What is Prediction
Prediction is a technique that allows to predict future scenarios based on the analysis of past data. It is a systematic attempt to understand future performances of a business or the behavior of a system. Such operation implies a detailed analysis of past and present trends in order to predict future events. In other words, Prediction works as a planning tool that helps decision-makers to prepare for the uncertainty that the future may bring, by supporting their decision-making processes.
Over time, the complexity of problems that can be tackled with predictive models has grown exponentially. For this reason, new techniques of prediction have been implemented, and each case now requires a thorough reflection on which strategy suits it best.
The choice of the best prediction model to apply depends on several factors: context of the prediction, availability of past data, the period over which the prediction is needed and the relationship between costs and benefits, among others.
The definition of Prediction
Based on the competences in aHead Research we can define Prediction as follows: to make predictions is to identify all future events in advance through the mathematical analysis of what happened in the past and what is happening in the present. The forecasting tools can help businesses in exploiting all future opportunities and mitigate risks associated with their activities by facing future uncertainty and taking more informed decisions.
Prediction differs based on the time span to which they refer:
Short-term predictions: this kind of prediction is needed, for example, to manage resources, production, and transport, but can also guide the prediction of future demand.
Medium-term predictions: this kind of prediction usually concerns the purchase of raw materials, machinery, and equipment, but also hiring planning.
Long-term predictions: this type of prediction is employed in strategic planning, considering market opportunities and internal resources.
Prediction methods are based on availability -or lack- of historical data.
Quantitative analysis is performed whenever information is available and if it is reasonable to assume that aspects of the past will still be present in the future. There are now countless quantitative prediction models, created within distinct disciplines with specific goals. The majority of such models is based on time series -data gathered over time- or cross-sectional data -gathered in a specific moment.
Whenever data is not available, one can resort to qualitative prediction. It consists of well-structured models, developed to obtain high-quality predictions without historical data.
Prediction: practical examples
The use of prediction techniques is very popular in areas that do not necessarily interest business. Here are some examples.
The prediction of earthquakes is a branch of seismology that specifies time, location and magnitude of future, within-range earthquakes and more importantly defines the specifics of future, stronger earthquakes in a specific area. At the moment, earthquake prediction is based on inferential statistical models used to determine the probability of a given event in every scenario.
Research on earthquake prediction is based on empirical analysis, which depends on two strategies. The first consists in identifying the events that anticipate an earthquake, so all premonitory phenomena. The second entails the identification of a trend or a geophysical model that might precede a large earthquake.
Another application area of prediction is the electricity market.
The costs indicted on the electricity bill usually depend on an estimate, therefore a prediction, based on the analysis of historical data related to a given utility, intersected with average consumer data of the market in a given moment.
Electricity is a non-storable resource. For this reason, a balance between production and consumption is necessary. Moreover, demand for electricity depends on weather conditions -such as temperature- and daily activities of each individual -rush hours, workdays and holidays-. These factors create peculiar price dynamics, with daily, weekly, and yearly fluctuations that distinguish this market from all other markets.
In the sports area, too, we find prediction used to identify the athletes’ performances. An example in this sense, is De Bruyne, midfielder for Manchester City. For his contract renewal until 2025, the footballer based his requests on technical data analyses demonstrating the importance of his contributions to the team. The renewal was in. fact determined by statistical data that analyzed past performances and predicted future ones.
Prediction in the business world
In aHead Research, we often resort to prediction activities to help both managers in the organization of their businesses and clients in taking decision while reducing the degree of uncertainty. An example is the prediction of sales related to clothing items for actors of the clothing industry. This activity is just as important as it is delicate, because of the high degree of seasonality in the industry. In fact, trends interest colors, prints, styles, and materials and they tend to drastically vary from season to season. The prediction of sales is therefore based on big data that make allow for the identification of future trends, the reasons causing such trends and the evolution of customer behavior based on the analysis of past purchase habits. The adoption of a Demand Driven Supply Chain model in the fashion world is a necessary requirement to let data define production and sales strategies, therefore improving the prediction of future demand behaviors and adopting models that can guide production and avoid a significant waste of resources.
Another example of prediction is described in our case study on bid optimization in online advertising auctions, in which we explain how a leading company in the digital marketing sector improved its ROI of online advertising campaigns.
Finally, thanks to the large amount of data available, supply chain system can predict the sale of full priced products and products on sales. The advantages of predicting on sale concern both product restocking and logistics.
The data used to predict products on sale consists of demand patterns such as seasonality and sales variation or concern external factors such as competitor activity and weather conditions.