Though you might not have realised it if you live in England, summer is here. Once upon a time, this meant kids would take to the sidewalks to hawk their non-alcoholic beverages. Other than providing their pint-size proprietors with supplementary income, lemonade stands are a helpful microcosm when considering the industrial applications of Artificial Intelligence (AI). While there is a lot of talk about potential use cases, what is this like in practice?
Let us consider two rival lemonade stands. On one side of the street we have Brute Force Billy and on the other side we have Analytical Ana Maria. Welcome to the clash of the cul-de-sac! We will assume their goals are the same, namely they are both looking to maximise their profits over the summer months. While their older siblings helped them build their lemonade stands, they still need to buy fresh lemons and sugar themselves every day. They muddle those with (free) water to create lemonade.
To understand the difference in their businesses, it is helpful to start with the key operational questions our intrepid entrepreneurs face. In this scenario we will limit ourselves to three, although we could probably list tens of questions! For one, how many lemons and sugar do they need to buy every morning? Secondly, how should they price their lemonade; should they sell at different price points? Finally, what products can they add to their stalls' inventories to increase their overall revenues?
Every morning, Brute Force Billy goes to the store and he buys 25 lemons and 500 grams of sugar. This makes roughly 10 liters, or 40 glasses of lemonade. Whatever the weather or the day of the week, Billy buys the same amount of ingredients and makes the same amount of lemonade. He sells this lemonade for £1 per glass, and makes a tidy £30 every day, except for when it is cold and no one wants lemonade. He sells Polo mints on the side because he likes Polo mints and he assumes others do too.
Next, we have Analytical Ana Maria. Ana Maria surveyed her neighbours and built a support vector machine regression with a polynomial kernel to predict the likely number of sales using the day of the week and the weather. Every morning her model tells her how many ingredients to buy. She keeps some extra ingredients in stock so that she can sell more lemonade should demand exceed her model's estimate. She does however keep track of her model's performance to improve its predictive power.
Secondly, Ana Maria is a savvy salesperson and has built a dynamic pricing model using a random forest. When she sees a customer approaching she uses the model to infer whether the customer is most likely to want to purchase a small glass or a large glass, and whether they are more likely to pay £1 or £5 for their drink. After all, we are talking artisanal cold-pressed Sicilian citrus juice here! While we can debate the data ethics of this profiling, the reality is that our hypothetical huckster is making a mint.
Finally, what other products should she sell alongside her lemonade? Again AI is helpful here. Ana Maria use natural language processing and some clever Python to automatically identify which sweets are being discussed — with positive sentiment – on her neighbourhood's Discord channel. This way she can ensure that her lemonade stand only stocks the latest and greatest sweets. It is no surprise that lately her sales of sweets have outpaced her lemonade sales, given their appeal whatever the weather.
As you can see, even though a lemonade stand is about as simple as an organisation can be, there are numerous possible applications of artificial intelligence. These types of decisions, regarding how many materials to procure, how to price inventory, how to package offerings, and what products to recommend crop up on a daily basis in almost any type of organisation. Business leaders would therefore do well to ask themselves: what can I learn from a lemonade stand, and how can I avoid being a Billy*?
* If your name is Billy, I apologise. You can be a Billy all you like.