Five steps for making data-informed decisions

As a data scientist, using data to make decisions – and to help other people make decisions – is an important part of my work.

Data can be really helpful in the decision-making process to provide context and reduce uncertainty. Figuring out the nuts and bolts of data-informed decision-making – including how and when to use data, as well as what data to use – can be tricky and unintuitive, but it’s an important skill and one that can be learned.

To help you get started, I’m going to share the five-step process that I follow to make better decisions with data:

Step 1: Define the decision to be made

It sounds simple, but before you think about turning to data to make a decision, it is important to specifically define the decision you’d like to make.

Defining the decision (including your goals, the available choices, and possible outcomes) upfront provides clarity that is useful for determining how data can help. A clear definition of the decision can also serve as a reference to save time and deliberation when you’re ready to start looking at data.

You’ll know that you’ve properly defined your decision when you can answer these questions:

• What is the decision you’d like to make?
• What are your possible choices or options?
• What is your goal in making this decision? What does a ‘good’ outcome look like? Alternatively, are there any outcomes that are unacceptable or must be avoided?

By listing out your goal and important positive or negative outcomes, it should be clear what decision you’ll be making and what you’ll be optimizing for or against.

Step 2: Define any additional relevant information and constraints

Most decisions aren’t made on data alone. The best and most practical decisions are made by combining data with additional information and context, including timing and business constraints, so it’s important to consider these ahead of time as well.

Knowing your constraints will also help you to determine what data will be most valuable to you.

I find it helpful to list out any decision-related constraints ahead of time so that I can consider them while I’m trying to figure out what data I need (and again when making the decision).

These questions can help to reveal relevant ‘decision context’:

• Do you have a decision deadline? What happens if you don’t meet it?
• Is your decision reversible? 
• What are your other practical constraints? (List them out!)

Once you’ve defined your decision along with additional context, it’s time to define what data would be most helpful for making that decision.

Step 3: Determine the data needed to make a decision

The trickiest part of making data-informed decisions, in my opinion, is figuring out exactly what data would be most helpful for a given decision. This boils down to a translation exercise: using the decision you’d like to make and given your constraints, what information will help you to reduce uncertainty and make a good decision?

Thinking in hypotheticals is helpful for determining what data could be useful. Imagine the following scenarios, for example:

• Is there data that would unequivocally tip the scale toward any given option (assuming your constraints aligned)? 
• If you’re leaning in one direction, what data would change your mind and make you choose an alternative?

Before you ask for or collect data, consider how you’d use it, and whether it would make your decision more clear, keeping your goals and constraints in mind. Thinking about the types of insights that would reduce uncertainty and make your decision easier can help guide you toward the data you should ask for.

It usually takes me a few rounds of hypothetically ‘auditioning’ data – ‘if I knew this, would it change things?’ – to figure out the right data to help with a decision. When you know what data would materially matter to your decision-making process, you’re ready to collect it.

One more thing: before you set out to collect data, consider narrowing down the number of metrics or pieces needed and defining their importance ahead of time. If you don’t know what data is most important and you end up with metrics that point in opposite options without a clear way to decide which is better or more important, it can complicate your decision-making process rather than simplify it.

Step 4: Collect and interpret data

Unless you’re collecting the data yourself (kudos!), you’ll likely need to enlist help from another person or team. If you aren’t well-versed on what relevant data is available, bringing a data SME into the process early can help to reduce iteration. 

My best advice when enlisting data help is to share as much context as possible when you’re describing the data you need. If you can share the decision pre-work that you’ve done, and especially details about the decision you want to make and your specific goals, do so.  Anyone pulling or analyzing data on your behalf will make decisions while carrying out their work, and the more context they have, the better those decisions will be – and the better the data will fit your needs.

You may have questions as you start to interpret the data. Sometimes the data you had in mind might not be available or match what you expected. Partnering with a data SME here can help to ensure that you end up with data that will help you to achieve your goals.

Step 5: Make your decision

Once you have the data you need, you’re ready to make your decision. Here’s where the work that you’ve done ahead of time pays off: your data, when combined with the additional context and constraints you’ve outlined, should reduce uncertainty and make your decision more clear.

And voila, you’ve made a data-informed decision!

Reflections

While this process takes investment, it pays off as you learn more about your data over time and become more efficient at determining how data might help your decisions. Ultimately, following this process will result in better, faster, more data-informed decision-making.