This interview has been edited for clarity and length.
Recently you have been writing a lot about transportation policy. And it seems like transportation is an area that people are very often angry about, and their actual behavior is different from what they say they want out of it. How do behavioral science and transportation tie together?
I suppose the first question is that we don’t fully understand the sources of our emotions. And so we might quite often effectively notice a correlation between X and Y. And therefore, when event X happens, my plane is delayed, I feel emotion Y, I am frustrated and annoyed. Therefore, I hate it when planes are late. Emotions donâ€™t come with explanations attached. They’re simply generated by a set of stimuli, for very good evolutionary reasons, I’m sure, but they don’t actually come with inbuilt explanations. As a result, it’s very easy for us to misattribute what it is we want from transportation networks, which is problem number one. Problem number two is the kind of things that are easy to measure in transportation networks, for example, punctuality, speed, distance, and time, don’t necessarily correlate all that well, with what people actually like about transportation, or what they dislike.Â
So it’s quite common for engineers to be pursuing one set of metrics. Even if those engineers research their customers, their customers don’t have the introspective powers to ask themselves, what is it that I find enjoyable or irritating about a particular experience. So there are several reasons you might be annoyed about your plane being late. And by the way, if a plane is five hours late, it is perfectly reasonable to grumble about the plane being late. If your plane is delayed by 30 minutes, and there’s simply a sign that says â€œBA247: Delayed,â€Â it’s more likely that you’re not particularly bothered during the wait by the delay itself, but by the degree of uncertainty. And therefore, you can solve that problem, arguably, in two ways. You can solve it rationally by not having the plane late in the first place. Good but expensive. Or when the plane is late, you can simply put â€œBA 247: delayed 30 minutesâ€ and people go, â€œYeah, fine. Right. Okay.â€ They may have to make a phone call, they’re probably not going to miss a meeting, because they will have built a 30 minute buffer into their day. It’s not that big a problem.
Similarly, with Uber, the problem of waiting for a cab was really a question of uncertainty. I can’t relax at this party, because now for the next 40 minutes, I’m unsure when my cab is going to turn up. And Uber solves that with the map. Now Uber could have solved it partly with predictive algorithms, which would have meant there were lots of cabs that were sent proactively to places where taxi demand was likely to be high. And Saturday night at 10:30, that will be a good thing. I’m not suggesting it’s a bad thing, although it might be wastefully costly and difficult to do until Uber reaches a certain scale. It may be environmentally unfriendly if they get it wrong, perhaps. But the Uber map solves that problem very largely from a psychological perspective, because when you’ve got some vague idea of how soon your cab is arriving, all that uncertainty and irritation kind of dissipates. And so the greater part of your emotional discomfort is probably created by the uncertainty surrounding the delay, rather than the duration of the delay.
Transport is a very, very interesting case where, at some levelâ€”don’t get me wrong. I’m not suggesting that the speed of transportation is irrelevant, or that it’s perfectly acceptable to have cars that drive at five miles an hourâ€”I’m simply saying that you tend to hit the law of diminishing returns with things like speed, and duration of travel, in many modes of transport. Those are simply the laws of physics. Whereas potentially, there’s no limit to how enjoyable you can make a train journey. There’s a limit through safety and simple physics and energy efficiency as to how fast you may want to make a train. But how pleasant the time spent aboard the train is, is actually potentially much more capable of a 10x improvement than a 10x reduction in journey time.
It will take you a little bit longer end to end, perhaps, to go from London to Paris by train. But on the other hand, all the time on the train is potentially of high quality.
Absolutely.Â The Moldova entry for Eurovision basically sold me a train ticket to Moldova. The train in the video wasnâ€™t that fast, but the scenery was good, you can move around, and there’s a dining car. It would have been a lot more expensive to make it fast.
Had they made it faster but less fun, it might have been an entirely counterproductive activity. So, in many ways, if you look at the comparative advantage of trains over aircraft, you can eat in a party of four, the Wi Fi might be better, and slower trains can stop at more intermediate stations. When you have very fast trains. It’s an extraordinarily costly exercise, both in terms of time and energy, for them to stop at intermediate places. So it’s a perfectly sensible point, just to say, if you’re competing with airlines, probably the best thing for a train to do is to emphasize the comparative strengths that a train journey, such as that from London to Paris, has over a flight, which is really quality of time. It will take you a little bit longer end to end, perhaps, to go from London to Paris by train. But on the other hand, all the time on the train is potentially of high quality. Whereas you can’t say that of a plane journey. The time on the plane may be pretty good. But that’s only 38 minutes or 47 minutes in the air.
What quite often happens is that the only way in which you can justify transport investment in the UK, is through reduction in time spent in transit. Now, that is far too narrow. And even actually, some pretty hardcore economists, like John Kay, have made the point that that is far too narrow a set of considerations to justify transport. What you’re probably missing out on there is the potential for a whole load of slower, likable, modes of transport.
Now, if journey time is your only criterion, it’s impossible to effectively justify investment in anything which isn’t fast. That would rule out e-bikes, potentially. New modes of transportation need to be considered holistically, not purely on the grounds of the journey time from A to B. And I think that’s happening increasingly. The other distinction it fails to make is if it’s a journey you only take infrequently. If a thousand people save 10 minutes on a journey they make once a year, that isn’t really a game changer. It’s not really meaningful, although it looks like a lot of time saved. On the other hand, if you reduce commuting time, it may affect fewer people. But actually one person saving an hour a day, 200 times a year, is a complete game changer in terms of things like property values, commutability, and access to employment. And so we probably over-invest in speeding up infrequent journeys.Â
You could argue that the Concorde was an extreme case in point in that case. Very few people spent long enough on the Concorde for the speeds to make a meaningful difference to their lives. There are probably 50 people who did, or 100, but I calculated it once that the person who was the heaviest Concorde user of all timeâ€”he was actually on the inaugural Concorde flight, and he was also on the final Concorde flightâ€”it saved him something like seven minutes a day of the period of his working life where Concorde existed. Concorde for him achieved a time saving which would have been equivalent to moving slightly closer to the office. It simply isn’t that exciting.
What I think is very interesting that we got wrong as well, is that a very intelligent, perfectly justifiable model of the future of transport gave rise to the Airbus A380, which was that the airline industry will increasingly be dominated by Hub and Spoke routes. And as a result, we need larger planes to obviate the problem of rising costs of landing at major hub airports like Heathrow or O’Hare or Dubai. Perfectly logical. What it turns out is really, that would be perfectly true if people were freight. But unlike freight, humans really hate changing planes. Actually, what you’ll find is, if you bang in, as indeed British Airways have done, a direct flight between Austin and Heathrow, or you bang in a direct flight from Nashville to Heathrow, as they’ve also done, pretty much Brits and and the Nashville-ites will simply time their trip to coincide with a direct flight. You know, certainly business travelers generally hate changing planes.
It adds to your uncertainty. It’s another opportunity to get stuck in a town you have no business in.
Right. And so I think, I think that failure, I think it’s it’s part of what you might call an endemic problem, where we deem important what’s measurable rather than measuring what’s important, and because we don’t have SI units for human emotions, you know, there isn’t a unit of uncertainty or regret or anger, we treat objective measures as a proxy for product quality. And it really doesn’t work. You know, I mean, a classic example of that would be quad play, the idea which led to things like the AT&T takeover of Time Warner, the idea that things would be more efficient if consumers all got their mobile telephony, broadband, landlines and entertainment from one megalithic company. It turns out the consumers weren’t remotely interested in doing that, for what could be a host of different reasons.
…we deem important what’s measurable rather than measuring what’s important
If I’m having trouble with my ISP, I want to be able to tether to my mobile provider. And I don’t want one billing portal to mess up all my stuff.
Let’s say you got into a billing dispute with your quad play provider. That person would have the power to return you to the stone age, right? Yes, they tend to hold your mobiles, all your landlines and your entertainment sources, thatâ€™s simply giving someone too much power. And of course, a question of brand credibility. No one’s queuing up to see the latest AT&T film. Itâ€™s a question of who has plausible credentials and in a given market as well. And actually may be self deception, which is we don’t want to know what we spend on all those four things put together. We actually donâ€™t want a single bill any more than we’d like it if Starbucks billed us monthly.
It sounds kind of like what we’re going through at the World Wide Web Consortium. We’re having a fascinating series of meetings involving ad tech providers, web browsers, various platform companies, all about how to completely reinvent the future of advertising. And every issue that comes up seems to be about well, how do we affect the privacy properties of this? Or how do we show that we’re making efficient advertising? But we really don’t have much voice in that process from the audience or from someone who’s actually making the advertising.
If you come from McKinsey, then broadband telephony, mobile telephony, and over the air entertainment may seem like very similar businesses. But to consumers, they don’t seem remotely the same, you know, they’re completely different brands. And I think there’s this very interesting question with online advertising, which is, if you have a series of advertisers, optimizing, for example, conversion rates, will that lead to an individual consumer, receiving similarly optimal advertising in proportion to where they spend their money.
And my contention is that the advertising I see online does not really bear any relation to where I intend to spend my money. It’s very heavily weighted towards outlets where I’ve spent money in the past. And it’s particularly heavily weighted to those outlets which are either slightly niche or very high margin or both. Because in that case, the value of remarketing to me is enormous. Whereas if Procter and Gamble brings out a new version of Tide, or, a new variant of shampoo, it doesn’t have the same probabilistic certainty that I’m worth advertising to, nor can it have than say, if I bought a high margin product online once three months ago.
If advertising is to work, it’s a reasonable request of advertising that the advertising I see should be reasonably proportionate and commensurate with where the contents of my wallet end up.
If advertising is to work, it’s a reasonable request of advertising that the advertising I see should be reasonably proportionate and commensurate with where the contents of my wallet end up. I accept the fact there’ll be a bit more advertising for high margin and luxury products, and a bit less advertising for low margin products. But nevertheless, there should be some sort of proportionality. And I don’t see it as a consumer. Do you? I mean, I’d be interested to know, because funnily enough, in the early days of the internet, the advertising was quite relevant. And that was because in 1996, they said, These people are on the internet, they’re probably male, and they’re a bit nerdy. And funnily enough, I am male, and I’m a bit nerdy, and so they tended to advertise things which were disproportionately quite interesting to me. And now, it seems to be a completely deformed view of my past purchase behavior reflected back to me.
I see you’re wearing a shiny set of headphones. How did you find out about those? Was it from an ad?
Theyâ€™re Bowers and Wilkins and I’ve got a vague idea. It might have actually been from an email from Sevenoaks Sound and Vision, which is the high end Hi-Fi place quite close to me. It’s actually a local retailer. I have bought high end Hi Fi. And actually, that works quite well. At the level of my more eccentric high margin purchases, online advertising works quite well. At the level of a Unilever. or Procter and Gamble it seems to work extraordinarily badly, because I do spend quite a lot of money with those entities. But I see nothing from them. So they might say it’s simply not worth outbidding Sevenoaks Sound and Vision to gain your attention, because you don’t have any comparatively high value as a consumer in areas where we’re seeking to make sales. But at the same time, if you imagine what happens if if a huge number of people are getting advertising, which is misaligned with their interests, then you have to say that for all its talk about targeting and quantify ability, digital advertising might end up as being economically less efficient at the macro level, than mass advertising models.
If every ad improves its conversion rate, then either people have to get more money to spend, or they have to start seeing less advertising. There’s no way that every ad can be as successful as the case study without somehow introducing more money into the system or taking some advertising out.
So you’ve got a kind of negative sum game, in that sense?
…the competition to appear efficient actually conflicts with the need to be effective…
If you think of the total amount of money in everybody’s wallet, and the total number of pixels of advertising that you look at somehow, then if the totals stay the same the average ratio isnâ€™t going to change.
And so you will have this misalignment that was always there, I think, where high margin products spent disproportionately more on advertising. Cars, which aren’t necessarily high margin products, overspent quite a lot. But some things become misaligned. And it’s probably because the competition to appear efficient actually conflicts with the need to be effective, I think. Because if you think about it, people who sell direct will always have an easier time justifying ad spend than people who sell indirectly.
Yes, because they’ve got simpler conversion tracking.
People who sell directly will make higher margins as well. And also the time between ad and purchase will also be shorter. So if you have an impulse purchase, for a niche product, which is high margin, which you sell direct, you will end up advertising a hell of a lot to your past customers, to a point where it may be impossible for anybody else to reach them. There’s a whole lot of system dynamics here, which I need to explore more. Because if you can only advertise to people where, you know, probabilistically, you will be able to generate revenue attributable revenue from them in the short term, then those people will basically dominate the field, which is kind of what they do, actually.
There’s kind of an oversimplifying assumption that we make in talking about web ads, which is that all sellers are making an honest offer of a legitimate product. And realistically, in the web ad system as in the real world, there are a lot of deceptive sellers. And they can participate in the same market for ad impressions that the honest ones can.
Yes, and you don’t actually have the small act of policing, which the media owners will undoubtedly have performed, of course, which is, if expert readers who are seeing the same ads as inexpert readers of a given publication can act as punishers effectively. And so effectively drive up the cost for dishonest advertisers.
If you read a paper hi-fi magazine, and there are ads in there for crappy headphones, then all the readers of that magazine will find out.
So that advertising to some extent worked because you couldn’t control who saw it. Which is kind of interesting, isn’t it? It was always said that great advertising will kill a bad product very quickly. In fact, if you’ve got a bad product, the worst thing you can do for it is good advertising. Because bad word of mouth and disappointment simply spreads faster.
As a result, there’s a reasonable proxy, which is if someone is running good advertising over a period of time, they at least have a tolerably good product. Because they’ve survived the act of advertising for several years, which they wouldn’t have survived if there were widespread punisher activity.
In fact, if you’ve got a bad product, the worst thing you can do for it is good advertising.
I just recently spent something like $20 extra per tire to get the Michelin tires because I remember Michelin was advertising in Car and Driver when I was in the eighth grade.
Absolutely right. So there’s a kind of Darwinian thing, which is that the duration you’ve survived. And also, you could argue that advertising is a display of commercial fitness, that if you were intending purely to make short term, quick bucks, you’d simply sell via other means, and it only pays to advertise if you’re planning to establish a long term reputation for quality and probity. So it is a real indicator, if you think about it this way, that you’re actually interested in a long term future, not merely a short term sale, which is of itself indicative of an honest trader.
In other words, a restaurant has to build up a reputation because it’s stuck in the same place. A pizza van can theoretically sell very bad pizzas to 300 people and drive off to a different town and perform the same trick all over again.
Yes, and that’s why pizza restaurants here in the USA anyway, very often sponsor Little League teams, because that that is putting their brand name on somebody’s Little League shirt that’s going to be in their closet for years, and establishes that the same company has been in the same business for long enough that it’s very likely they’re selling a good product.
But then, of course, the other thing that you’re not distinguishing between, when you’re very keen on measuring your sales, is you’re not distinguishing between sales you’ve created and sales, you’ve merely accelerated. Which I think seems important to me. If you’re the Hilton chain, getting someone to stay at the Doubletree rather than the Hilton or the Hilton versus the Doubletree actually looks like a fantastically successful bit of marketing. But actually, it’s more or less worthless, because they would have stayed with the Hilton brand in any case. Whereas getting someone to stay in a hotel, rather than AirBnB, is potentially worth a few hundred dollars. But it’s not really possible to distinguish between because we don’t know the counterfactual, what that person would have done had we not advertised.
Part of it is how we manage marketing and advertising. The CMO who’s responsible for results this quarter, or the agency that might be with a brand for a short amount of time, isn’t really incentivized based on long term value.
And actually, if your jobâ€™s at risk, short term incremental improvements are actually good career advice, in a way, aren’t they? If you’re nervous about your job, and you don’t have any thought to futurity, then the safest career approach is effectively to say, this quarter I did things 3% better than I did last quarter. Actually, what you may be doing is simply bringing sales forward from a subsequent quarter, or, or identifying people who are already likely to buy.
No, the incentive to justify your own existence, always trumps the incentive to actually be a deep seeker after truth.
And to be fair, there are companies that are doing more sophisticated stats around that. But the incentive to get the data science right is not as great as the incentive to show results.
No, the incentive to justify your own existence, always trumps the incentive to actually be a deep seeker after truth. Absolutely true.
So we’re now in the process of essentially completely redesigning the web advertising market.
Partly in response, presumably, to the disappearance of cookies, is that right?
The disappearance of cookies is one part of a two part trend. And one part of the trend is privacy regulation. And the other is privacy technical changes, where browsers and mobile platforms are eliminating some potentially risky tracking activities. But the response to this that you see at organizations like W3C is how do we invent completely new market mechanisms for how advertising works?
Very interesting. So the interesting question, therefore, is how do you recreate that fairly benign circle, the opposite of a vicious circle, where the interests of content owners and creators, the interest of advertisers, and the interest of consumers and readers, were all fairly well served? And in fact, you know, we’re in a benign relationship, because we need to consider the interests of all three. What’s undoubtedly happened to content creators and owners is their aggregating power, and their halo value has been completely discounted by current modes of targeting. I think that’s probably fair to say, isn’t it?
Yes, every impression is in competition with every other at some level.
The other question, which I think needs to be addressed, is I don’t think in the long term, monthly subscription, as a mode for paying for content, is actually fully scalable or sustainable for psychological reasons. However rich people may be, there is a limit to the number of ongoing monthly payments they’re prepared to commit to.
Almost every user reader of subscription publications I talk to says, it would be great if we could just pay by the article. And then when I talk to people who run subscription based publications, they look at their analytics, and they see people who are now paying for a monthly subscription, but would be paying much less if they were just paying for what they read. Therefore, they decide that selling content in units smaller than a month is cannibalizing monthly readers.
The only thing I’d argue against that is that everybody said that about pay as you go with mobile phone tariffs. It was only with pay as you go that the mobile phone became a mass device. There were simply too many consumers who wouldn’t commit to a rolling subscription to make the mobile phone a universal device. Secondly, many of those people are prepared to pay for autonomy, by which I mean there were many, many people on pay as you go arrangements who spent more with the provider than they would have done had they been on subscription, but showed no inclination to convert to subscription.
In mobile games, there are people who will not spend $4.99 on a game, but then they’ll get a free game and spend $700 on in-app purchases.
That strikes me as fundamentally, an economics influenced finance guyâ€™s excessive belief in cannibalization. In other words, you know, in other words, I think that that it’s fundamentally a mistaken assumption in that many of the people who subscribed beforehand, you might have to slightly change the terms of their subscription to say that, you know, in the months where you don’t visit, we refund you, or something of that kind, you might have to slightly sacrifice a small amount of revenue. But broadly speaking, many of those people will stay on a contract tariff with those people they want to subscribe to, I don’t think I’d come off my Netflix subscription if I was allowed to pay three bucks per film, because the ability to explore is simply worth too much for me.
Or even a â€œpause your subscriptionâ€ feature where you could say, â€œI’m not going to read this for a while, I’ll put it on pause.â€ And the next time you click through to something, you can re-activate.
So you could just change the terms very slightly for people on subscription arrangements, and effectively pay as you go would I think generate, first of all millions and millions of more users for paid content, but also millions and millions of more dollars, ultimately, if you looked at it in aggregate. Let’s take an example, say, The Wall Street Journal. In the Wall Street Journal, or The Economist, there are many, many articles, both of those entities, which I would pay $1 for. In fact, there may well be 10 articles a month, for which I pay $1. Part of the reason I donâ€™t subscribe is that I won’t subscribe to something if I feel I’m getting disproportionately worse value than other subscribers. Just by dint of me being a Brit, and not working in finance, the Financial Times, or The Wall Street Journal is going to produce a hell of a lot of content, which is of no interest to me whatsoever, and I feel I’m paying for it. So the relative value of the articles I do want to read is actually diluted by the negative value of the articles I don’t want to read, but don’t want to feel I’m paying for. I don’t want to pay for an article on the prospects for central banking reform in Ecuador, you know, because it doesn’t really have much bearing on my life. And so for that reason, I think there are some tragic mistakes there in terms of not understanding the psychology of value.
One way to do it, of course, one way to solve the problem, which I suggested in the UK, as I said, in the UK, why don’t you put up the BBC license fee by $10. And give everybody $20 worth annually of some Pay As You Go currency. Now, at that point, there’s suddenly you know, 400 million pounds of potential revenue floating around, and everyone is going to suddenly make their site accessible to it.Â
I was slightly unimpressed that the investor community was aghast at the fact that Netflix lost subscribers when lockdown ended. You had conditions for about two years, which were as perfect conditions for subscribing to at home content as you could possibly imagine, which is that you couldn’t leave your own house. And for the investor community, and for Netflix themselves, to be surprised by their falling subscriber numbers, at that point, didn’t say a lot for their understanding of human behavior because it was a little like saying, â€œisn’t this extraordinary that when you release people from prison, they no longer trade cigarettes.â€ Trading cigarettes is a condition of being in prison. It’s necessitated by being in prison. When you free people from prison, they don’t do that so much. It’s completely bizarre that both the finance community and Netflix themselves had failed to predict that.
Peloton bikes were another example of a pandemic product sort of returning to normal conditions.
Because, apart from anything else, there are a lot of people earning money working quite hard, who had no money to spend on things like holidays, restaurants and everything else. You have to be a bit of a grumpy dad in conditions of lockdown to say no, kids, you can’t subscribe to Netflix. Given that you weren’t going on holiday, you weren’t going out for meals, you weren’t fueling the car. And so, that struck me as highly surprising, which is, it doesn’t strike me as requiring a huge amount of human insight into behavior, just about that fact that there was going to be a fall off in subscriptions.
Yes, it’s surprising how much more home life has returned to normal than office life has.
I suppose I predicted that actually. What I predicted, I think, was a little bit of this kind of great resignation revolt against five days in the office, which is that essentially, if you think about it, everybody in an office job assumed that their working hours were a given, and that their working location was a given. And therefore, the one variable over which you compete, over which you would compete was salary. The other things were basically static. And then suddenly, we have this realization that actually, no, this isn’t. Previously, you could have said, â€œI think I can work remotely,â€ and people would say, â€œyeah, in theory, prove it.â€ Well, two years of working remotely did kind of prove for quite a lot of people that it was perfectly feasible, in some cases, to be more productive, and happier and richer, working in a different way.
Donâ€™t miss part 2 of this interview. Rory explains more about the new, remote work environment. What the three â€œFreesâ€ that more and more people are negotiating with their employers, and how they can help you? Plus, will an under-rated Internet invention, more than 20 years old, do for offices what the electric motor did for factories? Continue to part 2.