no 2.6 million by the middle of 2020. On the still 4 million people that are temporarily away from work that may or may not be going back to their jobs. And 9 million people had to borrow more money. This was I think, in December 2020, the ONS producers that that said, 9 million people had had to borrow more money because of the Coronavirus pandemic. And 10% of UK adults are effectively digitally vulnerable. And we’ll talk a little bit perhaps about digital vulnerability into that. So there’s a huge backdrop to heightened vulnerability, and how organisations both both both measure and respond to that vulnerability, not just because they’re regulated, to do so it’s because it’s the right thing to do. Yeah. And I think that many of many of our customers yes, that they’d regulated, yes, they have to do they have to respond to vulnerability appropriately, they have to treat customers fairly, they have to do what the regulator lays down to them. But but socially, and ethically, it’s the right thing to do. So

9:14
I think it’s quite interesting. You just You just mentioned it around, you actually saw that in some of the some of your data. So if you look across your customers, you saw that some of your data in terms of the greater report empathy coming through as a result of it. So I think we have responded quite well to that mean, people are dealing with people on a normal human basis, aren’t they?

9:31
They are. But but but you know what, one of the things, Chris, that we see, you know, at scale is vulnerable. Customers don’t always offer up the fact that they’re vulnerable, and the skills and the training of the agents to be to be asking the right question to be doing the signposting to actually truly understand the nature of someone who is vulnerable, but perhaps doesn’t want to just say sort of straight out right, that vulnerable. So it’s hugely important. And I think the other side of that vulnerability equation is, if you take it as fact, and I think it is fact that more and more people are vulnerable, and therefore agents are presented with more and more challenging calls, customers in challenging circumstances, how you support the employee, to be able to not just deal appropriately, that but to counsel and guide and coach those agents and give them a break, when you know that they’ve had a, you know, a particularly heavy workload of particularly demanding calls from vulnerable customers, I think there’s two, there’s two real sides to that, to that vulnerability. There’s the there’s the how do you measure, capture and responsive unlimite to the vulnerable consumer? And then how do you support the agents who are dealing with evermore vulnerable consumers? Well,

11:04
and obviously, speech analytics can play a big part of that, which is what you guys do. I mean, so how do you sort of see that sort of fitting in with I suppose, with digital journeys, and particularly quite interesting things like how this works when you got a like, an agent at home, as well as we’ve got a lot of distributor call centres now. Whereas it used to be sort of onside and you’d have that sort of like support, you’d have floor walkers, you’d have people sort of on site. Now it’s much more district distributed, you know, how does it How does that link in with things like, you know, some of the analytics and some of the data that you can extract from the calls? What’s, what’s been the, the experience there and how it all fits together?

11:38
Yep, so so so let’s, let’s talk about the these these digital journeys. First of all, if you don’t mind Cresa, that so 10% of the UK, adults are digitally vulnerable. And that came from the m&s in November. And when we say digital vulnerable, digitally vulnerable, that they, they net, they do not have the skills, perhaps, or the means to access digital channels. And it’s really important that when you’re designing customer journeys, you take account of this, this this digital vulnerable, the people who need to speak to people that will always be there, the automation of the mundane, the transactional, and recognise that most people even if they’re not digitally vulnerable, when they arrive with the contact centre, it’s generally going to be with with a complex query would be with the one thing I would say, in terms of analytics, yes, you know, we, we, we’ve got a whole host of content that we provide our customers to allow them to pick up on the signs of financial vulnerability, or health vulnerability, whether that be physical or mental. And also, digital vulnerability, it’s really, it’s really surprising, you know, but perhaps not surprising that, that when people arrive in a contact centre, having had a broken digital journey, they’re very quick to say, within the first 30 seconds, serve a call, Hey, you know, I was trying to change my tariff on the website, and it didn’t give me the tariff I wanted or, you know, your IVR threw me out, or whatever it may be. When you’re using analytics, when you can capture that sort of thing at scale. You can, you can see it happening consistently and persistently, you can use that insight to go back and fix the problem, of course, digital channel that led to the breakout into an agent assisted channel. So that’s, that’s the way analytics plays on the journey. And where analytics plays on on vulnerability, it helps to capture at scale, where, you know, what, how many customers are experiencing vulnerability? And perhaps where they’re at? Did the agent respond appropriately to that vulnerability?

14:13
And when you’ve got agents at home? Or like that it’s like a distributed call centre, you got agents at home? Is it is it really sort of seamless in terms of not really mattering, that they’re at home? Or does actually help the fact that you’ve got additional analytics at home to then even pick up on that? Because because it’s tough working from home, right? And people are in tough situations, right? And they’re, you know, they might be working from home, they got family at home, you know, those kind of things. It’s, it’s not the same as environment as being in the office and, you know, can it can it help with that, in terms of identifying, maybe even, you know, challenges that employees have as well? Oh, absolutely.

14:47
So if you think about about the way that the organization’s have traditionally done, done, quality assurance that they’ve taken, you know, a small percent, you know, five 1015 calls per agent. Now they’ve called been listened to, they’ve been assessed, and then they have this sort of coaching and guidance session with with the agent. Now. Now, when you’ve got your agents working from home, you know, you can’t have that one to one, conversation. Analytics has a huge part to play in this that we, we, for our customers, we provide holistic, and analytics. So we analyse all of their calls. Yeah, so so that would allow you to have a very targeted and focused conversation, not just around the one or two calls that we listened to, but around, you know, a consistent holistic and objective feedback. But then also put in the tools to allow that coaching to take place, you know, we have a module in our solution called coach, which which allows agents to see their performance trends against KPIs that allow them to self adjust, or let the supervisors use a data driven coaching session. So where you can’t be with the agent in person, you can provide very targeted and focused guidance and coaching, that’s data driven, is is how we approach

16:18
that. I always like the most like that the quality assurance always like that, that the coach or the side by side coaching, we’re both looking at the same information, you sort of like using that then drive improvements as much as you know, like the the top down kind of approach, which has been advocate of speech analytics for that. Yeah,

16:37
I mean, speech analytics, as a coaching tool. I mean, you know, when it avoids all of the, the, oh, you’ve just picked my worst calls this week? Or, you know, Why do you treat Chris better than me? I had a bad day on Monday, Monday’s calls, when you take a look at this, and we’re gonna have a conversation, Chris, about all of your calls, you’re really empathetic, you take ownership of your customers, but But you speak a bit quickly, sometimes, and there’s a there’s a lot of, you know, sorry, I didn’t quite catch that language from the consumer, just slow down, you know, that that is a much more valuable coaching session than saying, you know, based on your five calls, you scored nine out of 10 in each of the following four categories.

17:24
So well, and you’re rolling that up into like scores like, like speed scores of vulnerability scores, those kind of things and use kind of like models for that, to come out with summary scores.

17:35
Yeah, so so we we we score, the way that we work is the language is, is you can tag the language, key words, key phrases. You know, there’s a, you categorise, we have a concept of categorization, which is how you’d find some categories, something that that is that you’re interested in, within a conversation, you can have as many categories as you want, you can have scores as you want. And the really, really interesting thing about numbers when I when you when you score something, it’s now if I scored 72, on empathy, it’s a number. So you know, my target is 80. And I have a coaching session, so I need to be more empathetic, I can see immediately whether or not my 72 has gone to 75 or 80, or 90, as a result, so the supervisor can see immediately the impact of the coaching session. Yeah, so can the agent so. So I think one of the big thing was analytics can be used for many things, you know, provide actionable insights, improve business performance, one of the key things that our customers taken advantage of, during the pandemic, is to use analytics to have data driven coaching, conversation, and

19:02
real time feedback. So that real time feedback that that you kind of get,

19:07
yeah, yeah, it’s that it’s that real time or near real time feedback that you simply didn’t, you didn’t get pre the pandemic, but you can now get it, no, post the pandemic with, with speech analytics tools.

19:19
And to create the scores, you obviously then need to extract the data is your level, the various attributes around the conversations actually actually happened, the detail around the conversations happening, that’s all within within within the system. And but these they often then get put into models that then could get sold for something. So empathy, or whatever it is. I mean, what I mean, use machine learning, obviously, to extract some of that, and where does it Where does all that fit into the to the models? And I suppose one of the questions I was going to get to is really around explainability as well. So why was my empathy score? 88 versus 80 or 90, and, you know, how do I know that that for me? I mean, how do you sort of have I just sort of explained some of that and solve for some of that.

20:02
So, so we’ve built our models based on based on hundreds of 1000s, if not million hours. So our model is a contact centre model. So now we haven’t just gone to, to just conversational language models, we’ve built all of our all of our machine learning around the language of the contact centre. So that in the engine, if you like, of this pitch analytic solution, you have a lot of, there’s a lot of hype out there about AI and machine learning, I think we were doing, we were doing machine learning almost before the term was was invented. So that in the back end, we know, we know. So for example, take a take take a concept like this satisfaction. There, there’s probably over 200 ways that a UK consumer will express that they’re dissatisfied, anything from saying, you know, I’m not happy about this profanities, or I want to speak to your supervisor, you know, so being able to capture all of that language at scale and build that into the categories. No, you know directionally, that when, when you’re measuring dissatisfaction, the customer is dissatisfied. Now, you can then over No, so, let’s say a dissatisfaction score of 70. That’s as a dissatisfaction score of 90, you can overlay within the within the dissatisfaction, you know, all the categories, like escalation. So, I’m not very happy, might score 50, I’m not very happy, I want to speak to your supervisor may score 70, I’m not very happy. And leaving, I want to leave, you may score 80. So you can you can adjust the scoring to the outcomes that you and the relevance and prevalence that you need. But the whole thing is, is backed by, by deep machine learning an AI engine from millions of hours of conversation,

22:11
and how does it change by by market? Right? So, so an expression of having worked in several different countries, right? I mean, it is quite culturally sensitive some of this stuff as well. I mean, how do you think about around sort of building it by markets or doing special dissatisfaction in the UK might be very different than say it is in Canada or in the US, or those kind of things? Just and even within the USA, you might be it might be different, even within the UK might be slightly different? Do you see those subtle differences? Or is it just zip perception?

22:40
You absolutely do. And, and the subtle differences don’t don’t just start don’t don’t? The total difference actually start with the language models themselves. So, so UK, English, is not the same as Australian English? Yeah, it’s not the same as of African English, the the lexicon of words and phrases that form a, a language pack for South Africa. So one of the things we do, I probably should have said this a bit earlier. Now, one of the things we do is for audio is we transcribe every word of every call. And we use a we use a speech to text language pack to do that, which has got a huge lexicon of words with it. But the lexicon for Australia, because they know what the Australian say, are different words, they use different words, and how they then express themselves. There may be I don’t know if this is fact or not, but I’ll use it as a as an example. There may be 200 ways that we express dissatisfaction within within the UK. Yeah, there may be 130, within Australia, and of that 130 30 that may be unique to the way Australians express themselves. So we absolutely see both both a a cultural and language difference from geography to geography. And, you know, we build our models in accordance with them differences.

24:11
Yeah, I think it’s quite, it’s quite interesting, like, how it changes and how people respond will change and even how they express their dissatisfaction will kind of be different as well. I mean, so what might seem like distracts dissatisfaction in one country might might actually not be in another country as well. I mean, you sort of is really quite different, really, I suppose. And so where do you think where do you think we go from here? And obviously, we’ve got this massive data that’s been collected like contact data has been collected, including speech data or which has been analysed, we got models that are now coming out to be able to determine theorist fact various factors around what’s going on with customers, but where do where do we go next? From here? What what’s what’s next on the horizon? Do you think

24:50
so? So for us, you know, I think it’s a it’s about it’s about extending the ecosystem. Now this data that we don’t We produce can be used in, in many different ways and can coexist. So, so things like, you know, I’ve already sort of hinted about about going back using the data to go back and correct sorts of self service journeys, for example, yeah, but the truly integrating, say, the robotic process automation, or the Chatbot, or the, or the NPS survey, infrastructure, with the data from the conversations, using API’s, so allow, allow that, that, you know, cut the cut the secondary and tertiary analysis, from the data and allow all of these systems to coexist and be interconnected. And it sounds like, wow, you know, that’s like, sounds like Star Wars. I certainly think that the, the, the advent of, of open API’s has made that a possibility, you know, we, we built our platform, from the start with with API’s in mind. And, you know, we, we integrate an interface to many, many different third party systems, whether it be CRM systems, or, or CX performance systems, or IVR, or voice biometrics, or fraud management systems. So I think really, it’s about it’s about making more structured use of that data to all of the other systems that former customers

26:39
visit, I suppose it’s a contacts or contacts no matter where it comes from, or transactions or transaction, where it comes from. And it’s almost like, you know, how you put that into the same engine to then solve for particular outcomes? I’m gonna suppose that and it’s like, and then can you use that technology to then find those outcomes and determine what the outcomes are? Right.

26:59
So let me give you example of that, then Chris. Yeah. Because I think that’s you touched on a good point. So let’s take a customer experience survey. Let’s take an NPS survey. Yeah. So so first of all, you know, you might get an NPS score, you might be a promoter, you might be an attractor, you may leave some verbatim comments, you know, maybe allowed to 105 140 words, or leave, leave a minutes worth of comments on the voice survey. The true nature though, of what really drove that outcome is buried within the conversation itself, loft and in the conversation that led to that promoter or detractor, because you’ve only got, you know, a 10 15% survey take up rate of the other 85% of customers that didn’t get surveyed. How many of them conversations have the same characteristics, and then that allows you to to predict detraction, or predict promotions. That’s a good example of using the data on the survey, and the data from the conversation, combine them together for for now to drive better, better customer experience and better business outcomes.

28:22
And let’s face it, it solves some as almost like respond to bias the fact that people respond are going to have a certain a certain they respond to because they want to give an opinion, but it’s the other 85% who didn’t respond, and then how does that answer? What’s your overall picture? Probably gives you a more accurate picture than just looking at the responses.

28:38
Correct? Yeah. And it’s but you know, it was a it’s a great question, because I think I think that that’s a really good example of, of how you’d let the, the survey system, if you like, the CX system, and the speech analytics system coexist and complementing each other,

28:55
and how much more how much more powerful is speech versus text? I mean, so because you get, you get, obviously, if you’ve got text being combined, you’ve always got the speech, what’s actually being said, let’s just say, but then you’ve also got all the subtle pieces that are happening in in, in conversations as well. And it’s just interesting as we move to much more of a digital kind of outcome, especially for some of the more straightforward kind of transactions. Are we losing information? We’re losing that subtle information a little bit around how customers actually feel? Well?

29:27
I don’t know because I’m not sort of a mystic scientist or whatever the scientists offer for tax. What I will say is, it’s absolutely definitive. Customers expressed themselves differently over different channels. How how I would express going back to my my, my dissatisfaction example, before I speak to an agent to say I’m dissatisfied. Is they different how to do it on web chat? Yeah, I might use lol or, or sad faced emojis or, or things like that. So you can’t analyse you can’t just say, take the language of audio, and put it into a web chat world and expect it to work because it doesn’t, you have to analyse with a different mindset and a different context within the within the digital world than you do within the voice world.

30:26
As you see that even in social media as an example, right, the way people respond on social media could be could be really quite tough sometimes, in terms of being quite critical, versus the way you do that on text versus the way you do that, if you’re talking to someone face to face, where you sort of a lot more sort of, it’s much more it’s much more personal and much more variation. You wouldn’t people don’t don’t say those things when you actually speak to people face to face. And I just wondering if you, you know, we decide channels based on preferences or on habits? I mean, what should we have? Should we really be thinking about channels based on what the the right outcome is, and what what the kind of dynamics we want to have, I’m not sure if that really comes into it as much as this what the customer wants,

31:06
I think that it did the with, there’s probably no papers and papers and papers written on this topic, but the way I look at it is people need human assistance. When things are complex, everybody leads busy lives. And aside of the digitally vulnerable, which we’ve already spoken about, you know, what we want is the convenience to do the mundane stuff simply easily and effectively. Through self service channels. When we break out of those self service channels, we want we want the agents to be knowledgeable to be helpful to be skillful, and to be efficient in how my how my transactions handled and the did the customer journey scientists need to need to be thinking about about that that that efficient, effective skilled agents nullable aid and there’s lots of technologies there. Okay, so this is a no a call that in the IVR or in the self service channel, I’ve said what I want to talk about it’s a it’s an insurance renewal it’s a multi car insurance renewal this needs to go to Chris was Chris’s Chris’s skilled about dealing with this, he’s got the skills, there are all sorts of skills based routing getting goes to the next level as well.

32:28
So it sounds like it sounds like the call centre is not dead yet. I mean, there’s there’s a bit of life and it’s around sort of specialising it. And, and and and really sort of leveraging the skills there for the right time.

32:39
Yeah, you know, one of the things that one of the things that we see that I know infuriates infuriates consumers and me as a consumer, not as a speech analytics guy, that as a consumer, when when I make a call into into contact centre, and I’m put on hold, or transferred, or there’s a lot of hesitancy, and we one of the things we measure now all of the wonderful things we measure, one of the things we measure is silence. Yeah, silence is dead air time, that’s doing me no good as a consumer. And my customers no good because they’re paying for time that they’re getting no value from. And it’s amazing that when you look at, you know, why does silence happen. And it’s generally because of two things. The person dealing with the call hasn’t got the skills or knowledge to deal with that call effectively, therefore, they’re hesitant, put me on hold, they transfer, or they haven’t got ready access to the systems that they need the data that they need, in order to pull up in order to deal with my call. And there’s delays and you know, pulling up my CRM record or finding my letter of complaints or whatever. So So now one of the things we help our customers where to avoid that friction in the experience, what are the drivers of silence? What what’s driving silence? And how do you know that that efficient and effective interaction?

34:06
Well, it’s fascinating just to talk a bit about some of the things that you’ve been seeing, I suppose from a speech analytics point of view, it sounds like it’s you say, it sounds like there’s still a lot to do. And then the data is only ever expanding as well. That’s, that’s what’s also interesting terms of diagnosing some of the things are going on. And I do think it’s been sort of an interesting inflection point, I think now having been sort of locked down for a year, in terms of like, where we go from here, I think, really, but it’s definitely seems like the last year has been a bit of a proof point in terms of using more data using more, you know, technology to really analyse how customers are feeling right. And I think it’s probably only going to be more important going forward.

34:46
Yeah, it’s hard to say Well, no. Well, I think we’ll see you know, certainly lots of my customers that they’re not going to go to necessarily to a full blown everyone back into the contact centre wants to pandemics over there. will be this mix of, you know, sort of hybrid work from home or flexible working patterns which which I applaud and I think is is the right thing to do. But therefore, the role of technology, some of the examples we’ve discussed, I think, become now become more important and more sustained to make that. Okay, success.

35:22
Frank, thank you very much. I really appreciate you taking the time today. And it’s been great to chat to you and we’ll speak soon.

35:28
Thank you, Chris.