Lightly edited for readability. Speaker 1: Welcome to The Business of Marketing. My name's John Horsley, I'm the host of the show today. We're recording live at Cannes Lions. We're in the beautiful surrounds of the Deput Secret Garden. I have the absolute pleasure as well of being joined by Alistair Hill who's co-founder of On Device. Alistair, welcome to the show today.
Speaker 2: Thanks for inviting me. This is a beautiful place and a lot calmer than being out the front .
Speaker 1: It really is. It's nice to be sat in the cool and the, and the shade. Um, I would normally ask this question as a, as a starter for 10. So we're obviously working within the world of, of marketing, media, advertising, you know, et cetera. What attracted you in in the first place? What got your attention? What made you want to become a part of this industry?
Speaker 2: Uh, that is a really interesting question, and I think I've got a bit of a silly answer to it to be honest. I was doing a business degree and, um, I think I quite fancied somebody who was in the marketing degree, and I changed course. And uh, that was a long time ago. That was 25 years ago, and then that got me into it. And after that I just found it fascinating. And, um, yeah, and then the rest is history I think sh- pr- probably, probably say.
Speaker 1: Yeah. And it'd be really great as well to understand the, the purpose or the raison d'être, as it were, behind your business On Device.
Speaker 2: Yeah, so if we actually go back to university which is a little bit ridiculous, I wrote my dissertation on measuring the effectiveness of digital advertising and, um, realized that the idea of using, um, metrics such as click through rates and things like that was gonna be useful for performance advertising but was actually gonna have no bearing whatsoever on brand advertising. And so although I wasn't the best student in the world, um, the conclusion to my dissertation is basically what my business is. And so we realized very quickly that, um, digital advertising needed some way of being compared against other forms of media and... Yeah, that's, that's what On Device does.
Speaker 1: Okay. That's very cool. Really interesting that it ca- came out of a university which is extremely rare. Um, many of the people I speak to start off from one direction. Uh, they don't, they certainly don't do marketing courses at, at unis. Um, and they end up f- for one reason or another, um, joining this particular industry, so that, that is particularly fascinating. So y- your story started there and you've spent many years now helping brands and agencies measure effectiveness. Um, so where do you think people from your experience still, still get it wrong, or perhaps where they're asking the wrong questions?
Speaker 2: Yeah. And so, um, if we go back to that just original comment that I made about how performance metrics are not going to be useful for brand advertising, um, what a lot of these metrics are are easy metrics. And I mean, basically this theme runs throughout my career and so when I was at Comscore many moons ago, there was a, um, wonderful piece of work, um, written about, um, how weird people are who click on ads. And it turns out there's an in, like an inverse correlation between, um, people clicking on ads and, um, the branding impact of, of advertising. Basically people who click on ads are pretty odd people. There's not many people who do that. And so using those sort of straightforward easy metrics as a way for measuring the business impact of brand advertising is completely mad. And you think because everybody knows this that this doesn't happen yet um, and I won't name names, uh, with a client recently talking about how wonderful their brand campaign was 'cause they had twice the click-through rate of any other campaign that they've had. And you're like, "Well why did you even mention that?" And so, um, what seems to happen in our i- industry is that easy things are quite often used rather than harder things which actually make a difference. And really that depends on the sophi- sophistication of that brand and often it actually depends on the sophistication and confidence of like the individual person that, that you've got there. And I talk about confidence quite often because doing the hard thing isn't necessarily the cheapest thing sometimes and often w- almost always is, you know, isn't the easy thing to do. Um, and so having more confidence in your job and in your role actually quite often leads to doing these slightly more complicated things which actually get th- the best result.
Speaker 1: Mm. Okay. Um, I mean that's, that's quite an interesting example. So they're just looking purely at click-through rate. Um, there may have been a huge difference in terms of the campaign that they're running this time versus the previous time. Perhaps it was a slightly different audience segment. Perhaps there was an offer attached, um, which could easily increase the, the click-through rate.
Speaker 2: Perhaps there was a ton of fraud.
Speaker 1: Uh, well I was gonna say there could be bot traffic or fraud-
Speaker 2: Yep
Speaker 1: ... or various other things. So in relation to that, um, or people who are looking at things in similar ways, how do you go about helping them and, and help them with their understanding of what to measure and, and why?
Speaker 2: So, um, uh, what we do, we're a Brand Lift company. Um, Brand Lift at heart is a, um, simple concept. What you're in effect doing is understanding consumers exposure to advertising generally using some form of passive technology. Um, consumers have no idea what ads they see so you need t- technology to understand, um, what, um, what ads they've seen. Um, then you survey people and you make sure you have a group of people who've definitely seen the ad, a group of people who definitely haven't seen the ad, and then you balance out those two groups to basically make sure that the only difference between those two groups is, um, the underlying effects, uh, of, of the advertising, the exposure to the advertising. Now, what smart brands do is they've done the work to be able to correlate the brand metrics and how that links through to the, um, business outcome that they want. So the general perceived wisdom on this is if you get a 1% increase in awareness or a 1% increase in, um, consideration that leads through to a 1% increase sales further down the line, generally over a three to six month period. There's generally half of that Effect happens in the first few months, so like say a 0.5% increase in sales, and another half of that after three months. And so there's kind of the short-term impact versus the long-term impact on it. So basically, the smart brands know that these metrics correlate to the thing that they're trying to do, and we like to think of our, um, brand lift metrics that we get as being a leading indicator of a brand's future success. So they can find that data up quite quickly, and they can do really useful things with it to be able to improve the effectiveness of, uh, of, of advertising in the future.
Speaker 1: So you feel that the smarter companies out there are looking at incrementality?
Speaker 2: Wouldn't say incrementality, but I mean, they are doing that as well. And what you're doing with a brand lift study is you're trying to find out what is the incrementality of people who are more aware of something or more likely to consider that brand that weren't before. So how many more people are you getting? So is it two, three, four percentage points? Um, yeah. And then they're using that data to say, "Okay, this campaign's working well, and the blue ad's working better than the red ad, or the video ad's working better than the display ad," or, um, "This channel's working better than that channel." And then they're adapting their campaigns as a result of that data to be able to improve the effectiveness in, in the future.
Speaker 1: So why do you think that, um, you know, within media still, the... There's so much emphasis on, I guess, what happened. Something's been clicked, some... You know, something's happened, um, but not what it actually did for the business. You know, it's moved it from here to here to here.
Speaker 2: Yeah. Um, I don't know. I guess there's a straightforward answer, and there's a answer which probably isn't talked about so much on podcasts. There's a, uh, there's a general backslapping thing that happens. And, um, sometimes, actually very often, the metrics are used to pat each other on the back and go, "Great stuff." And you'd be surprised of how often that happens. And you'd be surprised how, um, often our recommendations are not taken into account and not used even though it's clear. And so then we all often ask, ask the question why. We don't get an answer very often. Um, but there's, there's reasons for that. Um, and yeah, it, it, it's sometime not entirely clear what, what those reasons are.
Speaker 1: Yeah. Obv-- I mean, I've seen this time and time again, frankly, where you could have data tell any story that you wish it to, frankly. Uh, I've seen within enterprise businesses people using data, uh, reporting to pat themselves on the back rather than use it to inform what they should be doing, uh, and what the strategy should look like, so using it reactive. Um, do you come across similar scenarios?
Speaker 2: And so our mission of our business is to measure and enhance brand effectiveness. So we choose those two words carefully. To measure is to give a score. To enhance is to give a recommendation. Or now, is to give our data in a format via APIs into, um, different, um, machine learning systems to be able to basically improve the effectiveness of it. Now, we've been doing this business for 15 years. There's nothing more frustrating than somebody taking a slide, going, "Okay, well delete those two, and we'll take this one, and we'll put that out there." And everyone will, will, uh, shrug their shoulders or pat themselves on the back, and they'll move on, and they'll plan the next campaign in a completely different way. It's, it's almost like, "What's the point?" Right? Um, whereas what is incredibly satisfying is seeing them taking that data and using it in a, in a way which basically means that the next campaign is dramatically more effective, and that's becoming easier over time. And the reason why that's becoming easier is the data used to be consumed by humans and consumed in a meeting room, and you're hoping that that person then uses it in that way. What's starting to happen now is that data's now being consumed by machines, and there are models and algorithms and systems that people are putting in place to be able to take effectiveness data and put those into systems so that the next one is automatically planned in a way where that data is used.
Speaker 1: Yeah. So I'm just thinking, uh... I'll just pause for a moment just to think about the recommendations that you might provide. Um, I think it's fascinating that you provide both sides. What do those recommendations look like in practice, and how are they actioned and act on?
Speaker 2: So what you're really wanting to do is understand the combinations of things that you can plan against, that you can activate against, that actually make a meaningful difference. So we've got a lovely case study of a soft drinks brand who found out that on a hot day if you did a radio ad in the morning, and then you did a display ad, um, later on in the day, that that combination of that really worked. Um, and so that's like one example of it, and that really helped them sell a lot more, um, uh, so- soft drinks. Um, but there are a whole variety of parameters that, um, companies now have access to, which enables them to basically triangulate those and use those as a way to be able to make it more effective. So if we go through an example recently, a CTV campaign. A client found out the best time of day to do it. They found the best channel, the best genre, um, the best frequency of exposure. So if you take all of those four things and put those together, and you basically train an algorithm just to show ads at that combination of those things, turns out you make things dramatically more effective. What you're doing is you're... I mean, everybody's always, they talked about cutting out wastage and so on, but you're basically using evidence to be able to do that, and you're using machines to be able to take that in, ingest that, and deliver that.
Speaker 1: Mm-hmm. So you mentioned, you know, fraud becou- could have been one of the reasons why, you know, the, the click-through rate increased, you know, 50%, 100%, 200, whatever. Um, so you're obviously monitoring for that as well within your systems. What does that look like? Um, are there particular areas of the ecosystem or particular channels where it's more prevalent?
Speaker 2: Um, so we are not a fraud company, so I'll just state that first of all. And we see oddities occasionally, um, which generally revolves around incredibly high frequency. Um, and when- Yeah. So I, I would say that we're not the best people to talk about ad fraud. We are, however, in a other crazy world, in the world of survey fraud and-
Speaker 1: Survey fraud?
Speaker 2: ... Survey fraud. Right. So most people do not know about this, right? But, um, uh, turns out there is a very large proportion of the market research industry, um, that is bots. Normally actually people in other countries who don't get paid very much money, um, for um, uh ... Don't have very high GDP per capita, who are basically spoofing that they're in another country and taking a lot of surveys. And so this ranges from 10% to up to 40% of, um, market research surveys, which is completely crazy when you think about it. Most people don't even realize this at all. Um, so one part of our business is we own panels. We have, um, consumers who take, um, uh, surveys with us, and we have a whole team of people focused on the defense of these apps because we're giving away money, um, to, to consumers. And so really understanding the fraudulent behavior of, um, these people is paramount to our business because if they start taking brand surveys and that gets into our data, then y- you're not showing the effect of the advertising. You're showing what a Nigerian thinks about a American campaign, right? For, for, for example. Uh, so yeah, so we, um, have a lot of defense mechanisms against that. Um, we get people to hold up IDs, um, so that we can unders- use facial recognition to understand that they are the person that they say. We make sure that we, um, uh, understand the consumer's web history. Um, it's quite difficult for somebody to spoof a realistic history, um, uh, of h- how you would use the web and, um, and a bunch of other different things. Um, so yeah, so we, we don't get so much involved in the ad fraud thing, but we, we definitely get involved in making sure there's really high quality data through, uh, understanding survey fraud.
Speaker 1: Wow, that's incredible, um, and very useful for people who are, are listening. It's also made me think that there's probably elements of modern slavery that could be involved in that supply chain as well, which is-
Speaker 2: Yeah, and it's-
Speaker 1: Very concerning
Speaker 2: There are some companies who've done some fantastic work into tracking these people down, and there's videos of people with racks and racks of phones taking surveys. Um, and when you talk about modern slavery, I mean, it's difficult to understand who these people are and who's in control and who's really making the money from it. Um, but it's highly likely it's along the lines that you're talking about. It's very much organized crime. Th- This is not simple stuff like give one example of this. We put a new defense mechanism in place, um, and within a day there was a YouTube video around how you get ar- around it. Um, and a thousand people had looked at it, and there was an advert for going onto somebody's Telegram channel to have the details of how you'd get around it within a day.
Speaker 1: Wow!
Speaker 2: Yeah. Crazy.
Speaker 1: What advice would you, uh, recommend to, to brands sort of re- running ads, uh, to better protect their services and stuff like that?
Speaker 2: So the first thing to understand is that brand advertising slightly nudges people along the journey to slightly want to be, uh, to, to want to buy that thing more. And so if you ever see results where you've had a 30% increase in brand awareness, for example, What I would do as that brand advertiser, I'd be like," All right. Let's go through that with a fine tooth comb because that sounds absolutely crazy." So anybody consistently giving you very high numbers of increases in effectiveness, um, I would be incredibly suspicious of. Um, if you're seeing it in your sales numbers great, right? But, um, uh, I think what we've understood from the industry is that, um, you need to have a high degree of skepticism of things like that.
Speaker 1: Okay. S- So you're looking for initial changes over a period of time rather than just-
Speaker 2: Basically, I mean, you might have d- been the one in a thousand campaign, right? And that may have happened, right? If we see something in our team internally where we've seen these huge increases, all of our team instantly go, "Okay, there's something wrong with the data. Let's go and relook at that and let's see what's going on." And I would say nine times out of ten they are right, and they'll find something that's an issue in the data and they'll redo it and they'll think about a way to be able to make that better. Um, very occasionally we see plus ten ing- percentage points increases.
Speaker 1: And so changing subject slightly, what's happening since AI
Speaker 2: So I think I hinted on it, um, before around how our data has now being consumed by machines rather than humans. But, um, but if we take a step back and think about what, um, what AI can do for any business really is it seems to fall into two camps. One is around business model and one is around operating model. And so business model being, are you gonna start selling something different in a different way as a result of AI? And operating model being the way in which you create that, is that going to change and, and, and, and be different? So if we take the first one, first of all, so, um, a business model, um, what we are seeing is the most sophisticated clients are thinking about how they create algorithms to target ads at people. Our business is predominantly about helping people to make smarter media decisions. So where do you put your m- media investment? Um, and the machines can crunch this data in a way which enables them to understand these combinations of things which I was describing beforehand. So, um, channel, time of day frequency, um, the right audience, the right target market, th- those sorts of things. And what can then happen as a result of it is what people are calling outcomes based planning. And you can look at the outcome, which is the data that we've got, the brand outcome, and you can work backwards from that and trying to work out how you can plan as r- as a result of it. And, and it seems these systems can do this pretty efficiently and we strongly believe that the future of, um Uh, the outcomes industry. So, the measurement industry will be around feeding data into these, into these systems. If we then take it from the other side, so we think about, um, the operating model, how is the operating model changing? Um, and our team are working continuously on that. A lot of this kind of falls into automation. Um, and is it AI, is it automation, is it not, so on. Um, but one example of that is our commercial team, um, takes maybe 3,000 media plans a year and turns those into 3,000 measurement plans. So here's a media plan, how are we gonna measure this? And that can take anything from, uh 20 minutes to do to a day to do, depending on the complexity of it. And what we've been able to do is basically feed in all of these basically email chains and put them into a system, and now you can take a media plan and you can press a button and it converts it into a measurement plan, and it's done 'cause it's trained on all of that data. Um, equally, there's a whole bunch of other things along our process where we're just adding, um, automation, AI and so on to be able to make that more efficient. And what that hopefully leads to is like a better service for our clients, um, hopefully for a better cost, uh, as well.
Speaker 1: Yeah. Are you seeing clients as well building their own, I'm gonna say orchestration technologies because there- there's so many, um, channels that they work with. There's a big e- ecosystem of suppliers as we can see here and how vibrant it is within Cannes alone, let alone globally. Um, and they have their own specific needs as well. So are you seeing clients actually build and adapt, um, perhaps using AI themselves to, to rapidly build prototype, get off the ground, uh, their own technologies to, to better help them?
Speaker 2: I think, like you have to take a lot of the sort of vibe coding thing with a pinch of salt. Getting to a prototype is relatively straightforward, but building that into a workflow and into a system and how you create operational change around that is, is hard and has always been hard, and is much harder than building the prototype. Um, and so a lot of people are sort of making suggestions that we can go a lot quicker. Um, we can build these custom things that you, you, you mention. Um, and I think to a degree you can, but you can't really do that with big stuff and you can't really do that with scale, um, uh, in terms of the way that that would be integrated into the way in which, which somebody works. So I mean, uh, I'm here to be like contradicted, but um, uh, I think you need to be a little bit skeptical of that because that's still really hard.
Speaker 1: Sure. Yeah, I was won-- When you talked about weather, time of day, various sorta factors and so on-
Speaker 2: Yep
Speaker 1: ...um, I was, I was thinking about agents that people could create-
Speaker 2: Yeah
Speaker 1: ...that are responsible for different parts of the workflow-
Speaker 2: Yeah
Speaker 1: ....and, uh, in real time switching w-what their ads, you know, what their ads are saying and to whom they're addressing.
Speaker 2: And, and that's exactly what we're doing.
Speaker 1: Yeah.
Speaker 2: Right? So we have a client who is building a system that will ingest our data and have a optimized button in their system. Ingest the data, you press optimize, it optimizes to it. What it's doing is like a whole bunch of different things and will change for each i- i- each campaign. What they have found out overall is that you actually need a different optimization for each brand and for each camp... I'm, I'm not sure I would say for each campaign, but for each, um, objective of a campaign. Um, and so if you're wanting to increase awareness for a CPG product, it's very different from changing consideration for a car brand, for example. And so that optimization has to be da- based on data which is associated with, with that brand or that industry.
Speaker 1: Are you also seeing, um, as you say, objectives, people's motives changing slightly? So, um, yeah, are you seeing more investment in terms of brand outcome?
Speaker 2: Yeah. Um, so is there more investment in what we do? Probably. Is there more investment in taking in outcomes-based data? Absolutely. Like, and massively. Um, and so yeah, the um... People talk about it, it's a little cheesy at lots of different, um, uh, conferences about this era of outcomes and so on, but when you actually get down to the reality of it, um, big companies are spending a lot of time and effort working out how they ingest outcomes-based data to be able to make their systems better, and we're seeing that across the board. Um, and so, um, yeah, it's, it's not like a, it doesn't feel like it's a fad in any way. There's like, there's big money, lots of proper teams going into making that work.
Speaker 1: Yeah. Uh, before it was very much more performance, um, orientated and kind of performs marketing at all costs. Um, and think that they've realized with the proliferation and, and the ri- rise of AI, um, that they need to become better storytellers. They need to get the, the right content out, um, that people's behaviors change. So, um, you know, th-the searching, they're not necessarily clicking through to a website. They're getting the answer there within a, an AI engine. Um, and the fact that, you know, if you have strong brand presence and, and recall that you already exist in that person's mind, are you front of mind, uh, at the point when they may be shortlisting or, um, going through a decision-making process, whether that's as a consumer or as a, uh, a B2B buyer.
Speaker 2: And, um, I mean, this goes back to, um, what we were discussing right at the start about Um, the sophistication of that company and often the confidence of that person who's in, in charge on it. Like, I, I don't think there's really any doubt now in like most of the marketing world that brand advertising makes performance advertising better, right? Everybody kind of knows that. So why then do you have like a panic, "Let's throw it all into performance advertising and not do any more brand advertising," and it's related to this confidence. Um, and if that person in that job happens to be really unconfident about their survival, then they'll do something like that. Equally, if the agency that's representing that brand is unconfident about their situation with that brand, then they start to do stuff like that as well. And yeah, we've, we've seen that quite often.
Speaker 1: Mm. Yeah. I mean, that's an interesting one in terms of agencies. So, you know, they're controlling oftentimes very large ad spend. Um, so they might want to backfill with, like, "Fine, we spent this on, you know, online, and it's been really helpful, and it's built your brand and, and various other things. But here's also a gazillion leads, uh, that, uh, that we've got you as well," um, to justify level of outcome regardless of necessarily where those leads have come from and what the provenance of, of those leads are.
Speaker 2: Hm. Yeah. Um, yeah. And it's difficult as a measurement company being involved with all of that and, um, I try to stick out of it would probably-
Speaker 1: Yeah
Speaker 2: ... be my take on it.
Speaker 1: Yep. No, that's fine. Um, as I said at the start, normally wind up i- in a number of, of, of ways. Um, I will ask you a question for younger people, um, who are maybe listening. But actually prior to that, I'm t-... I'm just gonna ask you a couple of quick fire questions. Um, so a, a brand at this point in time that you believe is doing great marketing.
Speaker 2: God. I'm not very good at that 'cause I now run a business, and like I don't get involved in the work as much as I used to, which is very frustrating. Um, I, I'm, I'm gonna struggle to answer it. That would be what I say.
Speaker 1: Or a brand that you recall and, and particularly admire-
Speaker 2: Um-
Speaker 1: ... that's closely linked to what you do.
Speaker 2: Yeah, exactly. Um, oh, do I have... Uh, this is sound ridiculous. I like HubSpot. I think it's great. Like, and I like their advertising at the moment 'cause they're in the Uber app and here in Cannes, and there's a whole load of people like us who use HubSpot a lot and spend a lot of money on it, and I think that's just a brilliant activation.
Speaker 1: Brilliant. We interviewed HubSpot last week in Rio-
Speaker 2: Right
Speaker 1: ... at, uh, Web Summit.
Speaker 2: Yeah.
Speaker 1: Uh, and one of the things that we found fascinating there was, uh, the woman that we interviewed wasn't just responsible for the region-
Speaker 2: Yeah
Speaker 1: ... or for Brazil. She was responsible for Portuguese-speaking territories.
Speaker 2: Right.
Speaker 1: And that's how they've, they've sliced up the responsibilities across the business-
Speaker 2: Got it. Yeah
Speaker 1: ... rather than going region, it's by language.
Speaker 2: Yeah.
Speaker 1: Which I thought was fascinating.
Speaker 2: Yeah, yeah, yeah. Completely.
Speaker 1: And in my mind says a lot about their marketing and-
Speaker 2: Yeah
Speaker 1: ... frankly, I, I, I think it's fantastic as well.
Speaker 2: Yeah.
Speaker 1: Um, perhaps a, a, a book, uh, a go-to resource, a podcast, something that's inspired you recently and something that's-
Speaker 2: Uh, Measure What Matters
Speaker 1: ... you've learned from.
Speaker 2: Don't know if you've read the Measure What Matters book. It's, uh, John Doerr. Um, he's like ... I, I don't know if I pronounce his last name right. He's like a VC. Um, but it's all about OKRs, and it's about how you structure the way you run a business, and it's incredibly effective. Uh, it's about, yeah, um, ob- uh, objectives, key results, all that sort of stuff, and it w-was published ages ago, and I read it ages ago, and then you c- ... we came back to it maybe nine months ago, a year ago, something like that, and it's just been fundamental to how we're running the business at the moment. It's a great book.
Speaker 1: Fantastic. Brilliant advice. Uh, and then lastly, um, a piece of advice that you would offer to a younger individual perhaps who are at the start of their career or looking to enter the w- the world of, of marketing, media, advertising, tech, you know, in essence, the cluster of, uh, industries that are here at the moment within Cannes.
Speaker 2: Um, I know this might sound ridiculous, but key to my c-career has, early career, was using Excel really well, learning how to split data, learning how to gain insight out of data, and learning how to tell stories with data, and you can really impress people through, through doing that. And it isn't, uh, a thing which is used, um, well by a lot of people. Um, you gain insight from a lot of data through splitting it by lots of different parameters, and once you can see that, you can then tell a story about how the world works. And the people who can do that well are utterly brilliant, and there are not many people who can do that well.
Speaker 1: Yeah. I actually only know one. I did know two people who could do it quite well.
Speaker 2: Yeah.
Speaker 1: Um, it's not myself frankly.
Speaker 2: Yeah.
Speaker 1: Most people only use a, a, a fraction of Excel and, and what it's for-
Speaker 2: Yeah
Speaker 1: ... when it is such a powerful tool.
Speaker 2: And it's not just Excel, but like, I mean, um, any data analysis tool.
Speaker 1: Yeah.
Speaker 2: Being good at that, being able to really see stories in data is just a wonderful, wonderful skill.
Speaker 1: Brilliant. Look, fantastic advice and, again, thank you for so much for joining us on the show today.
Speaker 2: I feel, I feel a bit sweaty now, but, you know-
Speaker 1: Oh, no
Speaker 2: ... this has been great.
Speaker 1: And, uh, wish you a successful and, and a fun, enjoyable time at Cannes as well.
Speaker 2: Thanks. I, I really enjoyed it. Cheers.
Speaker 1: You're welcome.
Speaker 2: Thank, mate