Smart analytics is like making coffee
Doing smart data analysis is like making coffee — matching good product with the right extraction method ensures that the end product best satisfies specific needs. In analytics, as with coffee, there is no one size fits all solution for every occasion. Sometimes you want a cup of French press to sip leisurely, sometimes you need an espresso before running out the door, or you may need a big pot of coffee for many guests after dinner. So, what are the keys to brewing smart analytical solutions?
First, know your data. Just as it is important to understand the sourcing, roasting, and flavor notes of beans to make good coffee, so understanding our data is essential to maximizing the insights we can extract through smart analyses.
If collecting the data, make sure that survey questions are clear and search taxonomies are well defined. If using data sourced from a client or obtained from a third-party, make sure to understand how it was collected, how it was coded, and any sources of systematic bias or missing data.
Second, be familiar with your statistical toolkit. Many of us have different ways to make coffee. Similarly, our statistical training, professional setting, and type of client work differs, so we are likely to have different statistical toolkits. The key is to develop a statistical toolkit that best meets needs efficiently. It is better to know how to use the tools you have, and not constantly chase after the next flashy statistical software or much-hyped statistical solution. With analytics as with coffee, newer is not always better. Instead, it is about understanding the capabilities and limitations of each method or apparatus to optimize extraction of insights from data or deliciousness from coffee.
I can make café quality lattes with my espresso machine because I use it daily, but would make terrible shots of espresso on the Moka pot or Aeropress because I rarely use them. In the same way, it is essential that we understand the statistical tools used most frequently. Learn the default settings in the software package, practice explaining model assumptions to non-technical audiences, templatize standard data processing and clean-up required for the models to run smoothly, and make checklists for post-estimation robustness checks and reporting guidelines to facilitate good interpretation of findings.
Third, learn from past experience to know when to use what statistical tool. From experience, I know that overly oily beans perform poorly in my espresso machine and that lighter roasted beans require more time in the French press for flavors to bloom. I also know that it is impractical to make Americano’s one by one after a large dinner, so it is more efficient to use the drip coffeemaker or French press to make coffee in bulk.
In the same way, develop your own set of analytical and research best practices. Learn and refine the most effective analytical approach to use based on experience across different research settings (early stage exploratory analysis or testing of hypotheses with final data), varying data availability, and familiarity or lack thereof with statistical science of diverse audiences. Through intentional learning from previous projects, we become more effective in solving novel problems and wise in leveraging smart solutions to meet new challenges.
Inside each box of Intelligentsia Coffee is a reminder that “great coffee is not the result of chance.” In the same way, smart analyses do not occur when we have whimsical procedures or get lucky. Instead, we execute smart analysis well when we take the time to know our data, be familiar with our statistical toolkit, and learn when to use what statistical approach. While the execution of data analyses is a science, knowing how to do smart analysis is an art. And just as with coffee, with practice, intentional learning, and experience, we can improve in our craft and produce wonderfully satisfying outcomes that enrich mind, body, and soul.