Translating Data into Action: The Data/Action Tripod
Understanding the core strategies of data usage: reporting, exploration, and prediction.
Richard Li•
#data#exploration#business intelligence
You’re a director at a medium-sized company. The executive team has spent the past few months working on the strategic plan for the year, and now it’s time for the unveiling at the all-hands meeting.
With great fanfare, your CEO stands up to the podium. DATA-POWERED AGILITY flashes on the screen. The CEO proclaims that this will be the year of data. The room visibly deflates. One by one, each executive follows the CEO, talking about agility, data, and analytics.
Afterwards, your colleague turns to you and whispers, “Did you understand what we’re supposed to actually do?” 🤯
In this post, we’ll look at what it actually means for organizations to be data-driven, and examine three types of strategies for using data: reporting, exploration, and prediction.
Today, organizations accumulate vast amounts of data. However, this raw data is challenging to act upon directly. To transform this data into actionable insights, organizations typically employ three general strategies: reporting, exploration, and prediction.
Reporting involves the systematic analysis and presentation of data to track performance, monitor trends, and make informed decisions. It provides a snapshot of past and current data, offering a foundation for understanding the organization’s status quo.
Exploration, on the other hand, delves deeper into the data to uncover hidden patterns, relationships, and trends. It enables organizations to ask new questions and gain a deeper understanding of their data, leading to more informed decisions and strategic initiatives.
Prediction leverages statistical algorithms and machine learning techniques to forecast future trends and outcomes based on historical data. By predicting future scenarios, organizations can proactively plan and allocate resources to achieve their goals.
You can visualize how these strategies work together in the Data/Action tripod:
“How many monthly active users do I have, and what’s the trend?”
Business metrics such as revenue, users, activation, and customer count are the lifeblood of every business. Thus, accurate and timely reporting of these metrics are the foundation of every data initiative. Reporting brings three major benefits:
Business intelligence tools such as Tableau, Metabase, and Looker are typically used to organize data into dashboards, which are then reviewed on a regular cadence.
“What types of user behavior signal purchase intent?”
Once a data reporting operation is in place, people naturally ask, “What else does the data tell us?” This is where data exploration comes into play. Data exploration (also known as “Exploratory Data Analysis”) involves delving deeper into data to uncover new insights and patterns that may not be immediately apparent. Common use cases for data exploration include improving customer experience (e.g., identifying that customers who use a certain feature tend to spend more money) and more targeted marketing (e.g., discovering that customers in the EMEA region have higher click through rates for product X compared to product Y).
One of the key benefits of data exploration is that it can spark new ideas that are backed with some empirical evidence that the new ideas generated will have a real-world, quantifiable impact.
The primary mistake organizations make in data exploration is adopting a reporting mindset. When approaching data exploration with this mindset, they often rely on their business intelligence (BI) tools to provide answers to specific questions. This results in dozens (or hundreds!) of charts that are used for research purposes. However, the essence of data exploration lies in the fact that the question being asked may not always be clear; the goal is to make sense of the data as a whole. This necessitates a more flexible and iterative approach to analysis.
Instead of reporting-oriented tools, tools such as spreadsheets, notebooks (e.g., open source Jupyter notebooks and commercial options such as Hex & Mode), and techniques such as regression analysis) are better suited to data exploration. These tools allow you to manipulate and visualize data in a more exploratory manner, helping you uncover valuable insights that may have otherwise gone unnoticed.
“What is the sales forecast for the next quarter?”
The final strategy in leveraging data is using it for predicting future results. This involves using historical data and statistical algorithms to forecast future trends and outcomes. Common use cases for prediction include sales forecasting, customer churn forecasting, and predicting customer lifetime value (LTV).
One key benefit of prediction is the ability to make adjustments before trends or issues are reflected in traditional reporting. For example, if churn is forecasted to be high in the coming months, organizations can take proactive measures, such as increasing customer engagement efforts or offering targeted promotions, to mitigate the impact.
In many cases, simple heuristics can be effective for prediction. For instance, using a rule of thumb like “25% of the weighted pipeline for the quarter will close in the next quarter” can provide a useful estimate. Spreadsheets are a powerful and versatile tool for these types of general-purpose predictions.
In some cases, there is value in investing in more sophisticated prediction models. For example, Uber has invested in forecasting across a brad range of domains to optimize driver allocation, marketing spend, and hardware capacity. These types of models typically are staffed with data scientists who can build, test, and maintain these models.
Effectively translating data into action employs three strategies: reporting, exploration, and prediction. Each strategy serves a distinct purpose and offers unique benefits. Reporting provides a snapshot of past and current data, aiding in tracking performance and monitoring trends. Exploration enables the discovery of hidden patterns and insights, driving innovation and problem-solving. Prediction allows organizations to anticipate future outcomes, facilitating proactive decision-making.
Tactically, when faced with a question that requires data-driven insights, it’s crucial to identify which strategy is appropriate and to use the right tools for that strategy. This ensures that the data is analyzed and interpreted in a manner that aligns with the desired outcome.
Strategically, organizations should consider how they will support all three strategies in the Data/Action tripod as they adopt their data architecture. This involves ensuring that the necessary data is collected, stored, and made accessible for reporting, exploration, and prediction. By implementing a holistic approach to data utilization, organizations can unlock the full potential of their data and drive continuous improvement and innovation.