The risks and rewards of dynamic pricing for business

Garreth Hanley:
This is INTHEBLACK, a leadership, strategy, and business podcast brought to you by CPA Australia. Hello, I'm Garreth Hanley, and welcome to INTHEBLACK. Today, we're talking to Jeannie Marie Paterson about the benefits and potential pitfalls of dynamic pricing strategies.

Jeannie is a Professor of Law and Director of the Centre for AI and Digital Ethics at the University of Melbourne. Her work centres on the regulation of new technologies and consumer-facing markets. Welcome to INTHEBLACK, Jeannie.

Jeannie Marie Paterson:
Thanks for having me.

Garreth Hanley:
Great to have you with us. Now, Jeannie dynamic pricing seems to have a lot of other names, including things like surge pricing, time-based pricing, demand pricing, variable pricing, congestion pricing, and even peak pricing.

Now, this is not new and markets and prices have always fluctuated, but recently, dynamic pricing has been hitting the news headlines. So my first question is, how has dynamic pricing evolved over time, and what are the factors behind its increased adoption?

Jeannie Marie Paterson:
So, dynamic pricing has been around for a while and we're probably all familiar with it. So hotels are more expensive in school holidays. That's kind of dynamic pricing because it's pricing going up when there's more demand. And we see that with airlines as well.

But what's happened with dynamic pricing with the rise of technology, algorithms, and data, dynamic pricing is becoming more nuanced to the variables that the merchant or the trader cares about. So they can change prices in real time and they can change prices in relation to a variety of factors, not just a prediction about, for example, school holidays.

Garreth Hanley:
And it can help businesses improve their bottom line or services, is that true?

Jeannie Marie Paterson:
Yeah, no, dynamic pricing can be really useful for businesses because it can make sure that their pricing is competitive compared to demand, and it can also help nudge consumers who aren't time sensitive, for example, into off-peak uses.

So if you think about dynamic pricing for electricity, off-peak pricing is a form of dynamic pricing which encourages or incentivizes customers to use electricity at times when there's not high demand for it. Or off-peak pricing for public transport does the same thing. If the price of public transport is cheaper at less popular times, you know that you can encourage customers to take advantage of that if you like slack in the system.

So it can be really useful. It can be useful for the trader, means that they're being competitive according to the market, but it also can smooth the market by pushing consumers who aren't time sensitive to use the service at times when it's not in high demand.

Garreth Hanley:
What are some of the risks and challenges associated with dynamic pricing, and what can be done and how can businesses mitigate them?

Jeannie Marie Paterson:
Well, if businesses are going to engage in dynamic pricing, first of all, I guess they have to be right about the demand levers. You don't want to be changing your pricing in a way that's not consistent with demand in the market or what your competitors are doing. So sometimes demand pricing is led by a prediction of when there's going to be a demand for the product or the service. And there's a question for businesses about how they identify demand.

They might do it by a prediction. Hotels know that they'll have more demand in school holidays. Freeways that charge tolls and engage in dynamic pricing know that there's more demand for the freeway at peak hours. But if you are an online retailer, for example, you might be engaging in dynamic pricing by your prediction about what goods are popular with your customers, or in fact by watching competitors. Sometimes dynamic pricing is predicted on what competitors are doing.

And you can get adverse feedback loops on that basis where the competitors are perhaps responding to signals that aren't really accurate about the market and it all can come tumbling down. You need to know that your trigger for changing prices is accurate. But there's some other external risks about dynamic pricing which comes from reputation with customers and also regulatory requirements about how dynamic pricing is conveyed to customers and indeed used. So there's also those external factors.

Garreth Hanley:
I mean, there you were talking a little bit about how the dynamic pricing operates under the hood, so knowing how that happens, is that part of that mitigation process that businesses need to engage in?

Jeannie Marie Paterson:
That's right. So if businesses are going to engage in dynamic pricing, they need to know what factors they're taking into account for price changes. Is it just time? Is it market signals? Is it competitors? Is it season? Is it real-time demand? And they need to know the basis on which that pricing is taking place, because if they don't, those external factors, customer satisfaction or customer trust, may be affected, and regulators may come knocking at the door if regulators think that dynamic pricing is being used in a way that's not actually appropriate in the market.

There's going to be higher scrutiny in highly regulated markets, so financial services are probably more highly regulated than consumer products being sold online. But even consumer products being sold online can have the regulator in Australia, the ACCC, looking at what's happening. And we've seen some examples of this. There's a number of examples where dynamic pricing has been used by traders and they've come unstuck with using dynamic pricing, partly because they haven't thought about how their customer base will receive that pricing and they haven't thought about the transparency of the process, which has brought the regulator in.

Garreth Hanley:
And there's a sense of fairness.

Jeannie Marie Paterson:
Yeah.

Garreth Hanley:
Public perception of fairness.

Jeannie Marie Paterson:
I think so. So we can talk about some examples if you like. One example that most people are familiar with is Uber. So Uber famously engages in surge pricing, so it puts the price of its services up when there's high demand. The incentive there is that drivers may come online if they see the prices are up, and customers who don't actually need to travel at that particular time will wait.

So in a sense, you are matching your supply to the demand through the dynamic pricing. But Uber got in a lot of trouble with dynamic pricing, or surge pricing as they call it, when it was engaging in that practice during time when there's critical events, so storms or cyclones or floods. So people really needed to use an Uber service to get to safety, and they found that the price had gone up exponentially.

So there was a lot of criticism of Uber during natural disasters for using surge pricing. And I believe, at least in the US, Uber has said they won't use surge pricing in times of natural disasters or at least they'll cap it. And that's a good example of a pricing strategy that had justifiable objectives. It was trying to match demand and supply but really lost social license when it was seen to be opportunistic by putting prices up really high in a time where people had no choice but to take public transport for safety reasons. So that's an example of that issue about know why you're doing it, but also be aware of public perceptions.

Garreth Hanley:
Are there best practices for implementing dynamic pricing effectively? Or are there common mistakes that people should avoid?

Jeannie Marie Paterson:
The Uber example is a good example of the common mistakes, which is that from a supplier's point of view, as I've said, dynamic pricing can make a lot of sense assuming that the levers have been well thought through. But the public perception of dynamic pricing and the way at which it's explained to the public are also really important. There are fairness considerations, if you like. Is your dynamic pricing strategy transparent? And if it is, if you explained it to a customer, are they likely to be surprised or outraged? Because customers lose trust in traders that seem to be taking advantage of them.

A good example of this is the furor that's arisen over Live Nation. So Live Nation is a concert organiser, so it controls ticket sales, it controls venues, and it acts for artists. Now, recently in Australia and in the UK, it's been suggested that Live Nation in its ticket sales through its ticket arm is using dynamic pricing. So people get in the queue to buy tickets for, say, Green Day and find that as they closer to the front of the queue the ticket prices are going up. It's been suggested that that's an example of dynamic pricing, that the number of people in the queue trying to buy tickets has caused the ticket price to rise.

Now, there's a bit of dispute between the ticket seller and the artists and the organisers about whose decision that was, but the public are outraged because they didn't know when they got in the queue that the prices are going to change according to demand. And then the regulator has got involved, the ACCC, and indeed the government has got involved saying, "This is unfair."

People have been subject to what's called in legal terms sometimes ‘unfair surprise’, in some way mislead or trick customers. And so now that practice is subject to scrutiny to see if it should be regulated. Why is that practice problematic? Because it's murky. People don't know that it was going to happen and they don't know why it's happening, and they don't know who's benefiting from that happening. And in the end, it just feels like price gouging to customers.

Garreth Hanley:
Is this a side effect of the technologies being used and the way technology is changing?

Jeannie Marie Paterson:
Absolutely. What technology has done is people can write pricing algorithms now that change the price in real time. So they can change the price of which tickets will be sold while people are waiting in line.

Uber can change the price according to how many people are seeking Ubers. So dynamic pricing used to be based on predictions about demand that were seasonal or based on historic patterns, and they're kind of predictable. We know hotel prices go up in school holidays. Now dynamic pricing is based on algorithms, so it can change in real time. And also, the processing of data, so we can use machine learning to make predictions that are quite nuanced about demand.

So we can get signals from the market faster, we can change prices faster, and we can also use data to make predictions about what people are likely to want to pay, who's likely to want to pay something and so on. So the whole process has sped up. But that means it's more opaque. People don't really understand how it's happening. That's why the Live Nation is a good example because the price is rumoured to have changed while people were sitting in line to buy tickets, which is closer to Uber surge pricing than it is to seasonal fluctuations in hotels.

Garreth Hanley:
We’ve done some recent podcasts on artificial intelligence, and a lot of our guests are saying that accountants and finance professionals need to still own the numbers, they can't just let the machines generate the numbers and that there needs to be a human eye over what's happening. Is that where people are going wrong with dynamic pricing as well, do you think?

Jeannie Marie Paterson:
I think it is, but I think there's also potential risk that haven't been uncovered. I think if you talk about that sort of dynamic pricing where you are increasing the price in real time as people are waiting for a product, that's problematic, but it can go a step further. So there's a fine line between dynamic pricing, pricing changing according to demand, and personalised pricing. Personalised pricing is where the price is actually set according to either the attributes of individuals or the attributes of cohorts of individuals or segments of your customer base. Now, there can be advantages in making price sensitive to the attributes of the customer base, and that's essentially how insurance works.

And it's also how credit works because often the interest rate for credit, the amount at which you'll be loaned money, is based on the risk that the customer presents. So the more you know about the customer or the cohort in which the customer sits, the more accurately the risk associated with the price of credit can be coordinated. And it's the same with insurance. The more you understand about the risk that individual customers or cohorts of customers present, the more the insurance product can be priced. So having more data about customers can help with that nuanced pricing, that personalised pricing.

And people can benefit from that, right? They can benefit from their circumstances being understood to give them a price that's really appropriate to who they are and the risks they present to either an insurer or credit provider or other providers of services, and that's all well and good. But the person or the industry or the trader or the firm that's using personalised pricing needs to be sure that it's accurate and that they know what factors are being taken into account. Because if they don't, they risk being found not to have engaged in personalised pricing tailored to risk, but discriminatory pricing.

And that's the big fear with using machine learning or predictive technologies to set pricing, that it might be accurate, but it might not be accurate. And worse still, it might be discriminatory or otherwise unacceptable. So there has been suggestions, for example, that Amazon, who's famous for using dynamic pricing, it will change prices according to what competitors are doing, for example, very quickly, but Amazon has also been found to experiment with the price that it will charge for particular products sold to particular customers. Now, Amazon has said that when it does that it's just doing effectively A/B testing, it's testing the market at what price customers will pay for particular products.

But there is a suggestion that perhaps some of that dynamic pricing is a bit more opportunistic, that they charge people more who are likely to pay more and people less who are a bit more discerning, a bit more savvy in the market. And it's sometimes been suggested airline pricing works this way too, that customers who are predicted to not to be price sensitive, often older customers or less tech-savvy customers, will be charged a higher price where younger tech-savvy customers will be charged a lower price because they're more likely to shop around. Now, businesses might choose to do that because they think they're making the best use of their insights from the market.

But if you sat down to try and explain that in the pub, "I'm going to charge you more because you don't understand technology and I'm going to charge you less because you do," most customers, they're going to go, "Wait a minute, I'm not sure that's fair." And that's really the danger here, that you're going to either offend your customer base because they're going to go, "This just isn't fair," or worse still, you're going to offend the regulator who's going to say, "You've misled people" or "You are discriminating against people."

Because I've used an example of age discrimination, but it could be socioeconomic background, it could be where you live, it could be your racial background, it could be your religious beliefs. Now, all of those things are protected attributes, you're not allowed to discriminate against people on protected attributes like race, age, religious beliefs, marital status, and so on. So very quickly, personalised pricing can become discriminatory pricing if it's not scrutinised properly.

Garreth Hanley:
It sounds like there's an ethical as well as a legal component in this and the line is something that people would need to really make sure they're not crossing.

Jeannie Marie Paterson:
But you can't determine whether you're crossing the line unless you understand how your pricing mechanism is working. So if you don't understand the basis on which prices are being set, there's no opportunity to say, "Well, do we have social license to do this? Are we breaking the law in this?"

That transparency piece that you mentioned, understanding the basis for the pricing mechanism is really important so that a business doesn't step from actually offering customers a better product into potentially offending, that's the unfair surprise point, or discriminating against customers.

Garreth Hanley:
Are there things that businesses can do to ensure that they are ethical and being fair for their customers?

Jeannie Marie Paterson:
Well, this is where we delve into the whole world of AI, that once we start using predictive technologies, there's lots of opportunities to better understand the market, better understand your own business, but there's also lots of opportunities to offend and to harm. And therefore, all businesses need to be AI literate. They need to understand where they're using artificial intelligence or, in the old-fashioned terms, predictive technologies or machine learning or data-driven insights.

There's a lot of different phrases and the language changes, but firms need to know where they're using those technologies, how they're using those technologies, and need to have governance mechanisms in place to scrutinise the outputs. So Amazon got in trouble for offering different prices to different people in the market. More recently in Australia, an insurance company was in trouble with ASIC because their pricing mechanism worked that they would offer a loyalty discount for people who continued to ensure with the business, but not all customers got the loyalty discount.

Some of them had their price increased by the amount of the loyalty discount that was given with one hand and taken away with the other. Now, there is a suggestion, and it still needs to play out in court, but there is a suggestion that they used an algorithm to decide which customers were price sensitive and which customers were likely to scrutinise the increase of their premiums. And customers that weren't likely to scrutinise an increase in the premium were offered the loyalty discount but didn't get it because their premium went up by the offered discount. Now, that is misleading conduct.

There's no nice way to skip around it. You can't offer someone a discount and then not provide the discount through a backhanded pricing mechanism. So there's ethical reasons, the outrage of surge pricing, and legal reasons, not misleading your customers, why firms using algorithms to set prices need to have a process in place checking that the algorithm is doing what they think it's doing. Because algorithms don't always get it right. They're simply processing the data that's available, they're not necessarily attuned to the objectives and needs of the business. So governance mechanisms are critical, and generally what they involve is auditing inputs and auditing outputs, looking for patterns of pricing that might indicate it's not working as expected.

Garreth Hanley:
So people really need to test their systems thoroughly before they let them loose on the world.

Jeannie Marie Paterson:
Well, they need to test them before they let loose, but they also need to recognise there's this thing called data drift, that the algorithm or the pricing mechanism may work for a while, but as the demographic it's being used on changes or perhaps it was never suitable for the demographic it's being used on, it can degrade over time. So something that works quite well initially can degrade. So it's actually an ongoing process that businesses need to engage in.

You can't be set and forget with something that's supposedly taking signals for a market and producing outputs, because it's simply a technology that's using statistical processes to make a prediction. If your market changes, then those statistical predictions won't necessarily work anymore. So that ongoing discretion is important. "The algorithm made me do it," is not going to be an excuse for getting pricing mechanisms wrong. That insurance example I mentioned earlier is a good example of that. So it's simply being engaged, I think. I think dynamic pricing or personalised pricing is beneficial.

It allows businesses to respond to their customer base, allows businesses to really explore new segments of the market, refine the products they're offering. And it's not necessarily bad for a price mechanism to nudge consumers into different patterns of behaviour, that's how we manage traffic flows on the freeway, manage the use of electricity, manage Uber drivers turning up at the right place. Those are all good things, but there are moments where it's going to cross over into unacceptable conduct. And sometimes businesses go, "Oh, it's really hard to work out when that occurs."

Actually, it's not. It's, would your customer base feel that they were being exploited or there was price gouging or would they understand the explanation? If you can explain it to someone in simple words, then it's probably going to be okay. If you can't explain it to someone in relatively simple words, then perhaps it's not, and perhaps you don't understand it yourself. And if what's being done is different to what the market expects, probably you need greater transparency because you don't want that gap between expectation and reality. Customer base doesn't like it, it makes them lose confidence, and the regulator doesn't like it. Whatever the regulator is in the particular market does not like the customer base being tricked.

Garreth Hanley:
And there's such a danger of the feedback mechanism when the data's feeding back in on itself.

Jeannie Marie Paterson:
That's exactly right. If we make a prediction about how the market works, then that's what the output's going to be, and it's going to re-enforce itself.

Garreth Hanley:
What's the future of this technology and dynamic pricing? If you had a crystal ball, what do you think is going to happen? Are there going to be regulatory issues for people? Is the technology going to continue to get better, or is it going to need to be slowed down?

Jeannie Marie Paterson:
I think it's going to continue to get better, but I think there's a limit to how far it can be used. So one of the things that we are sometimes sold on is, and we are still in the personalised pricing place, is that more data must produce a better outcome. So the more we collect, the more we are going to know about people and the more accurate our pricing will be. There's a limit to how much gain I think comes from more data. There's a point at which you might be collecting more data, but it might not be getting more nuance, it might not be giving more insights, and it might actually be losing what social trust or social license or whatever you want to call it. And I'll give you an example.

So, there's been some suggestion in the US that credit pricing, but also housing pricing, has been nuanced to information about customers, including their social media feeds. Now, yes, you want to get the right tenant, and yes, you want the price of the properties offered to the tenant or the loan offered to the borrower to be accurate relative to risk and willingness to pay. But we might ask whether knowing about a person's social media feed or who their friends are on Facebook is really a great way of setting prices. There's lots of people who will tell you, "Well, actually who someone's friends are and what they on social media tells you a lot about a person and their capacity for risk and so on and so on."

But you want to be really sure of that because otherwise people feel like they're being spied on and that there's just a data grab for the profitability of the business or indeed a data grab to discriminate against certain people according to their social media feeds. Neither of those are particularly useful or palatable outcomes. I'd really ask whether the gains outweigh the loss of trust and confidence that comes from businesses trying to use social media feeds or lots of sneakily obtained data to set prices. There's got to be a limit to where it's just not really giving any more information and it's just making customers uncomfortable.

Garreth Hanley:
So good quality inputs is really important.

Jeannie Marie Paterson:
Yeah, and not sort of jumping on the latest fad of, "Oh, well, if we collect more information, we're going to have a more nuanced price." As I've said, there's going to be a limit to that. Think about what price signals are useful for dynamic pricing and what are just noise and complicating the matter and make it more difficult to explain how the pricing mechanism works. Because if it's too complicated, it's going to go wrong.

Garreth Hanley:
You need to be able to explain it.

Jeannie Marie Paterson:
You need to be able to explain it, yeah. And change it. Because actually, if you go back to the banking Royal Commission, one of the outrages in the banking Royal Commission was fees for no service. And that was a legacy system, it wasn't a dynamic pricing system. It was a legacy system that the bank was deducting fees from certain people listed as customers and nobody had ever gone in and changed that mechanism when customers stopped being customers.

But it's also a sign that if your mechanisms for pricing get too complicated so that left hand doesn't know what right hand's doing, errors may result. Dynamic pricing is useful. It's useful for keeping up the market. It's useful for nudging or pushing demand in certain ways. But if it becomes too complicated, then it may fail to perform its function and instead may just result in unlawful activities.

Garreth Hanley:
You need that human oversight.

Jeannie Marie Paterson:
You need that human oversight, human on or in or under the loop, whatever the popular term is.

Garreth Hanley:
Yeah. Well, Jeannie, thank you so much for joining us. It was really good chatting to you.

Jeannie Marie Paterson:
It was nice to talk to you too.

Garreth Hanley:
Thank you. If you've enjoyed this episode, help others discover INTHEBLACK by leaving us a review and sharing this episode with colleagues, clients, or anyone else interested in leadership, strategy, and business. To find out more about our other podcasts, check out the show notes for this episode. And we hope you can join us again next time for another episode of INTHEBLACK.