21st century consumers demand more personal service than ever before. The ability of companies to provide this high level of personalized service, is highly dependent on their ability to predict the customer’s demand on an individual level. Since consumers get more individualistic, the variety in demand increases, adding a lot of complexity to the estimation. Moreover, markets become increasingly transparent, which increases the effect that price has on the demand of your product. An accurate demand estimation enables you to optimize for example your services, pricing and acquisition planning. Given this dynamic market environments, how can we predict future demand?
The three stages of demand estimation
Demand prediction proceeds in three stages. In the first stage there is no estimation. One does not bother about predictions at all. Commercial decisions are made based on experience and gut feeling. In the second stage, traditional demand estimation, you predict the total amount of sales. This helps you in determining your stocks and the focus of your investments. However, with data science it is possible to make demand predictions on individual customer level. In this stage, individual demand estimation, you do not only know the total amount of sales, but also understand the type of customers that have that particular demand. In this article I will show you how you can get to the third stage of demand estimation. I will start with a simple model, and will increase the complexity step by step. Throughout this article, I will use the demand of flight tickets as my case.
Let’s start with an example of traditional demand estimation. Traditional demand estimation helps you to forecast future demand by identifying patterns in your sales data. For instance, it might be that there is a seasonality or trend in the demand. Flight tickets may have an increased demand in the summer, because of the summer holidays (seasonality). Moreover, the demand might increase overtime due to a growing population and the increase in popularity of far destinations (trends). If we project these seasonality and trends on the future, we will get a basic understanding of what we can expect of our future demand.
Traditional demand estimation
The previous model is still very general and can be improved, because different kinds of customers have different interests, different behaviour and thus different behavioural patterns. For example: families travel more in the summer time, students in holiday periods, and businessmen throughout the year whenever there is business to be done. By knowing who your customers are you can improve your demand estimation, and make distinctions between different customer profiles. In the graph below, I first show the average demand for one year and then show the average demand for different types of customers that jointly represent the total demand. Can you imagine how knowing these differences can improve your service?
Individual demand estimation
An interesting puzzle isn’t it? However, we are not finished yet. Since only working with trends and seasonality (even on an individual level) assumes that nothing changes in the way we market our products. But what will happen if you change the price of your flight tickets? Or the way or timing that you promote your service? Or what kind of other products you have in addition to flight tickets? Of course we could improve model above by adding more influencing variables.
Suppose that we decrease the price of the flight tickets to Berlin from €110,- to €90,- , what will the effect be on the demand? Again, we do not calculate this effect for the whole population, but on an individual level. Since for a student price might be a more influencing variable than for business men. Being aware of this differences influences the pricing decisions you make. For example, due to the decrease in price, the demand from students will probably increase heavily, while demand from business men stays almost constant. This could be explained by the fact that business have to go to a certain location at a certain point in time to do business and price often is not the major variable where they base their decisions on. However, the decision to lower the price influences your turnover and profits in that period. So it is important to know who you are targeting before you are able to estimate the effects on your results.
However, this model already adds a lot of value to your business, it still can be improved. For example, will decreasing your price with another €20,- (to €70,-) have the same effect? Given the transparent nature of the flight ticket market (think of comparison websites such as cheaptickets.nl), changes in demand due to pricing, are heavily correlated with the prices of competing suppliers. Decreasing your price probably won’t be as effective if you already are the cheapest supplier in the market. Imagine that the cheapest flight tickets in the market costs €100,- and your price is €110,-. In this case we might expect that decreasing your price to €90,- gives a larger increase in demand (which student is not choosing the cheapest ticket?) than dropping your price with an extra €20. By knowing this it will influence the way you price your products, don’t you think?
You can imagine that predicting demand can be an interesting game, with a lot of different models, variables and data sources coming together! For now, I have used flight tickets and the influence of price changes as example case, but the same game is played in a lot of different industries with many different influencing variables. Such as recommending products, the way you promote your products et cetera. With data science it is possible make accurate demand predictions, so you know up front what the effect of a change in price or change in promotion strategy will have on your business results like profit or turnover.
I hope that this article explained the value of an accurate demand estimation model, and that you now have an idea how it can improve your offers and services!