Companies have adopted analytics faster than the Road Runner outrunning Wile E. Coyote. But choosing the right questions to ask analysts isn't as relaxing as watching a Saturday morning cartoon. Any given set of data can raise a considerable number of analytics and data exploration questions. Gathering meaningful responses to questions therefore becomes daunting.
Over my years consulting on analytics I have met clients who were truly afraid of asking questions, unsure if they had the right words or understanding of what to ask. They're unclear what tasks have been done or remain to be done. So how do you find the right questions to keep your team moving forward?
To get to the right questions, you need the right framework for the dimensions and metrics represented in your analysis. You also need a good framework for the questions related to those dimensions and metrics. Let's look at the details to create both frameworks.
Identify the Dimensions and Metrics You Want to Talk About
Dimensions are features or items you want to measure, while metrics are the measurements associated with those features. The dimensions and metrics should relate to the kinds of questions your team is seeking answers for. For example, a digital marketing team would compare conversion rates (metrics) of different referral traffic sources (dimensions) to answer the question of how customers are discovering a website.
In instances when data is imported into a programming language, such as R, Python or SQL, you will likely select dimensions and metrics based on how you arrange the data. An object table in which queried data is placed would treat each column as variables for a statistical analysis — which is usually describing metrics — and each row of the table as dimensions. So if you were creating a table in your program related to cars, the columns would contain vehicle specifications (metrics) such as length, width, miles per gallon, anything that would be of interest. The rows would contain the types of cars or identification of a specific set of cars you would want to analyze.
Once the dimensions and metrics are mapped out, data acquisition and exploration tasks become easier to select. You can ask yourself if there will be a need to transpose data, add column for calculated metrics, or join tables together.
Initial ideas for data visualization can be better anticipated, too. Let's say that location data is included in our car dataset example. You could consider if a bubble map that has two dimensions of geolocation data with a metric would be a good visual for ongoing reporting.
Related Article: How Marketers Can Plan Data Mining With R Programming
Now Let's Start Planning for Open-Ended Questions
Any discussion, be it a casual chat or formal presentation, should include an open-ended question that relates to objectives from the data. You might ask, “Did you discover any opportunities to connect with customers online during our fall sale?" to see if a campaign was effective.
This isn't to say you can never ask a close-ended question. Sometimes time only permits a yes or no answer. But a binary, yes-or-no response glosses over the reasons why trends and deviations develop. The response can also overlook the technical details of querying a database for advanced models, answering concerns on what it takes to gather and update data. Responses to good open-ended questions usually include key details.
For example, open-ended responses to digital marketing reports note trends and deviations among the metrics of interest. Trends are a general increase or decrease, like those seen in marketing data. You can also look at data distribution to understand any specific movements in the campaign. Deviations are a sharp change from a standard behavior of data, like a spike that alters a trend or an outlier in a data distribution.
Related Article: Stop Torturing Your Data and Other Tips to Reveal True Data Insights
You Can Better Define Discussions that Move Your Work Forward
It may take a few back and forth conversations to reveal all of the details behind an answer. But open-ended questions focused on reported dimensions and metrics streamline your efforts for generating timely and thoughtful insights. The responses will initiate ideas to segment next steps into immediate and long-term tasks.
Within the context of digital marketing analytics, an immediate task can be adjusting a digital ad campaign to save a budget, while a long-term task would involve a strategy, such as attracting an audience of a given demographic to download an app and use its features. Within the context of data models, you would have responses to the data being explored that may or may not trigger immediate tasks. That would depend on how the data was acquired — is it easy to access from the original source or does it require extensive tech support set up to acquire it? The long-term task may involve developing the model, especially for a machine learning model, which usually takes a long time to produce results.
Many of these immediate and long-term tasks frame how to manage other operational needs that can arise from analytics discussions — I outline a few in this post on communication. How you approach the topic of dimensions and metrics can also enhance your soft skills, which I also explain here.
Overall, the best open-ended questions associated with dimensions and metrics should raise curiosity and give you confidence to work on next steps. You'll avoid burning out from chasing every minute detail in your data. Maybe you'll even spend more time relaxing ... at least, until the next analysis.