How AI helps us in audience intelligence
What does audience intelligence mean for the insights drawn from it?
In this episode, Paul and James discuss understanding sentiment accurately and how it can unlock insight at scale. When you can understand how people feel, it helps make better decisions for everything from marketing to government.
This episode of Know your Audience also discusses customer experience, broad trends and how happiness can be a window into a world.
The guys touch on the concept of AI and natural language processing to better understand an audience – bringing it all together to understand how you can unlock insight to make better audience decisions.
Know your audience is a production of D/A and Dr James Piecowye.
AUDIENCE FIRST MARKETING PODCAST TRANSCRIPT
Dr James Piecowye Welcome back to know your audience with myself, James Piecowye.
Paul Kelly And I’m Paul Kelly
James And Paul, we’ve been having quite an interesting conversation here. And the thing that I want to jump back to is a question that I’m sure is in our listeners’ minds as well, is the whole idea of audience surveillance, audience intelligence, and sentiment.
Paul Yeah, I guess like, if you take a step back,like we’ve been talking about, I guess, through this whole series of know, your audience, and everything like that is thinking about consumers of your product or service, whether that’s another business or whether they’re out in the world, as an audience.
When you start thinking like that, you become more aware of thinking about things from a perspective of who is consuming your product and your message. So it’s like an audience for this podcast, you know, it’s the people sitting in wherever walking or whatever they’re doing, we think about what they might be interested in, and we tailor the content to perhaps give them something to educate.
So when you understand that audience, which is what audience intelligence is, it’s getting an insight into that audience. It’s not just data, it’s not just numbers, it’s insight, which is actionable measures of something that will make a difference to their life or make a difference to your business. That is the key of what audience intelligence is.
James A lot of people are already doing this in their own businesses. (Asking) how many people are calling today? What are they concerned about? What’s going on? What’s the customer feedback?, all that sort of thing. Surveillance?
Paul It’s intelligence. And the difference is, I think surveillance has a nefarious connotation, but not with language, because it sort of implies that things are being watched. And that’s not really the point of intelligence. Intelligence is about understanding. And I know that also has a nefarious thing, if we think about CIA or something,
We’re understanding and we’re getting information about their behaviours, and potential behaviours and or trends, issues that might affect them from multiple data sources. So the thing you talked about about customers? Not retention? What’s it called the surveying, you know, customer feedback? That type of thing is a data point. And what we’re talking about here is multiple data points across your own first party information, which is what you’re talking about, right? Which is where you ask them for feedback, your sales data, your supply chain data, like all those various points coming into one.
James So that you can take the data and do some form of analysis.
Paul Yeah, it’s okay across all of those and creating them, using technology in our case, to sort of bring a better understanding of what all these different data points together might mean and give you information on. But then it’s also really interesting to go beyond that into the third party realm where you’re sort of looking at information about you know, from the social web, or from the internet, or from new sources and things like that about broader trends, broader societal shifts, and then how you then can break that down to people’s opinions and thoughts,
James Which is where sentiment comes in?
Paul Yeah, because what we want to do is we want to see how people feel about something without necessarily asking them because, you know, this is great, I think line of psychological thought about how we behave, which we’ve talked about extensively about the system one, system two thinking, and how we’re almost always all on autopilot, right, except when someone asked us a question.
James This is, the sentiment side of things, it’s something that we’ve we’ve always been doing, we do it as human beings, we’re sitting in front of each other, we’re having a conversation, you’re talking about this, I’m looking at your face, I’m judging … Oh, he’s smiling in a grimace, you know, how does he feel about it? He slowed down. So we’re very adept at making an analysis of the sentiment of people that we interact with.
In this case, though, we’re taking audience intelligence to another level, because it’s not just sort of data points that are looking at, (things like) Do you like something? Do you not like something? What people have said… you’re now going in and examining social posts. And getting a sense of, for example, what people think of things.
Paul Yeah. And this is where it becomes a great use, because, like you said, that there’s,
I guess verbal and nonverbal cues, right? And in our behaviours that we see with people we interact with, and sometimes those cues even in our situation, I’m not necessarily showing you my true self. Whether that’s conscious or unconscious, I’m not doing that, because we’re in a social environment, there are social norms that dictate our interaction with each other. And that’s, that’s the multiplier.
So yeah, you might be a business and have an irate customer, of course, like, but that could be one of 1000. But it focuses all your attention on that.
And you begin to sort of understand maybe that person’s frustrations with your product or service. But when you’re talking about what people think about broader trends, like,, if I’m thinking about the economy or people’s feeling of job security, at a really macroscale, which is something that we do with a consumer sentiment index, it’s about understanding how people feel towards that, and at a very large scale. But you can also filter that down to a category, for example, so are people turning off some kind of cooking oil, because of you know, like palm oil and health related things? How do they actually feel about the different types of fats and things like that, because that can really give an insight to a business and what a future trend might be.
And if you’re positive, we simplify it to positive or negative, but you could also think of happy or unhappy when I see this all the time, happy, neutral, unhappy. And particularly where we are in the United Arab Emirates, there’s a big push on happiness, but that is positive sentiment, right? And the happier you are towards something, then the happier you generally are.
And what we can do with technology is understand this at scale through what’s called natural language processing.
James Explain that; natural language processing, explain how we’re now taking audience intelligence, we’re putting sentiment into the mix as well, and now …
Paul We arrive at sentiment through a process called natural language processing. And what that does is, if it’s able to analyse, let’s use social data. And let’s let’s drill that right down to say something like Twitter, because it’s text based, largely. So let’s, if we focus in on that, if you could imagine a million tweets on a topic
for example 100 million tweets on the economy, from all the GCC countries. Gulf Cooperation Council for those outside, so the country’s, the Arabian Gulf? If you there’s no way possible that you could read all of them. Right? Well, they were I mean, doubting that some savant somewhere can sit down and go through those spreadsheets. Yeah, but it would take a long time, they wouldn’t take minutes or hours, you know.
And what technology enables us to do, and what AI at its core is, as we’ve discussed before, is the ability to process information at a fairly fast speed, at a fairly high degree of accuracy.
And at and at a level, which no human could possibly do on their own, or maybe a team of 100 could do in a few weeks.
Then what part of that technology is called natural language processing is, what it is, is it looks at the context of the words that are used.
James So that’s the key word right there ‘context’. Yes. Because as as we’ve been talking about ourselves in this conversation,
when you take the natural word context, and you take it, let’s use the GCC as an example, or even a little bit wider. And we’re talking, we could talk in Arabic language, which, as you’ve said, over and over and over again, we run into this huge issue of a perception of homogeneity. (it’s hard to get that word out), a perception of homogeneity with this region because of language yet, that while the language is Arabic, the way the Arabic language is being used,in different countries is different.
And this creates a really interesting challenge, because you’ve got the same words with slightly different intonations with slightly different spellings. But that might be used in different contexts, to say similar things. Which, if you’re listening, and you don’t understand that, you could be thrown off your research.
Paul Yeah, exactly. Yeah, there’s a huge number of points there if we have time to unpack them.
The first step, I guess in what we could do is in a future episode, we’ll jump into what a technical layman’s terms technicality of natural language processing, but in a really simple way, it’s the ability to understand what someone’s saying within the context of a broader conversation. And what it’s able to then do through various models that are created within this field which is within machine learning, okay, is for that machine or that algorithm to understand what is meant by something. So it could be a phrase like; “I’m not happy about this”, you know, which is very clear, right? I’m not happy.
There’s a lot of things like sarcasm, (that can confuse what would be a simple statement)
James Yeah, “don’t use this” all related to the same idea that I’m very unhappy. But they’re iterations in the language.
Paul So they’re very clear statements of intent, but then you have these statements that are very unclear where you’re introducing like emojis, (like a crying emoji can actually mean ‘very happy’ as an example), or, like, you know, “pleasant surprise” or that type of thing. And there’s also ways that people talk, as well. So with sarcasm, for example, there’s lots of examples in the world where sarcasm (saying the opposite of what you mean by how you’re saying it). And in the online world that sometimes most often doesn’t come across well, but you can also pick that up if you’ve got the technology that’s smart enough, because you understand the context.
Usually that’s in a reply to something or somebody sharing a piece of news with a sarcastic comment. And it’s clear to a reader that it’s sarcasm, but a machine unless it’s built smart enough, can’t understand that. Now getting to the I guess your bigger point there is about language. And this is what happens unless you’re aware of it is that in a lot of parts of the world, not just in the Middle East, particularly in the Middle East, because Arabic is an underserved language, which is a bigger issue than this, but
James And it’s an incredible number when we start looking at, you know, 400 million speakers in the region. I mean, this is this a huge number of people, and I think that are underserved
Paul Yeah, and that’s our focus. So technology often follows commerciality. So the English speaking world is larger, for sure. It probably now is the global language, it’s probably over taken French I would suggest.
And that a pure amount of people who speak English, the amount of people who speak Spanish and also the geographic spread, for example, Spanish, and then the sheer number of people who speak, for instance, Mandarin, okay. And languages like that tend to shift focus away into those areas. So even though it’s still a huge number, the commerciality sits within probably wealthier, you know, because if you look even at the Middle East, right, you’ve got a probably a cluster of countries that are extraordinarily wealthy, and then you’ve got a lot of other countries that are not. And so the commerciality then follows, probably where. And so what happens then is that there is a modernist view of this region in the sense that Arabic is a language, which is, but Standard Arabic, and he tells you part of the story, there’s over you know, well, there’s 2022 main dialects across the region. But then even when you go into somewhere like Saudi Arabia, you’ve got several dialects within that country. So there’s one on the east talks about different someone on the way like
James I had a conversation with my students yesterday, exactly about this. And students in Dubai talking about their cousins in Abu Dhabi. They were talking about the same thing using totally different words. And they’re not quite sure what they’re saying. And that’s 150 kilometres apart.
Paul Yeah. And it becomes, I guess, the history of, how the raging came to be, and tribalism and all that sort of thing. And the interesting thing about Arabic, which I think probably happened in, in ages before, in, say, Europe, (with Latin and things like) that is that there is a language of religion, which overarch has everything, which is a very formal form of Arabic, which then becomes a unifying language, which is then come into some sort of formality. It’s not necessarily as formal as say, The Holy Quran, but it’s a bit more, it’s less formal than that, but there is something called Standard Arabic, right?
So people can understand things and it’s largely and also popular culture starts to fade in and you get this massive melding of things. And like you just said then you’ve got people who particularly younger speaking in those dialects with slang, right and so what you get then is this slang, Emirati or slang Kuwaiti or whatever, if you don’t know about that, and the technology that you’re using to understand sentiment, how happy someone is towards whatever it is whether it’s a huge topic like world affairs, or it’s a tiny topic like how they spreading the butter on their toast.
James Yeah, important if you’re going to be putting your Marmite on the very edge.
Paul Vegemite like, come on. (Am I okay, sorry). Team Vegemite here.
But, getting back to that topic, you’re not necessarily going to… you’ll miss out. It’s not that it’s wrong. You just miss out on a lot of data. What that can do is shift the results.
James So you don’t have to , your data is not as rich.
Paul No. And that as a result and impact downstream of saying, for instance, if we’re looking at, maybe the spreading is a bit niche, but if we’re looking at, say, for instance, people’s preference of ice cream, and how happy they are towards, I don’t know, Nutella or ice cream or something, then you might get a data pool, a small data pool, right, or 1500 people or something like that. Because you’re using standard technology that relies on Standard Arabic. And then it relies, you know, without that enrichment of the culture around it, you might be potentially missing out on 2,3,4 or five times as much conversation happening,
James Because it doesn’t understand sentiment.
Paul Yeah, and it doesn’t understand the dialect.
So it can’t pick up the sentiment because the words don’t make any sense. To somebody reading it to understand that they do make sense. You know what I mean?
James So right, so if you were going back to old-school and doing interviews, you might be able to catch this stuff, if you understand the dialect.
Paul And then you have the problem with interviews, if somebody’s not telling you the truth, social desirability bias, where I think that you want to hear something from me. So it’s different from what I actually think, right?
(EG) “I use reusable shopping bags every time”… like, I don’t, a lot of the time. Because I forgot, I’m busy.
You know, like, it’s not a conscious choice, I understand about plastic and things like that.
But society would make me feel bad about saying that. So I would watch what I say if somebody asked me that in a formal situation, you know, like, instead of addressing the bigger topic of, actually, if it was more convenient for me not to somehow, then I would rather not use that single-use plastics, and all that sort of thing.
Sitting here, drinking a water bottle out of plastic, you know, it’s the convenience element for somebody, and then, you know, a broader trend might have me answer that question a lot differently. Because I think that you want to hear me say something else.
And that’s the problem with interviews
What the great thing about social data is it gives us this data, this huge pool of data that which we can understand, if you understand behaviour on platforms that people are a bit more honest on Twitter about their beliefs, and political biases, and how they live their daily life, as opposed to somewhere like Instagram, where they’re perhaps projecting an image, but in the comments, they’re not right.
And once you understand all that, you’ve got this richness of data of how people behave without actually asking them. Something like sentiment, that means you can understand how happy or more probably more importantly, how unhappy they are towards something, whatever that thing is, whether that’s an economy or whether that’s a bottle of water, you can understand that and you potentially miss out on a lot of data by by that emergent, like, you know, that homogenous view and language.
James And so by having this very heterogeneous view of what’s going on, and having this really much greater data pool, I can now make better decisions, Yeah. And I can, I can see where the niches are.
Paul Exactly, and sentiment is the start of how people feel, are you happy or unhappy?
Not necessarily sad and just happy on the positive, you know, positive or negative? Then you go from there, right? Like some people are just neutral, there’s things shared without comment, it’s just factual recordings. So that stuff gets excluded. It’s just where somebody has an opinion on something, is that good or bad? That gives you a launching point then to start looking at perhaps more deeper emotions, you know, like, anxiety, anger, those types of actual human emotions, you know? And then when you can begin to understand those you can sort of, you know, if you’re a customer facing organisation, and all that people do is complain about you online, which is what happens with say, I think banks, for example,
James Everyone’s got an issue with the bank you.
Paul Yeah, I think that’s a global thing. I think, I think in different parts of the world, there’s probably bigger pockets of people who might be really happy with their experience. But, you know, money is one of those hygiene, you know, like, where it’s just like, not everybody’s happy.
James Don’t mess with my money.
Paul And there’s that feeling, right, of ownership, and I think things will change eventually, but you can enrich how you know, by looking at this, you can look at benchmarks for example. So, alright, we’ve got to change this, we’ve got to change how we behave towards our customer, where can we look so you look at adjacent industries, you know, what I mean? Where people are happy, and they’re happy to share their experience, right, like, you know, things like travel. You know, there are a lot of people who complain online, but there are a lot more people who have shared their positive experiences and travel or events or even shopping in different categories, you know, like so, footwear is often a positive experience, right? Even though it actually probably a lot of the time isn’t.
People still share that you know what, and you can learn from that to apply those learnings from to customer experience, for example. So you start to just from this one notion of happiness and unhappiness in sentiment, when it’s done accurately and unlocks this whole world of understanding about a customer or a service or whatever that is – i’m using consumers here, I guess it’s an endpoint, but it’s as applicable to business to business interactions and understanding what the pain points of other businesses are,by how their employees and things like that might be interacting with the world.
James Do you think people realise these tools are available?
Paul Yes, and no.
James The audience, the intelligence I think people get, because it’s happening a lot. But, the sentiment of being able to do intelligence on the sentiment of the audience, I wonder…. I just get the sense people don’t realise that’s possible?
Paul No, I think a broader population may not and I think people inside say marketing does. But there is without doubt, and 90% plus, (I’ve not spoken to any, and this is my own choice here because I obviously work in this area.) but people we speak to are very unhappy about sentiments (which are) very important to them.
James Right? Because everyone wants to know, what are people saying about us? How do they feel about it?
Paul Yeah, our competitors, our category? Sure. Adjacent categories? Sure. Because it unlocks the trends and unlocks trends. And then from trends, you understand what might be coming, but also what’s out there now.
James And so that’s the key point, right there sentiments unlock the trends. So you can see what’s happened, what’s going on now. And project what’s going to come?
Paul Yeah, and depending on your industry, it might be way more interesting to look at the unhappy people, even though there might be a smaller proportion of unhappy people. That’s where you get an indicator of the future.
behaviour. And just to finish, I guess that last point is that there is a general unhappiness with how that works in Arabic, because the current tools that are globally available, and there’s no criticism whatsoever, it’s our choice of specialism here is that you couldn’t possibly sit and invest in dialect understanding, if you’re also trying to service 67 other languages that perhaps have more economic value to you, if you know what I mean. So if you’re in Europe, you’ve got those 14 main languages there or probably haven’t more than 18, anyway, that is more important, you know. So you’ve got this massive, like you think of somebody like Switzerland, for example, you’ve got French and German speaking people in the one country without their own language, and then you’ve got, you know, exactly, so the focus shifts, right.
But what sentiment is, (used in the right way) is a leading indicator. So, in research, you often hear about the leading and trailing indicators of behaviour and patterns in the economy or whatever. Sentiment gives us a really good understanding, depending on what it is about what might happen in the future.
So a good example I can think of is employment. Okay, so how confident does somebody feel in their employment situation? Because it’s always future focused. You never think about the past, right? You’re thinking today I’m thinking about how secure my job is, right. And each day, obviously, I think about every day, but certain events might happen, like, you know, timely events, I guess at the moment, like conflicts and things like that.
James Famines, supply chain interruptions.
Paul Yeah, whatever. pandemics, etc. Yeah, it gets people thinking and when people, for instance, worry about their job, they stop spending on big ticket items. So you find a contraction in things like automobiles or luxuries, not luxury items, because they’re very, usually the people who can afford them.
James Continue to afford them.
Paul Yes, there’s no there’s sort of inelastic to I guess, worries. But the things like I guess that might be splurges like travel, for example, is a big one, you know, someone might downgrade a holiday option, or someone might not splash out, you know, decide to hold on to their car for a bit longer, just a bit longer. You know, maybe I don’t need a new TV right now, you know, those sorts of things, which, I guess, somebody’s able to not (buy) They didn’t need them today, right? It’s not a concern. It’s not like rice. So by understanding leading indicators, you’re able to better help yourself in some sort of prediction in the future of what might be happening.
James And sentiment plays a big role in helping to understand.
Paul It is the role. not just as a leading, yeah, and use that with trailing indicators, right? So you back that to your sales data, which might lag three to four months, because that’s how I’m feeling today. I’m not going to change my behavior potentially on something I’m going to buy tomorrow, but it may affect something that I’m just beginning to come into the market.
So some of these things like automobiles are great examples, because you don’t go to a car showroom and buy it that day, generally, there’s a decision making process that goes on for quite a long time. Even if you’re brand loyal. Yeah, there is a long time before you actually design because there’s so many options, right as well. So, you can’t just go into a car showroom and drive a car out, it doesn’t work like that. And there’s lots of items that are like that, you know, like furniture, etc.
And so that decision process gets affected. So instead of me going, I was thinking about buying a car, I was going to pop to the showroom on the weekend, I think I’ll just go to the beach instead. And I might just put that on ice until I can have a bit of certainty.
That will show in sales figures in three to six months, right. So you’re able to go, Hmm, might be an issue here, maybe I need to think about what I’m offering or my incentives or that type of thing to make sure the customers still keep coming. Or just to be aware that this might be a bump in the road for my plans, and that can trickle down through everything, because then you know, as people sort of, for instance, can track their spending more than they eat out less, for example. So if I’m a restaurant chain, then I’m probably looking at trying to think about okay, well, maybe after some of there’s going to be some issues here. So do you know what I mean? So it helps you with a leading indicator, and then your trailing indicators help.
James So if I’m, if I’m doing audience research, and I am not thinking about my audience sentiment, and looking at my leading and trailing indicators, I’m not really not doing effective research.
Paul No. You can always look at issues and I guess, pain points and trends outside of sentiment,
James But someone might convince you that that’s really important. But honestly, yeah, it seems like we’re missing. You’re missing a whole area.
Paul Yeah.. And I guess, it becomes like a data point that can really help decision making from a business point of view. But it can also begin to unlock things, really specific things like sentiment towards, for instance, (like I mentioned earlier), fats and oils, for example, could be if I was if I was a manufacturer of like vegetable oil, or whatever, sunflower oil, or whatever it is, what are people actually worried about now that I could potentially address that worry with some new product development right, over the next six months or a year? Like where do I focus my r&d? So it can actually unlock a bigger question.
So even as part of a bigger study, where you’re perhaps looking at the how people have used a certain product over amount of time, which isn’t related to sentiment, when you add in the sentiment layer, that’s a richness, which gives you an ability to, I’m not saying 100% forecast the future, but you’re like, Well, you know, boy,
James Right now you’ve got richer data to make those analyses and start thinking about decisions that you want to make in R&D, in sales, where you’re going to allocate resources, maybe you need to hire some more interns. Yeah, etc, etc.
Paul Yeah, and you say this stuff, and it just tracks perfectly.
We publish something called the Consumer Sentiment Index across the Gulf. And what that does, is it tracks consumer sentiment across the business, private business, government. (So the economy, not the government’s performance, per se, but the economy) and employment. And without fail employment, if that dips one month, perhaps one to two months later, you see a dip (resulting dip) in business, and then the government is last. And then what happens is the actual employment figures are a trailing indicator because people, the last thing they do is fire, you know, it’s not the very first thing to do, right? You cut other costs, and then you go, Okay, this isn’t getting better, I need to lay off.
But what we’re doing here is getting how the actual employees worry, not what the business person is saying, right? What I’m going to do, you know, like, I’m going to reduce my headcount this year. That person would probably have been worried about that job, their job, that means in the previous year, and then you’re able to better sort of mix everything together. So it’s interesting.
James When we’re talking about sentiment, we’re talking about knowing your audience. It all comes down to having the necessary tools, as we keep saying, to make informed decisions. Yes. And those tools exist.
Paul They do. They do. And it’s about if you use any of those tools, it’s about understanding and asking the right questions so that you’re getting the right information. And if you are, great, if you’re not, then there’s always something out there.
James If you’re not, you should be and maybe you need to be asking questions about the resources that you’re using exactly why you’re using those.
Paul Yeah. And whatever you do, I guess, don’t pay for something that you’re not getting value from. Or an accurate viewpoint.
James This is this has put a lot into focus. When we start thinking about sentiment, we start thinking about the audience. I’ve got a lot more questions. We’re going to get to those very soon. I’m James Piecowye.
Paul I’m Paul Kelly
James And this is ‘Know Your Audience’.