Using Artificial Intelligence for Price Discrimination
There was an article published by the Scientific American earlier this year titled Will Democracy Survive Big Data and Artificial Intelligence?, and it’s probably worth a read if you have 45 minutes (yes, it’s that long). We gave up about halfway through when we came across the following specimen which CB Insights affectionately refers to as “bad data viz”:
Before throwing up that monstrosity, the article talks about how all your big data exhaust is being used to manipulate you into buying more stuff. Because the article is written by zee Germans, it manages to not spew forth some political ideology every 7 seconds, but instead puts forth some interesting ideas that should both alarm and amaze, regardless of which echo chamber you happen to be living in. The article opens with a few very powerful sentences:
The amount of data we produce doubles every year. In other words: in 2016 we produced as much data as in the entire history of humankind through 2015. It is estimated that in 10 years’ time there will be 150 billion networked measuring sensors, 20 times more than people on Earth. Then, the amount of data will double every 12 hours.
That should blow your mind. Think about how much data you generate yourself, today. That will depend on where you live of course, the types of technology devices that accompany you, and what you willingly disclose. Some of what you willingly disclose may not be so obvious.
Extreme Ad Personalization
In a past article we remarked that it’s not just the paranoid who think their smartphones are listening to them. Many smartphones are sitting next to your side waiting for you to say “Ok Google”, which means they are listening to you. Why do companies want to hear what you’re saying? It’s because what you say, what you think, what you type into search engines, all that is used to determine which ads to stick in front of your face that will result in you parting with some of your cash. We are now subjected to extreme levels of ad personalization and we all seem to be just fine with that. But what about things like price personalization?
Before we delve into this topic, we need to clear the air by defining some of the terminology. Here it goes:
- Price discrimination – primarily used in the world of economics to describe a “situation where identical or largely similar goods or services are transacted at different prices by the same provider in different markets” (link)
There are three main types (or degrees) of price discrimination:
- Price personalization – also called first-degree price differentiation, this is where you are selling to each customer at a different price based on information specific to that unique customer (link)
- Product versioning – also called second-degree price differentiation, this is where you create slightly different versions of the same product at different price points (link)
- Group pricing – also called third-degree price differentiation, this is where you price things differently based on groups of buyers (link)
Then there is one more term we’ll use as follows:
- Dynamic pricing – the price changes according to variables that are specifically NOT customer-related (link)
In common parlance, the use of the word “discrimination” will have negative connotations, usually as a result of the criteria being used to determine how the price is personalized. As you would expect, maximum outrage will surface when any group is charged more, but nobody bats a long fake eyelash when “ladies drink free” on Wednesday nights. What everybody should be able to agree upon, is that certain types of price discrimination are fair. The last few seats on a flight should be auctioned off to the highest bidder. Intuitively that makes sense. But what about charging each person a different price for the same widget, in particular, high priced widgets like laptops?
We recently went to buy a Lenovo laptop from their website and completed the machine configuration all the way to the payments page. Over several days we continued to modify the configuration while looking at the device every single day, the same way most people would when mulling over a high-ticket item. After deciding against the purchase in favor of waiting until year-end, we left the page and didn’t return. An email follow-up 24 hours later offered a $100 off coupon if we purchased now. It was exactly what we needed to pull the trigger on that purchase and we did. Then we thought to ourselves, is that price discrimination? Would offering a $100 off coupon for hesitant cheapskates like ourselves be considered discrimination? Going back to the Scientific American article:
Personalized advertising and pricing cannot be compared to classical advertising or discount coupons, as the latter are non-specific and also do not invade our privacy with the goal to take advantage of our psychological weaknesses and knock out our critical thinking.
We’re not sure how weak-minded you would have to be for someone to knock out your critical thinking using big data, but point taken. Our psychological weakness is value, so we could suppose that the Lenovo coupon took advantage of that. But what if Lenovo knew based on our social media conversations, that we’ve historically been very critical of price when it comes to large purchases? As we can clearly see, everything comes down to what criteria is being used to determine each “personalized price”. The whole situation sounds like the perfect opportunity to introduce some artificial intelligence (AI) algorithms, so we decided to take a closer look at a startup using AI for “dynamic pricing”.
Perfect Price and Dynamic Pricing
Founded in 2013, San Francisco startup Perfect Price has taken in around $2 million in funding to “deliver price optimization as a service, leveraging machine learning expertise previously applied at Twitter, Microsoft, and Disney“. In order to understand who might benefit from such functionality, they present the following four questions to help determine eligibility:
- Is your demand predictable?
- Is your inventory perishable?
- Are your fixed or sunk costs significant compared to your variable costs?
- Does demand vary by time?
Does that last one ring any bells? Remember how the legions of Uber critics were hating on the world’s biggest startup because they were employing “price discrimination”? Well check this out:
If a business is able to make more money as a result of higher demand by servicing the highest bidders, there seems to be nothing wrong with that. Put yourself in the shoes of the business owner, and you’ll immediately lose any notion of “discrimination”. It’s no different from how those horrible airlines price seats for their flights. It all comes down to figuring out what is the maximum spend you can extract from a customer based on each unique “buying moment”. It’s called “behavior based pricing” and Perfect Price introduces the idea in their 47-page “Ultimate Guide to Pricing Strategy” which you can download here if you have the time. If not, here is the basic gist.
If you sell widgets online like Amazon does, you would intuitively look at the people who actually bought a product and try to increase sales based on what worked for successful purchases. According to Perfect Price, you would be “ignoring as much as 99.5% of the behavior data by only focusing on what’s selling”. Here’s the idea behind “behavior based pricing”:
If, instead of pricing everything on cost-plus-margin, you set prices based on observed willingness to pay, you’ll have a variety of unit margins in your store. But each one will have a purpose, derived from behaviors of your customers, and overall your contribution margin (and, possibly, revenue) will be higher. But retail is more complicated than that.
They use the example of the 2016 Porsche Cayenne. With prices ranging from the $58,300 base model to the $157,300 Turbo S version, where do the majority of sales happen for such a high price sports car? As you might expect, it’s right in the middle of the range. The top and bottom prices simply act as anchors to drive the customer to spend the maximum amount of money once they’ve decided to buy a Porsche. Their behavior throughout the entire buying process will telegraph whether or not this is someone who is price sensitive or not, then you can nudge them in an appropriate direction. This is a good example of “product versioning” or the “second-degree price differentiation” we talked about earlier.
Perfect Price is using artificial intelligence for dynamic pricing, and seems to be heavily focused on hotel rooms and car rentals where price discrimination should be expected. You can add airlines to that list as well, and anything else that has a limited supply of something you rent (yes, like an Uber). It’s such an interesting business model that we may come back to this topic and do a deep dive on the Perfect Price platform. Before we do that though, what we’re more interested in is the following.
AI for Perfect Price Discrimination
Are there any startups out there that are using AI to customize prices based on factors that are more dubious, say the customer’s social media feeds as an example, or past buying habits? Are there any startups out there pushing the boundaries when it comes to pricing widgets based on what the customer is willing to pay using all the big data exhaust available? Drop us a line and let’s have a chat.
Pure-play disruptive tech stocks are not only hard to find, but investing in them is risky business. That's why we created “The Nanalyze Disruptive Tech Portfolio Report,” which lists 20 disruptive tech stocks we love so much we’ve invested in them ourselves. Find out which tech stocks we love, like, and avoid in this special report, now available for all Nanalyze Premium annual subscribers.