Quick Answer: "Correlation does not imply causation" is the golden rule of data. Just because Ice Cream Sales and Shark Attacks both increase at the same time does not mean ice cream causes sharks to attack. Both are caused by a hidden third variable: Summer.
The Danger of Observational Data
With the rise of "Big Data," computers can find patterns in anything. Did you know that the divorce rate in Maine perfectly correlates (99.2%) with the per capita consumption of margarine? This is called a Spurious Correlation. It is entirely random. If you act on this data by banning margarine, you will accomplish nothing.
The Hidden Confounder
Often, two things are correlated because of a "Confounding Variable." For instance, a hospital finds that patients who sleep with a nightlight are more likely to be near-sighted. Should we ban nightlights? No. It turns out near-sighted parents are more likely to leave a nightlight on, and near-sightedness is genetic. The genetics caused both the nightlight and the vision issue.
How Do We Prove Causation?
The only reliable way to prove that X definitely causes Y is through a Randomized Controlled Trial. By using random assignment to split people into two groups, you mathematically destroy all hidden confounders. If the randomly assigned drug group gets better and the placebo group doesn't, you have proven causation.