Beyond the Mantra: Embracing Correlation in Decision Making

Beyond the Mantra: Embracing Correlation in Decision Making

The phrase "correlation does not equal causation" has become a common refrain in the world of data analysis. While of course it's accurate, it's often used to disarm analysts and bring analyses to a screeching halt. This article explores a more pragmatic approach, focusing on when correlation is enough and how an overemphasis on causation can lead to paralysis in decision-making.

The Showstopper: "Correlation Does Not Equal Causation"

When stakeholders utter this phrase, it's often code for "Your insights aren't good enough for me to act on." This popular belief has led decision-makers to think they need causal insights to make decisions with data. In practice, this requirement isn't always reasonable. Consider A/B testing, where the requirement for causality often results in flawed implementations and eventually, in delayed, gut-based decisions.

The Downside to Causality

Causality is an impractical requirement when making decisions with data. Proving causality requires a higher level of statistical rigor and a lot of carefully collected data, leading to extreme delays in decision-making. The reality is that causality is very difficult to prove, and the pursuit of it can be more of a hindrance than a help.

A Heuristic Approach to Using Correlations

Rather than dismissing correlations, we can use them in a business context to maximize our chances of making the "best" decision. Here's how:

1. Be Intentional When Testing Correlations

• Avoid Random Correlations: Focus on correlating things that are already connected, such as customer actions.
• Avoid "P-hacking": Don't search for surprising correlations that are likely due to chance.

2. Correlate Conversion Rates, Not Totals

• Understand Customer Behavior and Always Use Conversions: Conversion rates provide insights into customer willingness to convert, while totals may be influenced by other factors, or may be misleading.

3. Ensure Trends Are Consistent Over Time

• Analyze Data Over Time: Look at how data has been changing over time to ensure that correlations are consistent and reliable.

4. Always Monitor the Results, and Adjust

• Track the Impact: Monitor the results of decisions made based on correlations and be ready to adjust as needed.
• Learn from Mistakes: If a decision based on correlation doesn't pan out, analyze what went wrong and use it as a learning opportunity.
• Minimize Risks: By acting quickly on correlated findings and vigilantly monitoring the results, we can minimize the risk of wrong decisions.

Conclusion: Embracing Correlations

In practice, we need accurate insights, and we need to act fast. Waiting for analyses that claim causality or lengthy A/B tests is not always feasible. By embracing correlations and using them wisely, we can make better decisions, faster.

For startups, where agility and responsiveness are paramount, correlations can be a valuable tool for decision-making. By understanding their limitations, being intentional in their use, implementing rigorous testing, adopting a pragmatic approach, fostering data literacy, and continuously monitoring and adjusting, startups can use correlations wisely.

This approach allows startups to navigate the complex landscape of business development with confidence and efficiency, turning data into actionable insights without being paralyzed by the demand for causality. It's a nuanced balance that empowers startups to thrive in a competitive environment, making data-driven decisions that resonate with their unique goals and challenges.

??Let's retire the phrase "correlation does not equal causation" and recognize the value of correlations in decision-making. By following best practices and being mindful of their limitations, we can turn correlations into a powerful tool for data-driven insights.