Can You Really Predict the Future? Lessons from Data, Causality, and Forecasting

How can we accurately predict the future? This article explores the relationship between description, explanation, and prediction using a case study of an online‑education company’s monthly revenue.

1. Description: Presenting Patterns

First, description is the initial step. By collecting and analysing data we can reveal underlying patterns. The monthly data (ad spend, user growth, revenue) from January to December 2023 show a steady upward trend.

The relationship can be approximated with a simple linear regression:

Revenue = intercept + slope × month + ε

The fitted line matches the observed data closely, predicting an average monthly increase of about 30.5 k.

2. Explanation: Revealing Causality

Explanation seeks to answer “why”. While the regression shows correlation, it does not explain why revenue grows. Possible causal factors include increasing market demand, brand awareness, advertising intensity, and seasonal effects such as exam cycles. Consumer‑behavior theory links brand perception and purchase decisions, suggesting that higher brand recognition drives more registrations.

3. Prediction: Exploring the Unknown

Prediction goes beyond simple extrapolation; it must consider uncertainty. The linear model assumes continued demand growth, stable advertising effectiveness, and no major external shocks. Scenario and sensitivity analyses are recommended to evaluate how competition, policy changes, or other shocks could affect future revenue.

Thus, description, explanation, and prediction represent three logical layers of scientific inquiry: description uncovers data patterns, explanation uncovers causal mechanisms, and prediction attempts to forecast future outcomes while acknowledging real‑world uncertainties and the principle of falsifiability.