The IPA Making Sense Podcast: Sam Knowles

What is it about studying Classics that sets you up for success in the insight business? 

Sam Knowles, Chief Data Storyteller at Insight Agents, joins the IPA Making Sense Podcast to discuss insight, innovation, AI, decision-making and much more.

Transcript

SPEAKER_01 0:10

Hello everyone and welcome back to another edition of the IPA Making Sense podcast, where we'll be making sense of the complicated things in media, advertising, and marketing today, and also hopefully just making them a little bit more fun. I'm Simon Frazier.

SPEAKER_00 0:23

And I'm Molly Bruce. And today we're joined by the amazing Sam Knowles.

SPEAKER_01 0:28

Now, Sam is the founder and chief data storyteller at InsightsAgence, where he helps organizations turn complex data into stories that connect, persuade, and also drive change. He's the author of the award-winning Narrative by Numbers, How to Be Insightful, and also the fantastic uh most recent publication, Asking Smarter Questions. He's also the host of the Data Malarkey podcast, and he's one of the UK's leading voices on using data smarter. So, Sam, it's great to have you with us today. Thank you, Simon. Thank you, Molly. It's fantastic to be here. Well, let's get straight into it. I mean, one of the things, Sam, and particularly that's always inspired me by the work that you've done, you know, you mentioned a lot throughout the books that you've published and also the presentations that you've given. Uh I've seen a number of them over the years. And the thing that that that's always kind of stood out to me is that you and Rory both share that kind of classicist background, and you kind of have this in innate kind of ability to bring, you know, to make that relevant in in data storytelling. How do you how you know what's the connection? Like what is it, what is it about classics that you think that really drives uh that ability to tell stories with data?

SPEAKER_02 1:34

Well, I mean it's lovely to be spoken of in the same sentence as Rory. Um we were both taught him a little bit before me uh by Mary Beard um when she was teaching papers on the body and antiquity and other and other things, um, and had a little elfin Bob was a very different character. Um what is it about the classics? I mean, it was always said to be um an education for the mind uh and and uh and a way of kind of time traveling and mind traveling and getting into other cultures, which I think is helpful. And I think there should be some strong caveats about classical literature. I mean, like you know, like most of history and most of literature, it's exclusively written by free middle upper class men. Uh no real indication of just what slave culture might have been like. You know, classical Athens had 30,000 citizens and about 300,000 slaves on whose backs they floated. Um but I think more than that, particularly when it comes to narrative, Aristotle set down what we understand about the three-act story structure in his little book, The Poetics, in the fourth century, um, the thesis, the antithesis, and the synthesis, the drawing together. But I think um uh in terms of piecing together the culture of particularly ancient Athens, ancient Greece, and ancient Rome, we're left with very imperfect evidence. And we need to um we need to bring together lots of different sources of evidence to get a sense of what that culture was like. Also with the language as well. I mean, decoding the language is like, I mean, for me, Latin, Greek, and Sanskrit are like kind of Python R and and machine code. I was at school, I was both a classicist and also a we weren't called coders. We were called programmers in those days. They're solving puzzles. And with imperfect evidence, I think that really sets you up perfectly for a career and a job in insight and data-driven insight. Um there's that line from Sherlock Holmes, you know, it's all very well paying attention where the dogs bark, but it's when the dogs don't bark that you really need to pay attention. So I think there's something about bringing together imperfect evidence and getting a view of the world that really helps you. I mean, you know, they they did a pretty good job in setting at least the Western tradition in terms of culture and philosophy and art and architecture. That's that's probably helpful, bringing culture into commercial communication. Um, and I mean, you've you you said you've suffered several times with me talking. I mean, one of my, he's certainly not a hero, but one of my kind of uh kind of yardsticks for uh for insight is is Socrates, that kind of curmudgeonly old ancient Athenian philosopher, who is said to have said many times, whenever on a quest to discover something about the eternal nature of truth or beauty or courage, he's said to have said, All that I know is that I know nothing. And I think it's it's not a rare researcher or insight or analytics person who starts in that position, but I think it's a brilliant place to start. Lots of different strands coming together, I'd say.

SPEAKER_01 4:36

Absolutely. And and Nat, you know, you're kind of talking through, it kind of makes it makes much more sense. You know, that connection of understanding kind of the the uh like the the societies and the civilizations. And I guess uncovering I mean, I guess there's similarities between archite um archaeology and insight as well. It's that, you know, getting that fossil brush, digging down, and then uncovering, you know, you're you're you're you're kind of like brushing away and finding the truth uh underneath that, I guess.

SPEAKER_02 5:04

Aaron Ross Powell You make a brilliant point. I spoke to a uh a really interesting guy called Carl Ehrman, a Swedish political scientist recently, um, who's written a book called The Afterlife of Data, What Happens to Our Data When We Die. Um, uh a fascinating thought. Uh and he's talking about, you know, the fact he's in this book, he talks about the fact that we currently, with our particularly social and digital media content, have already created the largest single archive of data about human behavior ever bequeathed. In 20, 30, 40 years, there'll be more dead people's profiles on Facebook than there are alive people's. Who owns it? What are the GDPR consequences? All that kind of stuff. But he says so w when you say that to people, they say, yeah, but it's all garbage. It's just what I had for breakfast or Molly was doing last Thursday. How interesting is that? And he said, well, if you talk to an archaeologist, they say, yeah, it's all very well to have the speeches of of of this person and and but apparently he was telling me that Montaigne, the Italian philosopher, said, I don't want to know what Brutus said in the Senate. I wanted to know how irritated he was with his slaves or how his wife ticked him off. So when you talk to archaeologists, they say, when we find a rubbish bin, or when we find a sewer, we can find out what people were eating. And I'm not saying that all of the content on the big plat of the big social media platforms is the sewer, but it's the day-to-day kind of detritus of our lives, our sort of the digital dandruff that we drop. Um and I think that that it's very similar, so for archaeology, um for uh and and also for insight, because you know, you ask people questions, or even if you look at them unmediated and you look at what they're saying in social or or or i in the the digital remains they leave behind, that's much more naturalistic than um the post that they put on LinkedIn that's been perfectly crafted. So yeah, no, I I I think I think s m looking to make sense. I think I I think classics as a as a training, I mean it was fun and there's all sorts of there's all sorts of entertainment and humour and and learning, but I think as a as a as a way of finding out where the dogs don't bark, as would say, it's it's it's a brilliant discipline.

SPEAKER_01 7:18

Absolutely. It's interesting that you mentioned kind of going through rubbish, because there was um an obsessive Dylan fan, I can't remember what his name was, who who defined this uh thing he called it garbology, and he would go through Dylan's dustbins and and then try and sort of you know, he made some wild, I think probably chemically stimulated as well, the guy was. So he made some wild assumptions about about things, but also it's it's undercover uncovering those strange truths that ex exist, I guess, in like what we would dismiss as mediocrity. Um I mean another thing that you that that I that you've mentioned there in relation to that uh quote from Socrates, all that I know is that I know nothing. We'll come on a little bit later to this uh in relation to AI. But it's strange the thing that strikes me as kind of unusual, particularly with the rise of all of the AI systems, is that principle doesn't exist anywhere, from what I can understand. Because it always says, Oh, I know something, and I know and it and it delivers that with a complete confidence that I guess had Socrates been involved in the development of Claude, it may not have been called, but it would probably say, hmm, let me just actually I'm not the right person to answer this or something.

SPEAKER_02 8:23

That's a I mean that's a that's a really brilliant point. Um I mean, Socrates didn't approve of writing, yeah. So he definitely wouldn't have approved of the internet, let alone large language models, um, because he felt that it meant that we weren't learning things by heart. I told you he was a curmudgeonly old so-and-so. Um but I I I th it's a it's a really important point that you make. Um I I was talking recently to a professor of of the I think she's the ethics of AI at King's College London, a woman called Sylvie Delacroix. Um, and one of her objections to the responses generated by large language models and other AI tools is that they have no humility or uncertainty markers in them at all. And it's like, you know, i I mean, market researchers have known this for many times for a long, long time. If you ask people a question, they will give you an answer. Particularly if you ask children a question, they'll give you an answer because they expect that the act of asking a question demands an answer. Now, it is possible, and I have experimented with this, it is possible to get your LLM of choice to be a bit uncertain. But it feels to me right now that that's like adding sort of random chatter in to the responses rather than it being a fundamental principle that's that's built in. Sylvie Delacroix runs this uh an organ a body called the Center for Data Futures at K at KCL. Um and she's actively working with the big pr providers to see if they can actually give not only more useful but kind of more human-like um responses. Because the typical response is either like, well, it's like a pub bore, isn't it? Well, it's funny you should say that because. And then and then they get something wrong and they say, Oh, you're quite right. So I mean I'm not I'm not a di I'm I'm a I'm a I'm a big user of a number of tools for a number of things. I know we'll come on to that.

SPEAKER_01 10:10

Yeah.

SPEAKER_02 10:11

But in terms of in terms of generating insight, I think there's a long way to go.

SPEAKER_01 10:16

Oh, absolutely. Um and that brings me on to the the next question that I wanted to to uh to talk to you about now is obviously you've built your career around helping organizations make sense of complexity through data and and also particularly what's always resonated with me is the storytelling aspect. I mean, what are the common kind of uh I guess cultural or structural barriers that stop you know data translating into insight? Because as you said, we've got more data than ever before.

SPEAKER_02 10:41

Yeah, there's this there's this limiting self-belief that says if only we had more data, better dashboards or or smarter analytics, then we'd make better decisions. And I I I think there's a fundamental lack of curiosity and willingness and probably ability, untrained ability to ask the the right type of questions. But I think one of the fundamental problems, particularly for businesses that aren't digitally native, that that have been around for more than 15, 20 years, is the silos and the different lexicons that exist. So uh there's a there's a uh an analyst called Ian Whitaker who writes for City AM newspaper. Yeah, writes on on media, and he runs training and and so on on learning how to speak the language of the CFO. That's his that's what he does. Um and I think the CFO and the CMO often do not speak the same language. And it's interesting, some of the younger um, the kind of the new cohort of, I'm not sure they call themselves data storytellers, but but but people who are working maybe who they've been trained in social sciences and they're working at the intersection between um uh between insight and or between insight and planning and creative are beginning, I've noticed, to call themselves data translators, and it's a term I've I've seen in many, in many places, uh, and in AI too, um, being able to to bridge not those two worlds, there are multiple worlds. But I think there are silos. There are there are there are silos, different lexicons, different data tolerance. You know, the finance team and the procurement team are going to have a much higher degree of data tolerance, and the marketing team traditionally, not explicitly but traditionally, would would have had less. So there's been no single source of truth. I mean, I think everybody wants to take credit for success, and you know, success has many m parents and and and um failure is an orphan. Um nobody wants to accept blame. And I think that a lot of a lot of the data sources that that many organizations have have kind of been bolted on rather than uh rather than a kind of endemically in. Now I think when you look at at digital digitally native businesses or entirely digital businesses, that can be less of a problem. However, I think one of the challenges there is that there can be an acceptance of tools and methodologies just because they're new or just because they and for example, I mean I'm not I'm not a I'm not an expert in this area, but for example, um the digital attribute attribution models, it appears to be a kind of toss of the coin whether you go first click, last click, or or or uh middle click, or I mean I I'm not I'm not an expert in that area. I've written and thought about that area, I'm not an expert in it, but I think there can be the not the shock of the new, but the rush towards the new for the for the Well, we've ripped up the rule book, we've started in a completely new way. But the what they might call the legacy businesses, the businesses that have been around for since before 1998, let's say, to choose a random date. Um there have been kind of fiefdoms and areas of control and people not wanting to share data. Uh and there being no uh I mean, where does data where does where does the where does the responsibility for data live? Is it in the IT department? I mean clearly that we've moved on from there. Um but I think I I think silos and structure is uh has been a bad thing. I think the other thing was, um, and we saw it happen with uh prompt engineers a couple of years ago being sort of sought out and paid like rock stars and footballers. Um but you know, data scientists used to be hoved off and kept in the basement, um, and they were you wouldn't approach them. And um and so I think the more that that they can be involved in the kind of understanding the role that they have in terms of analyzing data in order to make business decisions. I think I think that there's been quite there have been quite a lot of, well, we've got this data, let's give it to the data team and make sure they make sense of it, and then they give it back. And they've done a beautiful job and created a fantastic Python script that does it elegantly, but has got nothing to do with the business outcomes. So the attachment of data analytics to actual impact within the business, I think is helping to change things.

SPEAKER_00 14:50

Aaron Powell Yeah, that kind of leads me on to something that I wanted to ask because your fellow classicist Rory Sutherland has described how data can sometimes be used as this is quote unquote ass covering disguised as rigor. Um so to effectively, I suppose, justify a decision that's been made um after the fact. So maybe you don't get sort of the negative consequences that can come associated with that. Um but how do you think you can sort of avoid falling into that sort of trap of using it to cover the decisions?

SPEAKER_02 15:24

I think that one of the things that those that have a facility with data can be very much guilty of is to sort of parade the scholarship to say, well, we've done this and we've done this and we've we've built this multivariate model. And me and some people get that, and other people really don't want to hear that. Um it will it will kind of put their hackles up, it'll it'll make them sweat, it'll give them maths anxiety, it'll it'll remind them of why they went in a different direction. Um so I think um the uh to croak, Rory, the the the arse covering disguised as rigor. Um we did this, we did this, we went down these different routes, we found these different things. That can be I mean uh I I understand why the scientifically minded, the analytically minded would do that, but it's not human, and it's not empathetic, and it's not respecting the data tolerance of the audience. Those are the three things I think are kind of characteristics of a of a brilliant data storyteller. Um the other thing is that I think too many leadership teams, boards, uh management teams work or have worked, maybe are doing so less now, but have worked with their Data Insight research analytics teams partners to prove hypotheses rather than test hypotheses without getting all Pauperian on you. Um I I think that, you know, find some data that shows that the tweenage yogurt market is ready for hedgehog and hazelnut flavour. Okay, we'll go and find that. Um and that and the and and so so we've developed a product that can do this and we want to establish a market for it, or or give us the data that can. Uh I I I mean, I think it's entirely unacceptable. I mean, I I think it's unacceptable because I I when uh after my classic spit and then working in comms agencies, I went back to school and and trained as a as a psychologist. Um and then I really did. That's when I both fell in love, was first terrified by and then fell in love with data. Um but my stats teachers would absolutely and they were great and they were funky and they used the Radio One playlist as the examples and so on. So they really anchored things in reality. Um But they would not countenance for one second going out to look for data. You know, build- I I remember building an experiment during during my PhD. I was looking at alcohol and memory, but that is another that's another story. Um But uh build designing an experiment that was I thought building on another experiment, and my quite irrascible um Greek supervisor saying, you can't be doing this. This is just not the way to do it. You've seen something and you're looking to go and prove it. You know, what is the hypothesis, not what is the thing that you want to prove. So I think I think that's bad. I I think the other thing is, and there there was a there's a lovely article written in the in the Harvard Business Review back in 2019 by a guy called Scott Berenato. Um and it's called I think it's called Data Science and the Art of Persuasion. And he talks about how you build a team, uh, the different skill sets you need to take to take the data that your business generates, that the market generates, that you can buy in all sorts of different data sets. Um and then how you get the workflow to actually properly inform business decision making. And he talks about, and this is you know familiar from robotics or from logistics, he talks about says that efforts fall short in the last mile when it comes time to explain the stuff to decision makers. And he and he and he's got some really compelling data in there, unsurprisingly, about how between 90 and 95 percent of any um uh of the time spent on analysing data to inform decision-making is spent on cleansing the data and running a new analysis. And uh and about between five and ten percent, if you're lucky, is spent on how are we gonna tell that story? What is the story that we're gonna tell? Not not not how are we going to prove the hypothesis rather than test it, but but what is the story that we're gonna tell? And so, I mean, the teams that I work with, individuals and particularly teams in finance and logistics and operations and sales and marketing, lots of different functions in lots of different sectors, uh I I I get them to do really simple things like, okay, you're producing a quarterly board report, uh, and we know, you know, you you know you've got the tent poles of when that's going to be. So a week before or a fortnight before you have to have all the data, just put that in your diary. And then, you know, a week before you need to have a working draft, and two, 48 hours before you need to have the final draft, and you need to rehearse it, and you need to say it out loud, and you need to perform it to yourself, to your phone, to your colleagues, to people who are naive to it. And they think, God, these are just simple, these are just sort of simple productivity hacks. But actually, if you instill that discipline and you don't fall short in the last mile, then you're much more likely to deliver something that addresses the key business question that's been given to you by the leadership team, the board, whoever it may be. Um, but you're also more likely to tell a more compelling story. Now, that story may not be what the CFO and her team want to hear, um, but much better that the way that you blend your narrative and your numbers, your stories, your statistics gets a response. It may well be under no circumstances darken my door with a request for more advertising funds this quarter. It may be that, but but you've made the case and you've shown not all of your workings out, but your relevant workings out.

SPEAKER_01 20:43

I I think that that there's um I can't remember where it's from, but there was a uh it it was like a sales thing or something I was I can't remember now, but it was they said, you know, the number every every time if you're selling something to somebody, the the objection will always be I don't have the money. But actually the fundamental is I don't trust you. I don't trust what you're telling me to actually want to invest in. Because yeah, there will be circumstances where where you will the genuinely that will be the objection, but most of the time people are thinking, I don't trust that the money I invest in you, because you've not taken the time to build that rapport or taken the time and I guess that storytelling element is trust building. It's saying, you know, here's a proven you know, track record, not necessarily just data empirical data, but actually building an idea that this person knows what they're talking about. I mean safe hands, actually, because you know, the biggest uh thing, I guess, in the in the economic space is risk minimization. Like they don't they don't want to go, this person's a loose canon. I you know, they you know, we may win. They're not think they're not thinking in that fat tail as Nadim Talib uh would say distribution mindset. They're thinking very much on the kind of the thin tail of it.

SPEAKER_00 21:53

Um that that sort of echoes, I think, one of your key messages from uh your Book, Narrative by Numbers, where you've argued that numbers aren't really what does the persuading and it's storytelling. Can you give us an example of a brand or organization that's sort of like radically transformed its impact by sort of really turning data into very compelling narratives?

SPEAKER_02 22:21

I'll give you I might give you two if we can. Perfect. One is a very well-known campaign, and one is one is a result of some work that I've been involved in, which a spoiler is um the difference between Agatha Christie and um Knives Out. Uh but we'll come but we'll come back to that between who done it and how catch them, but we'll come back to that. So the first is the is the uh British Heart Foundation and the work they did kind of serially, but starting, I think, maybe a dozen years ago, um, with Vinnie Jones with their staying alive um uh campaign. So it the BHF, British Heart Foundation had realized that um the proportion of the population that knew what to do, when, as Vinny says in the ad, some geezer collapses and has a heart attack in front of you, had fallen from a half to a third. Um this wasn't something that people went to evening classes to learn how to do. Defibrillators had been put around the country and people thought that was enough. People didn't know what to do. Um and uh so and they didn't have a big media budget, but they wanted something that was going to have a viral nature. They would do some cinema, they'd do some, they'd clearly have they clearly run the uh the ads online. And so they um and they had a number of different data points they needed to convey. And who better than than Vinnie Jones to reprise his Lockstock and Two Smoking Barrels uh role wearing the same camel hair coat in the same Chiswick bus garage. I mean people in the advertising community will appreciate this, um, where he offed some people in lock stock, um, with two very tough black dressed portly henchmen and some geezer is presented as having had this heart attack, um, and he shows you what to do. Um and I mean I know from I mean, it's not it's not difficult to know this, but particularly from my psychological research that that that that an emotional memory, an emotional experience is going to be more memorable. Um a really great emotion is happiness and laughter, and so you have Vinny, rather than having m come kind of love and hate or mum and dad, he has hard and fast, which is how you should push down a geezer's chest. Um uh and and uh this person who's supposed to have a heart attack comes in. Uh and it's an amazing piece of data storytelling because uh I mean, in uh in 90 short seconds um uh he shows you how f how fast you should push. Now, the the planners, I can't remember the agency. Anyway, it doesn't matter. The planners went into the BHF and said, we've got two tunes that we could use that have the beat. You want 110 to 120 beats a minute. You've got two tunes that we could use. So the one that we're gonna recommend you use is staying alive, because it's it's a meme, everyone's seen their dad, uncle, granddad dancing at a wedding. When you're under that moment of extreme stress, you really need to be able to remember it and to feel it. And everyone can everyone can hear it. They said, Okay, yeah, okay, that's we we like that. We like the humour, staying alive. It's usually he's usually bumping people off. I we get that. Tell us, what's the other tune? And they said, Well, it's got much better drum and bass combinations, just as well known. And it's Queen's Another One Bites the Dust. Now we think you'd they said, no further questions. So you've got that you've got that comic juxtaposition. You've got the you've got Jones and Hoodlums and Bus Garage and Bloke dying, but actually rather than killing, they're helping him to stay alive. Uh and then he says, push down a couple of times a second, like to the beat of staying alive. So there's a really there's a there's okay, that that's why. Um push down five i he knows his audience, right? Very important to know your audience. Um Britain's a strange place. I well remember uh when I was growing up Blue Peter, where there was the great debate about should we have metric or imperial, and they ended up with both for their recipes, as all recipe books to this day still have in this country. Um uh and so he says push down um a couple of in uh five or six centimetres metres, that's a couple of inches in old money. Okay, so he knows his audience, it's a broad audience. Better a cracked stur better better a cracked rib than he kicks the bucket. I didn't actually know that you needed to do that. But that gives you a sense of how far that is. He doesn't say push down on the sternum, he's in character, because n Vinny Jones's character in Luxstock would probably not know the words, but on the sovereign because he's wearing a medallion, because he's an East End hoodlum. And then he says, and there's an and there's another brilliant bit. So how, you know, the question in the brief, how are we going to encourage people not to do mouth-to-mouth resuscitation, which which many people find revolting, it can lead to it kind of all sorts of unpleasantness and vomiting, sorry listeners, um, uh if it works. Uh, but you need to be properly pro professionally trained to do that. And and at one moment he turns to the camera and he says, and no kissing, you any kiss you miss is on the lips. Um now that's not sexist and homophobic, it's practical advice, right? Um and that is an I think just a glorious piece of data story. Love that. So that's example one. The example two is an example for some training I did with the Global Insight and Analytics team, uh Center of Excellence, indeed, um, at AstraZeneca in three different locations, two in the States and one a couple in the UK. Um, and 130 people in this team. Uh and their principal internal clients are commercial, and the commercial people in Pharma, they really want to know is this a blockbuster or not? Am I retiring or not? I mean, not quite. But uh we've launched this in Europe. It's done really well. Is it going to work in the States? And super whip smart uh analytics PhDs, biochemists, amazing data-driven stories. Now, their approach before we started working with them, and we ran three half-day workshops in cohorts of 20 in lots of different places, great. Their approach, and I I said, give me, as I always do, give me uh some typical examples of the reports that you're pr you're producing, and tell me the impact that these have had. And so they did, and you know, some of the chart some of the slide decks went into three figures. Uh I think one of them had 162, but I just remember that as a sort of, wow, I've never seen one that deep. Could you make it that big? You know, Microsoft should stop it at 20. Um and some of the charts had eight histograms on them, uh with no kind of lasso around to say, look at this one. Um and so my analysis of that was that they're trying to tell a whodunit story. They're doing an agat they're doing an Agatha Christie or a knives out. Um and by that I mean with a whodunit, you're presented with, you know, something something happens and you go and see, is it is it the butler? Yes. Spoiler. Uh is it the is it everybody, Death and L is it um uh is it somebody who we only saw fleetingly? And you followed the detective making mistake after mistake after mistake, and then you get a grand reveal at the end. Now that that was the approach, and that is the approach that many who work with data, including many in the market research community, I love working with market researchers, some of the worst offenders and the quickest learners when it comes to to uh data storytelling. I mean, uh but they want to take you on that journey. They want to, I may I use the phrase parades of scholarship, but they want to take you on that journey and say, we looked at this, we looked at this, we didn't find anything. We looked over here, didn't find anything. And then we looked over here and we did. And the commer and I and I sat in on some of the meetings, and the commercial teams were, you know, rubbing their back of their heads with their hands, almost making themselves go bald like a baby in a bassinet, you know. Um and I said I said, Well, there could be another way. So you're doing the whodunit. And you know, I love a whodunit. I'm very I'm always fond of watching Knives Out and reading the odd Agatha Christie. But it's like me going to a magic show. I don't expect to try and solve it. I'm just going to be taken along. However, um, and I'm old enough, just about to have watched Columbo as a kid, uh uh 70s, 80s cop show, this bumbling guy, Peter Falk. Um, and there's an there's a new uh series, Pokerface, uh season show that Natasha Leon and so those so Columbo and Pokerface are they're not who done it, they're how catchums. That that's what that's the narrative style of uh of that uh murder mystery. So before the first ad break, or before the first the first break, you uh as the audience have the authorial position of God, uh or the or the view is the view position of God, and you see who did it. You see, you see that it was Miss Scarlet with the lead piping in the ballroom, right? You you see what's happened. And then the the tension and the frustration is that either Natasha McGillon's correct or or Columbo, and there are others too, but those are too archetypal. Um you see them not picking up on a piece of evidence. How could you miss that? Yeah. Um but you see everything. You uh you know what the answer is, and the interest is in in is how they get there, rather than you go on the meandering journey. So I said to these AZ guys, why don't you completely change the way that you have your commercial meetings? Um and they said, Well, I'm not sure we could do that. And I said, look, say slide one, this is what we're here to do, drug X in Latin America. Say, on the next slide, I'm gonna tell you exactly what the commercial opportunity over the next five years is. Suddenly the commercial team set up on their seat. But I'm only gonna show you that if you stick with me for three areas or three data points or three additional bits of information that have led me to that conclusion. Right. And that led to a complete Vault Fass and a flip from Insight and Analytics being the party poopers to them being commercials' lead partner within the business. Complete transformation. I have my my work there is done, I have nothing else to do there. Um they are na they are now all how catchums rather that rather than uh who done its. But I think I understand the motivation of those that work in research analytics data, data science, um, data cleansing, I understand the the desire and the need, but it's it's a kind of it's an it's an empathy fail, uh and it's a failure to grasp that that the audience doesn't they trust you. No, it's like if you employ one of the big MMM consultancies, they've showcased their credentials. They don't need to sh give you their R code or to or to to to tell you every bit of analysis they've done. The CMO wants the answer to be able to go to the CFO and say we need to do this because it's working here, it's not working here, therefore. Um I understand that motivation, but it's it it it Cognitive Psychology 101 says that those who haven't been involved in that process of analytics will put their wall up and will not listen, will not hear. Um so two examples.

SPEAKER_01 32:57

Fantastic examples. And I think one of the um things I mean, I'm certainly guilty of this, is is you might do a piece of research, you've got loads of data, and then but the the fundamental that I think I I've discovered just by talking to you just there is you put together a presentation, you think, all right, I'll put all of these charts in. The thing that you haven't got is a story. And and actually I think when people go to uh it's uh I mean I'm reminded of that that uh line from Gladiator, but you know, where when people go to see a presentation, whether it's a conference or something, they want to be entertained. They want to they want their brain to be stimulated in a similar way, I guess, to how if you watch a you know an exciting film or something like that. You want something to make you feel, you know, emo an emotional response to it, apart from boredom or you know, at worst, disgust or something, you know. But but telling that story, or you know, with the data, removing and just going, actually, I guess stepping back for it and going, what am I trying to tell like what am I trying to tell? What's the story here rather than what's all this? How do I just get all of this data into a 30-deck?

SPEAKER_02 34:04

Exactly. I I mean my my my uh I've said it, I've quoted it many, many times. I'm a big fan of uh Dan Pink, the American business writer. Um and he's I mean lots of books, When and and and Drive and so on. Um but the one quote that comes back to me time and time again uh is from um to sell as uh to sell as human. And he says we're all in the moving business. Uh and we are all in the moving, uh moving and persuading business. Um if we work in commercial communication, we want people to do more of something, less of something, start doing something, stop doing something. Uh we want them to consider the rational and the emotional reasons for doing it. You know, um uh I I'm always reminded, I can't remember the comedian now, but comedian who said that, you know, smoking campaigns that show that anti-smoking campaigns that don't say, you know, smoking makes you dance like your dad at a wedding are missing a trick here because because I th I there this I've clearly been uh either dancing at weddings to make that that reference twice. Um but um uh you need to give people information, absolutely. But if you want to check if you want to motivate their a change of behavior, new category, new product, totally new way of doing things, you know, um uh uh you know, Netflix in its different in its different evolutions from from love film through to to streaming as the technology allowed. If you want to motivate people to change behavior, you need to give them information. One, absolutely, but telling them doesn't just work. You need to give them motivation, there needs to be a reason for that, and you need to quite often give them the b the behavioural skills on how to do that. Um information alone is never enough, and I think that's a common data storytelling fail. That's how I see it through my glasses anyway. Fascinating.

SPEAKER_01 35:52

Now, in your obviously your most recent book, uh um Asking Smarter Questions, um, and this is something I that I I attended I attended South by Southwest not so long ago, and I attended a number of the education seminars there talking about you know the impact of AI on the education system and the and I also there was something I've watched recently relating to uh kids who were exposed to technology from a very young age versus those that weren't, and the ability, like the innate questioning ability, uh and the kind of that uh willingness to always be on a kind of a learning mode because you know that you don't have you know, it's they say scarcity is the mother of invention, I think it is. You know, that having less available to you makes you more, more um, you know, makes you you what you have better. I mean, do you think that increasingly as well with it with the world of AI coming into things, have we fallen into a trap of asking dumb questions? You know, are we are we um yeah, you know, are are we just going, well, what's the easiest question to ask rather than the the the hardest one? And and I guess what I'm what I'm getting for is aside from obviously buying your excellent book available at all good retailers, um what what kind of if you had to give like one small tip just to get people started on that journey of asking those smart questions, what would your one be?

SPEAKER_02 37:13

So I don't think AI is to blame. It may be an accelerant, and I've and I've I've got an analogy I'll give you shortly, but um I think this has been going on for a long time, um uh kind of ever since schools and workplaces existed. And I mean it in this way, because first in education and then in the world of work, we are rewarded. We are given an A star, we're given a house point, we're we get we get a GCSE, we get a we get a degree um based on the answers that we give. We are asked questions and we and we and we uh we're we're validated and rewarded for the answers that we give. Our bosses, our teams, our clients pose questions to us and we give answers to them. Um and in school and in university and the workplace, um when we ask questions in response to questions, we're seem to be asking out or or we're given diagnoses for for um uh not being able to pay attention. And that often is that often is absolutely true. I'm not minimizing the ADHD um misdiagnosis and now probably correct levels of diagnosis. I'm not minimising that at all. But we are there's not time in a particularly in the state school setting, but actually in any school setting, there's not time for all of for everyone in the class to ask a question. There's a curriculum to deliver there. So we shut things down. Um and and I and I think and I think we are rewarded and validated for the answers that we give rather than the questions um that we ask. I think AI is accelerating this, and this is my analogy. I mean, I've talked about the kind of rock star data scientists, rockstar um prompt engineers. Everyone was panicking two years ago. We've got to get a prompt engineer. We'll pay £100,000 to have a prompt engineer, we've got to. And then there was a slow kind of dawning realization that that maybe our chatbot LLM friends could tell us what sort of questions we should be asking ourselves. Um and I think the accelerant factor is here because if you don't if you if you let's say if you write a brief using AI uh and then you review what's presented back to you using AI, is this good, how could it be better? Um or if you um if the questions, if your key business questions are determined by your LM's response, well that's okay, but it doesn't have they they don't have motivation or context or emotion or all the other things that's necessary. You ask the question or or you you get the questions framed and you refine the questions using your copilot or whatever, and then you ask them and then you present that on. It's a bit like a restaurant critic, this is my analogy. A restaurant critic saying to some junior in the operation, could you go and eat the meal, and then somebody else, could you go and write it up, and then I'll and then I'll just I'll fire it by lunchtime, it'll be great. Um I I think I think AI is is incredibly useful for an awful lot of things. Um but I think in a in a an education and then into a work culture that validates answers rather than questions, um, it's natural that we would have undervalued questions. Um and yet that's how we make sense of the world. It's how from a very young age we make sense of the world. We want to know what the causes and effects are. We ask you know, we ask um uh uh if I do this, what will happen? We we do that. We are you know the metaphor of children being miniature scientists. We that's how we make sense of the world. Um uh education and work have have put that in a in a bad position, I think. Not for everybody, but but as I say, the information crisis is not a we haven't got enough data, it's that we're not curious enough or don't ask the right type of questions.

SPEAKER_00 41:10

Do you think there's a way that we um or that adults can sort of intentionally rediscover that curiosity and and the sort of yeah, desire to keep on asking more questions?

SPEAKER_02 41:20

I do. I mean I do. Um there's a there's a Harvard um uh education school um prof called Howard Berger, who wrote a book called A More Beautiful Question, which is all about the question why. Um and he found that across cultures, um, be it pourquoi, pequé, or why, by the age of five we've asked that question 40,000 times. And then school and work squash that, and it takes another 13 years to do the same. Um that sense of curiosity. Um, Simon Barron Cohen, um uh the who runs the Autism Research Centre uh at Cambridge, uh wrote a book a few years ago that was looking at um what he believes is the thing that accelerated human cognition past other hominids, um, but but particularly really accelerated human uh cognition, which was he calls it the systemizing module, our ability to uh to understand that there are these what he calls the if and then contingencies. So as a small child, if I pull the cat's tail, am I gonna get a scratch or is it gonna purr? If I put my sister's fingers behind the plug and switch it on, am I gonna get extra tea or am I gonna go to bed early? Is it the naughty step for me? So we want to make sense of the world. Um we we come equipped, we come equipped. And it's not just a human thing. I mean, you you look at kind of play fighting in in, you know, you look at any Attenborough documentary, there will be that moment when the young cubs are roughing and tumbling and apparently ripping their throats up, but actually no, they're learning. Uh and they and they come equipped uh or birds learning to fly, uh, and and and so so I mean we come equipped as creatures with the cognitive architecture to be able to learn how to do the things that we need to do to survive to pass on our genes, but we also uh f for for us a lot, that's supplemented by language. We don't just have to pull the cat's tail, we can say to our sister, if I pull the cat, no, don't do that, because I did that and I got this bad score. Okay, we we we we learn. But I think uh I mean I think there are some there are some principles, some fundamental principles of of asking genuinely smart and smarter questions. Um one of them is is about open-mindedness, um, this kind of Socratic paradox that all I know is I know nothing. Um a second is about kind of pre preparation and working, you know, preparing for the scenarios we're going to be in. Another one is about is about um curiosity, there's I mean which is which is really driven by this this desire to know the world. Open questions, not not that's kind of like hypothesis testing. Rather than proving you ask an open question to get an open answer. Being really simple in your questions. I always remember at the during the COVID press conferences in the UK, particularly in the first few months, when they were all done online, you had, you know, Prime Minister and Chief Medical Officer, Chief Scientific Officer, or somebody standing in for the Prime Minister. And the journalists were allowed on a Zoom screen or allowed one question. But of course the journalists wouldn't ask one question. They'd ask six. And then the politicians and advisors would choose the one that they wanted to answer. And I think that cluster questions, too complex a questioning framework, allow big data databases, qualitative interviewees, focus groups, to give you the answer that they feel most comfortable doing. So simplicity is good. But the fundamental, the at the root of it all is actually listening. Is actually, you know, you put an you put a query into your database and it spits out, and then you un you take some time to understand what that means. And I realize I'm, you know, all of the things I'm saying about, you know, ask more questions, spend time listening, prepare your presentation, don't fall short in the last mile, as Scott Baronato says, um, when it comes time to explain this stuff to decision makers. That sounds sort of almost antithetical to the modern hyper-paced, AI-powered workplace. But I think if you do build those things into your workday, your ways of working, your ways of thinking, your ways of leading a team, you'll paradoxically have more time because you won't have to do them again and again and again because the answer doesn't satisfy um uh those that have asked the question. So yeah, I think I I think curiosity really matters, but the real fundamental thing is listening and and seeking to make sense. We need to move from so what, what do the data mean to now what this is what we should do as a result. Um and that is elegantly simple, but takes a huge amount of time and effort to do. Yeah.

SPEAKER_01 46:04

And it I guess it goes back to that thing of whether like my mum will always say, you know, it's better to have a small amount of something really good than a lot of a large amount of of stuff that's just so-so, you know, not very good. Um now we're seeing a an explosion, obviously, in recent years, you know, particularly accelerating in the last two years, of available data, of dashboards, of AI-driven insights, and and you know, here at the IPA with our touch points, but we're certainly going through a almost a kind of a a revolution, but also a crisis of identity, of working out how data sets fit into the modern uh modern landscape um as we go forward. Um I mean, do you think all of these, you know, the as possible said earlier, like that scarcity being the mother of venture, the more we have, and I guess it leans into what you said about Socrates, the more we know there's the more we don't we know we don't know, in a way. Um has it made it harder to make uh for leaders to make good decisions these days, do you think? Like because there is so much stuff.

SPEAKER_02 47:02

Yes. Yeah, I uh yes. I mean I could just say yes, but I'll I'll I'll understand a bit more. Um I mentioned this limiting self-belief about about more dashboards, more data, and superior analytics. Um we will therefore make better decisions. Decisions are I mean, this is uh uh another one of those paradoxes, but decisions are usually judged on outcomes, um, which is a a fine metric of success. We want to know we ran this campaign, what did it do? Did it have a positive effect or a negative effect? We incorporated this data set, did it enrich our understanding, allow us to do new things in in new ways, or not? Um But I think that actually one thing we don't focus on nearly enough, and sometimes at all, is how good those decisions are that we make and the criteria on which we j we we we we make our decisions. Because, for example, um events, dear boy, events can get in the way, as um as as has been said. Uh you can you can um you can decide that plastic is that we're gonna we're gonna we're gonna bulk buy plastic and we're gonna um we're gonna uh um uh uh to to machine print this amazing new product. And then the US and Israel decide that it's probably time that we had a bit of a conflagration in the Middle East. Or however however those things happen. And the and the price of oil goes from 70 to $120. Uh and you can't afford that plastic. And it will therefore have been a bad decision to make that because the outcome was affected. Or I tell you what, we make a really good artisan cheese in Somerset, and our dairy in 2015 has won every award, and we're gonna open up massive export markets, and then suddenly you can't get your cheese out of the country. Um, I I think there are some criteria on which we can judge the quality of our decisions that actually enable us to do a kind of pre-mortem on our decision-making. So I think there are five independent levels. The first is how clearly are we expressing the decision that we that we need to make. Um so uh I mean, a classic example might be, I don't want to cause controversy in the advertising industry given that some companies had a mandate to come back to the office five days a week and others were more laissez-faire, but the decision, the decision to come back to the office was generally a low quality decision. However you cut it, however you look at it, it was a low quality decision. It what it wasn't, it was largely CFOs saying, we've got all this empty space, what are we gonna do with it? Are we gonna get gonna get rid of two floors? Or and so let's let's let let's make that out to be a decision that and let's have some consultation. But so how clearly do you express that decision? What are the live options? So do you have is it at the toss of a coin? Do you have 20 different options that you can do, or do you have a more manageable, sensible two or three options and maybe five options that you can do? Sometimes we're operating in very low evidence environments, usually not these days. There's usually plenty of data, but is the evidence fit for purpose? Is the data that we have does that enable us to assess, uh judge between the uh the the different options? Fourthly, who's uh going to be accountable for this? Who's responsible for this? Too often, I think, decisions in business, but no and in life. I mean, the you know, this decision quality decision quality is not just about business. Too often decisions are made in the passive voice. It has been decided that every employee of our agency group will be working three days a week, Tuesday, Wednesday, Thursday. I mean, uh it's a specific case. I don't want to get hung up on that specific case. But too often decisions are presented in the passive voice, and nobody's nobody's bum is on the line for that. Um and I think that really, really, really matters. And then the fine the the last one of these dimensions, and they're all di they're all designed to be independent of one another, is kind of review reviewability. So actually, are we are we gonna learn from this? We've made this decision. We've this is this is what we we want to look at. These are the options that we had. We had this evidence that stacked up against it. Molly's gonna be responsible for this. Are we gonna learn about it? Yeah. Or not. Or not. Um and um I mean I you very kindly mentioned books. I said that I was never gonna write another um business book, and three was a trilogy, and two was anyway. I've decided that I am gonna write something about about because I I've developed a a standard of decision-making quality that allows, using those five dimensions, that allows you to, as I say, do a pre-mortem. You can you can look at, have we clearly expressed it? No, we need to do some more work on that. Do we have is it flip flip of a coin? Uh, or are there 30 different options we could do? Um uh is the evidence fit for purpose? Is Molly taking responsibility? Is Simon gonna learn is Simon gonna learn? Um and and and you can score and you can score yourself or you can you can get scored in it, but I I think it's I think the quality of decisions we make based on you wouldn't be surprised for me to say this, the quality of the questions that we ask, the fact that we have the right data, and that that data isn't all kind of um collinear as the as the statisticians say, isn't all pointing on on the on the in the same direction, but these things are independent of one another. It's remarkably simple. Um, but it's amazing how when you get people to stop and think, either they think we've got everything we need, we can run, we can launch our MVP tomorrow, or do you know what we just need to do a little bit more? Um so that's that's going to be absorbing me for the next year or so.

SPEAKER_01 53:00

Aaron Ross Powell And I guess it's the idea of you know journey versus destination thinking, because sometimes you know it's like, oh well, this will be the destination. Oh, we've got to the destination, we've failed. And then it kind of is that becomes the oh well, that's where the business is in. And I guess you see true entrepreneurs are the ones who go, life lesson from that thing, right? Apply that to the next thing. Keep going, like keep going on, always learning. And actually, you know, most failures are not catastrophic. And particularly in the advertising industry, most failures are not catastrophic. They they should be kind of encouraged as a kind of a test and learn approach.

SPEAKER_02 53:34

So long as it doesn't become a fetish, I think that's I don't mean anything deviant, but I do sometimes feel that the West Coast uh the fail, fast, and break things. Yeah. Um is that famously on the wall of meta, I think it is. Um but I think kind of every tech startup uh really wants to embrace that. I think sometimes sometimes if you haven't failed, then you've then you've not got your badges of but your battle wounds. Um I th I think most people don't want to fail. No. But admitting to well, taking respon taking ownership and responsibility and uh and uh in the accountability bit and then building that learning to make sure we don't do it again, I think is i is absolutely vital.

SPEAKER_00 54:15

I've got a final obligatory AI question, which is like where do you think AI will genuinely be improving sense making and what implications does that have for the skills um that strategists and analysts will need in in the next few years?

SPEAKER_02 54:32

Save the the biggest and the best to the last. Um so I think um I think the tools that I have experienced, and I wouldn't say that I'm at the bleeding edge of having used everything, but I make uh use of quite a lot of things. Um what I've experienced is that the the tool, the the mass availability subscription, walled garden protection protected tools are really good at I mean, they are really good at convergent thinking. You know, predict predict the next token models are really good at boiling stuff down. They're really good at taking inputs and boiling stuff down. Um but insight I mean I've d my work-a-day definition is this profound and useful understanding, profound because of so what, useful because of now what. Um joining old and old together to make something new, um the evidence I've seen so far is that even with brilliantly scripted sets of prompts there there there's no so far replacement for for the for strategists and planners. And I think that they are often lousy at um insight because human cognition, human innovation, I say call insight is the superpower that drives innovation. Um it integrates data, but also experience, um, context, drive, and kind of selfhood or representation of selfhood for target markets. You know, we have embodied experience, we have emotion, we have empathy, we have proper motivation. I think another place where AI falls down. I was talking recently to a guy called Angus Fletcher, who's he's a professor of story science at Ohio State University, written a brilliant book called Primal Intelligence. All that that it's all about working where working, getting to insight in kind of homes, Sherlock homes like low data environments. Lots of our environments are high data, but in low data environments, or in volatile and uncertain environments, we need in his model intuition, imagination, emotion, and common sense. And those are four qualities that I don't think Anthropic or OpenAI are boasting about, that the their latest tools. And so in low data environments, neck next to hopeless, kind of random guessing. There's a lot of people about talking, you know, do the grunt work so you can do the front work. Really, really, really, really useful. But too much kind of cognitive offloading, saying, write me the questions, um, create me some synthetic audiences, uh run some run some focus groups and tell me what we should do. Um Kelly Beaver at Ipsos talks about ultra-processed data. I I think she coined that phrase, I love it. Because ultra-processed data, you know, it it kind of it looks like food, but all of the essential nutrients are stripped away, like context, like emotion, like like humanity, like any kind of connection to uh an actual consumer. I mean I think there are amazing tools that really uh enable us to do things quicker. Absolutely, absolutely. Um and they will get better and better. But I I don't think that the strategy and planning teams in agencies in-house are mortally threatened. I think that when they I I think I mean my my exposure to them, which is which is quite frequent, is that they're making quick and rapid use of lots of tools. Some are faster than others, but but that's down to probably market success and ability to invest. Um But that ability to join old and old together to make something new with context, with motivation, with emotion, with humanity. I'm not saying it's safe forever, but I don't see it going away anytime soon.

SPEAKER_01 58:29

Well, thanks so much for coming in today, Sam, to talk to us. Um really exciting episode. Um so just a quick one. So obviously you've got Data Malarkey, massive podcast, really, really good, uh humongous range of guests there that we've uh we've been we've been looking through and listening to episodes recently. Where else so you've got the um the three books already out Narrative by Numbers, Asking Smarter Questions, and How to Be Insightful. How to be insightful, sorry, I thought I could remember them all. Um and and then the the fourth one on the way, is it where else can people find you?

SPEAKER_02 59:00

I mean, the simplest place, I mean, it is said uh uh that I'm all over LinkedIn like a cheap suit. Uh so Sam Knoll's data story on LinkedIn. I'm quite I'm quite active there. It's the only social channel that I have these days. Yeah. Um and then I mean everything lives in a recently redesigned and revamped website at insightagents.co.uk.

SPEAKER_01 59:22

Fantastic. Well, thank you so much, Sam, and we'll speak to you again soon.