Set a goal and get expertise
Step 1 in the data science lifecycle is to define a business goal by using subject matter expertise. Having a clear goal as you begin your data analysis is essential. In the garden example, the goal is to produce as much lettuce as quickly as possible. However, growing conditions aren't absolute. If your baseline production is 1 pound of lettuce in 14 days, you need a way to speed up production.
You've heard that talking to plants might make them grow faster. Although it seems possible that sound affects plant growth, the likelihood of sound affecting your outcome might be too small to be worth considering. You decide to experiment. You find that if you play classical music at 50 decibels for 30 minutes every 3 hours, you get the same pound of lettuce in 13.5 days. In the end, adding a music schedule to your garden is a complex solution that probably isn't worth the hassle to shorten growing time by half a day.
The garden example demonstrated that having access to people who are subject matter experts (SMEs) in factors that affect your goal is important, so you don't modify variables that won't yield major changes.
If you use the same lens to look at rocket launches and weather patterns, you see a two-part goal:
- To increase the likelihood that the day chosen for launch will be a good weather day.
- To know what conditions should stop a launch.
The expertise that's needed to reach those goals is found in the contributions of meteorologists, physicists, biologists, rocket scientists, and many others. SMEs help by scoping factors that potentially affect a launch and so require critical attention, thereby minimizing the number of variables to analyze.
For example, SMEs might determine that the amount of direct sunlight on the launch pad might not make a difference in the successful outcome of a launch, but that the amount of moisture in the air does. They might also know that there are ranges of important data that can be disregarded. For example, a launch site temperature of less than 30 degrees fahrenheit rules out consideration of all other factors. There are no mitigating factors when it's too cold to safely launch a rocket.