Three stories about Artificially Intelligent pricing. First, how it is affecting your business today. Second, a history of the revolution in pricing. Third, an unlikely hero. Come and learn how AI pricing is disrupting how we sell online now.
Speech by Thierry Arrondo | Managing Director, Vendo
Thierry Arrondo Speech Transcript
Okay, hello everybody, thank you for staying.
I’m here today to explain 3 brief stories. All of them are on artificial intelligence. And how artificial intelligence, applied to price, is changing the way we sell and the way we do business.
So, I work for Vendo.
Vendo is a processing company, a biller. Thank you for staying, I know that’s not very attractive. Thank you for staying.
I have 3 very interesting stories, very short stories to explain that will drive you through these 3.
So the first one, I will present what is the problem we’re trying to solve.
The second one we’ll go through a time journey about pricing. How was pricing 500 years ago, how pricing is today? I’ve got 10 minutes, it’s gonna be very short, don’t worry.
And the last one I’m gonna introduce you to an unlikely hero in this story.
Fact: Fixed Prices Destroy Value
Okay, let’s start with the first story. And, I’m gonna start with a fact. Fixed prices destroy value. Who agrees with me? Nobody, okay. I’m gonna try to explain that.
Take me as an example. I live in Barcelona, I’m gonna spend 2 days here with all of you at this amazing conference. And I also spend the weekends in Costa Brava. Costa Brava is a place 100 km in the north of Barcelona.
Me, as a consumer, I am different in every single context that you put me. So, I will buy a beer tonight at a disco and I will pay $10 for the beer.
That probably is a lot for beer but let’s assume this disco is expensive and I pay the $10.
If I’m at Costa Brava in a nice bar in front of the sea, I would probably pay $5/$6 or Euros. And, in Barcelona, where I live at a grocery store, I will pay for that same beer, maybe it’s not German, but for the same beer, I will try paid €2.
Same Product, Same Consumer, Different Context, Different Prices
Same product, same consumer, different context, different prices.
So, if you treat me exactly the same way everywhere, you’re probably gonna lose money. You’re probably gonna make less revenue.
If you treat me the same, probably one of these 3 things will happen.
The first one is that you will price high and I will not buy. It will be too expensive. I’m not gonna buy a beer for €10 at the grocery store, that’s not going to happen.
The second option is that you will price low and yes I will buy. But you guys, advertisers, and affiliates will make less money because you could have a price higher and I would be willing to pay more.
And the third option is that you offer me exactly the price that I was willing to pay. But that’s 1/3 of the product is or it may be even less.
So, I’m gonna assume that now that we’re all in agreement. And using fixed prices hurts your revenue.
Myth: Fixed Prices Are A Myth
Let’s go for the second statement which is a myth. Fixed prices are a myth.
What are the fixed prices?
Fixed prices are the idea that a product can cost the same. Can that be real?
Well, let’s think about us. We’re all here today, some of you guys came from the States, and some of you guys came from somewhere else.
Even if you took the same flight to come here, even if you guys are staying at the same hotel, probably none of you will pay the same amount of money for the service for being here today.
Let’s go to another example that will bring things to the ground.
So let’s, with our imagination, fly to Manhattan. Let’s fly to Central Park.
And, I want you guys to join me in a walk from the southeast corner of Central Park, where the Apple store is, nearby the Plaza Hotel, we’re going to walk 4km to the north, okay?
We will start at the Apple store at this postal code the average income is $180,000.
Then, we’ll start working to the north and 4km away, only 4km, half an hour walk, the average income is $28,000.
$90,000 difference in income. Do you think the prices, the services that you’re going to find in these 4km are going to be priced the same?
The beer that I had in Costa Brava will have a price in the south, another one in the north. The hotels will be different. Of course, real estate will be totally different.
The kindergarten, if you want to keep your son or daughter somewhere, will be totally different.
So clearly fixed prices are not real.
Change: We Must Change The Way We Work
Now if fixed prices are not real and using fixed prices are hurting your revenue, why are affiliates and advertising, you are using fixed prices and the same offers to sell your products?
That doesn’t make sense. So that brings me to the third point.
We have to change the way we work. Given the current reality, as President Obama will say, we have to stop doing stupid shit.
Why? Because we will lose money if we don’t change the way we do it.
Let’s go to the second story now. And this is about pricing revolution.
And I’m going to start with the definition of pricing.
So pricing is the process whereby a business sets the price at which it will sell its products and services. The source, Wikipedia. We all trust Wikipedia.
Now why a company will do that? Why a company will develop, will invest efforts in setting different prices for their products depending on the context?
First Rule Of Marketing: Know Your Customer
Well, probably to meet the first rule of marketing. And the first rule of marketing says that you have to know your customer.
If you know your customer, you will be able to segment your customers. And probably, you will be able to sell better.
So the objective of selling is, sell more and make more money. Source, Thierry Arrondo.
Well, that’s my presentation, I’m not sure that’s reliable, that’s what it is.
Time Journey: Medieval
Okay, now let’s go into a time journey. And then, I want you guys to think about how people were selling in a medieval market.
We’re going to go 600 years ago. And, we’re gonna think about a dusty medieval market. It’s the afternoon, you have a seller that has his stall, and he’s selling fruit. There are no prices because probably he can’t write.
And you guys probably can’t read. One of you would be approaching, he will be looking at you. You will be wearing nice clothes, you’d probably come from the castle, you’re probably rich.
He remembers you because you came last week and took his best product, his best oranges.
The seller will say, “Hey welcome back, what do you want today.”
And the seller is already thinking about what is the best price and product that can I offer you to make more money.
He’s trying to increase the conversion ratio and he’s trying to increase the revenue that he will make in that transaction.
There will be another person coming from the village, dressed in his clothes. Maybe that one will not get the best offers but will get fruits that are 2 or 3 days old and at a decent price.
Again, what we want to do is maximise the amount of money and maximise the conversion ratio on a one-to-one process.
Okay for many, many, many years, this is the way we were selling.
Time Journey: Modern
So let’s jump to now 100 years ago, industrial area.
We have here the production line of the Ford T Car. That car that you could get in any colour as long as it was black. A car manufactured by Ford.
In that area, what industrials like Henry Ford and others did, was to focus on cost.
It was a cost gain. The idea was to be able to produce as much as possible, controlling the cost. And then they will fix the price, the same price for everybody.
The same price for everybody and they will manage their profit by the number of cars they will sell.
There is a better way to do that. Although, I agree that Henry Ford did a good job.
Time Journey: Pricing Today
Okay, let’s jump to today, what pricing looks today. And to do that, I’m gonna put an example.
So this graph shows the price of a general electric oven or microwave oven that one customer saw throughout one day in 3 different sites.
Amazon, Best Buy and Sears.
In Amazon, he saw the price changing 9 times. And in Best Buy he saw the price changing 2 times. At Sears, that big return that we all want to be, they never changed the price.
There was no human interaction in those price changes. It was all automatic.
Okay, how does that work? How can it happen? What are the tools that we need to make that happen?
To be able to offer the different prices for each product, at each time.
This is the vision, this is knowing your customer to a point that every individual, every single customer that will hit your systems, will hit your site, will go through your affiliate system, will be unique.
And, you will be able to target him and to offer him what will make him buy.
Well, there are 3 pillars.
The first one is artificial intelligence.
Why? Because the context that you need, all the data that you need, is totally beyond human capabilities.
If we think about a small site receiving 100,000 visitors a day that means that you will have to make a decision every second.
Every second you guys are going to have to make a decision about what is the price that I’m going to have to show to that end user to be able to make him buy.
There’s no human, there’s no group of analysts that can do that manually. It’s impossible.
You need a machine to take care of that.
Then on top of that, ideally you need a system. You need algorithms that will bring feedback loop that will learn, that will understand if a context has changed to be able to provide different prices to different contexts.
So, in the first attempt to do that we used in the past, A/B testing. And probably most of you are familiar with A/B tests.
I want to see among 3 different experiences, what is the best one for me or for the end-user.
I’m gonna split my traffic in 3 and I’m going to wait until I have statistical significance in the results. And then once I find the winner, I will send all my traffic there.
That way of doing things is not really efficient because the exploration phase is pretty long. And meanwhile, you’re losing time.
Is there a better way to do that? And yes, there is a better way.
The Bandit Problem
And you may also be familiar with this algorithm, quite popular through an analyst in Google called Steven Scott that he publishes about the bandit problem. Named after the casino slot machines.
What is the bandit problem? Imagine that you’re in a casino that you have 20 different machines and you want to know which one, where to put the coin to get more rewards, more money.
Well, the idea is you can do an A/B test and you can split your money and see which machines will give you more money to a point where you have a confidence level.
If you’re testing a drug for cancer, you want a very high confidence level 99% to make sure you don’t kill people.
And if you’re testing different experiences or different prices may be a 70% is enough.
So A/B test, the alternative is for every coin that you will play on the machine, you’re going to wait to see what the result is. And for every result, you’re going to calculate the probability of that machine to be better than the others.
With that information, you’re going to redistribute the traffic. So, if one machine has a higher probability, you’re going to send more traffic to that machine.
Okay, what do we need for artificial intelligence and this testing work? What do we need?
You Need Time, You Need System To Learn
We need time, you need the system to learn. So you need time, you need data. The more data you get, the better. Because systems need data for the exploration phase. Then find the winner and do the exploitation phase.
Ideally, you want to set a KPI. With the example of the slot machines, if they give you the same reward, conversion ratio is enough.
But if the machines give you different rewards, then you will also have to take into account the revenue.
And in that case, what we usually use will be the earnings per play or the earning per click. So, you will take into account the 2 aspects. How much I convert and how much money I’m making.
And, all this is pretty complex. So you need the system to be in place to do that. You need the time, you need the machines.
But this is the only way really to get additional results from the alternative to using pricing.
The Unlikely Hero
Okay now, let me go to the third story called the unlikely hero.
Let me introduce you to Bill. Bill is short for Billing, his surname is Processor.
And, he’s not very popular. People do not like Bill. He’s been doing the same for years.
So basically, he takes credit cards from end users on behalf of merchants. And the process the transaction to a bank and he gets money.
It’s a commodity. Nobody likes him, and nobody invites him to parties. And I know that because I’m one of them.
Okay, what makes Bill interesting in this story?
Well, Bill is interested in this story because he has access to all the data that you need for artificial intelligence. And not only that, he has access to this data in the format that you need.
Billing And Pricing
So, Bill has access to millions of customers. Bill has access to hundreds of advertisers, of merchants. And he can compile all these data in the same format.
And an analyst will tell you that 80% of the hard work in analysing data is putting the data in the right format.
So in that case, Bill, by using the right tools, artificial intelligence, machine learning, becomes a cool guy.
He’s not anymore that boring guy that is a cost centre. He becomes a profit centre.
How do we do that?
Billing Is A Data Business
Well, Bill is here. Every time an end user hits his system, he knows where he’s from with the IP, what time of the day it is, the device that he’s using. And with all this data we can segment people, in real time.
And then by segmenting the people, we will be able to show the right price to each user.
This is one result for one test, a real test, this is from an advertiser.
The first phase shows an exploration phase, where the system is learning. As you can see sometimes you lose money.
This shows the lift against a baseline. We define a baseline, where we don’t touch anything. And then, we calculate what the lift of the system is compared to this baseline.
So, you have to be patient. Usually, people freak out at the beginning, “Oh, I’m losing money, this system is not working. We have to shut it down, please, please, please.”
But we have to be patient, we have to let the system work, we have to let the system learn. And once it learns, you will get a revenue lift.
The revenue lift that was seen on average is 2 digits, 12% in this case.
Some of you, if you’re an affiliate, you will ask, “Well, why is this affecting me. I work on conversion based, so I don’t really care if the price is high or not.”
Well, that’s not true because there’s an unbreakable law of economics that caps the revenue that you can make and caps the revenue that an advertiser can make.
You will never make more money than what you can get from with the end-user, from the consumer.
So if you have techniques to increase that, independently of your business model and your relationship with the advertisers, you’re gonna be able to get more money.
Pricing: Past and Present
Last slide, let me show you how pricing looks for us in the future, in the past, present.
I will probably not be surprised by that but we’re seeing, going back to the medieval market.
But with billions of users hitting your pages and with the right tools, artificial intelligence, machine learning make much more money.
Leveraging the first rule of marketing. You will know your customer, you will know what price, and you will know what offer you have to show him to make more money.
Thank you very much.
Hope you enjoyed. And, I hope you enjoy the rest of the show.