One of our most popular blogs was about algorithmic pricing and how it had gone awry during a strike of the subway workers in London. Markets are driven by the law of supply and demand. Consumers communicate to the sellers through a product's demand curve. Your demand curve is a graph of the quantity of a good or service you are willing to purchase at any given price. When you do not purchase an item, you are telling the producer you do not want the good or service at that price. Sellers read the buyer via changes in their inventories. When inventories rise, it tells the seller that their price is too high. Sellers communicate via the supply curve, which is the number of goods or services the seller is willing to provide at a given price. When the price of a good or service is too low, the seller may choose to reallocate its resources. It is telling the consumer that it cannot make a profit at a given price. Communication between the buyer and seller has taken place for thousands of years, whether it has been haggling in a market, or the building up of inventories. Algorithmic pricing has been used for several years, using artificial intelligence. We have asked our British contributor to follow up on her popular blog.

Algorithmic Pricing – A Follow Up

If you read our previous blog on algorithmic pricing, you could be forgiven for thinking that it's a disaster waiting to happen – leave a computer in charge of setting rates and sooner or later your prices will be spiraling out of control. Or, even worse, that widespread use of algorithmic pricing will result in skyrocketing prices across an industry, as algorithms mine data from one another in order to effectively fix prices higher than supply or demand would dictate over an extended period.

That scenario isn't completely impossible – and it's not the intent, but as our previous blog illustrated, it can get out of control. Authorities are keen to keep an eye on the potential abuse as technology develops. For the time being though, algorithmic pricing is really only a big problem when it goes wrong; or when it's poorly designed and executed. Most of the time it works just fine, and when the algorithms used are well-tuned and finely-balanced there's actually very little difference between adjustments due to algorithmic pricing and the normal adjustments that are made to prices due to natural fluctuations in supply and demand. The adjustments just happen sooner and more frequently. Recently the models have become more sophisticated by introducing artificial intelligence (AI) into the programming.

In traditional supply and demand, of course, as the price of a good or service rises, the quantity demanded will fall. When demand falls, so too will the price as retailers or service providers drop their rates to increase business. The inverse is also true: when prices decrease, the quantity demanded for a product will increase as it becomes affordable to more consumers. This feedback system ensures that prices remain in line with the level of demand and supply that manufacturers and service providers are willing and able to produce and consumers are willing and able to buy.

A well-designed and balanced algorithm is, in fact, designed to do exactly this. The only difference is that it doesn't require the intervention of a human, and so can adjust prices many times a day in accordance with fluctuating demand and supply. Programs with AI aren't just reacting to consumer behaviour, they are predicting behaviour. Gas stations are a good example of algorithmic pricing done well – they adjust their prices frequently according to supply and demand, but the algorithms used are balanced and well-designed, and as such remain close to the market price. In the absence of computer algorithms to determine price, the same process would still occurandhellip; but it would be slower and more laborious, and not nearly as effective. For example, Sam Schechner reported in The Wall Street Journal that a gasoline station in Rotterdam using sophisticated algorithmic pricing with AI surprisingly raised its price when a competitor across the street lowered theirs. The reason: the software anticipated lines at the competitor's pumps. Models are also used to adjust the prices of related products. For example, the price of milk, which is sensitive to price changes, could be set less than competition, and the price of cereal, its complement may be increased.

As with any new technology, glitches are almost certain to occur as the use of algorithmic pricing becomes more and more common. It's possible too that unscrupulous individuals might misuse algorithms to fix prices by colluding with other companies. The excuse, "the computer set the price" is unacceptable. In 2015, the US Department of Justice prosecuted David Topkins for writing an algorithm that violated antitrust laws for price fixing (New Yorker). Artificial intelligence will make it more challenging to prove collusion. As the technology develops, however, the ability of authorities to monitor and intervene when this happens will increase. Ultimately algorithmic pricing remains a practice to be embraced: when it's used correctly, it represents an advance on the traditional methods of responding to supply and demand, and can be a positive tool for retailers and consumers alike.

To learn more about the laws of supply and demand visit any of our twelve free lessons related to supply and demand.