# Why Data Modeling Is Not Enough for Business Success Shift focus from predictive data models to decision-centric models for better ROI. - **Source:** [MIT Sloan Management Review](https://sloanreview.mit.edu) - **Published:** 2026-10-19 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/37a303ad-e994-4687-b794-81281b2fe9a1 ## The Data Science Paradox Many organizations invest millions in data science, yet 60% to 85% of projects fail to deliver tangible results. While teams build highly accurate predictive models, these often lack actionable insights for management. The core issue is a disconnect between technical accuracy metrics and the actual business decisions leaders must make daily. ## Defining the Communication Gap Data scientists typically focus on prediction and machine learning metrics, while managers prioritize cost, profit, and market share. This linguistic divide causes elegant models to sit unused. Experts note that projects fail primarily because both parties do not understand the real business problem, leading to solutions that solve irrelevant technical challenges. ## Decision Science vs Data Science Decision science models the choices managers face rather than just the data itself. It relies on data science for inputs like predictions but adds logic, mathematics, and domain knowledge to frame problems. This approach ensures that analytical outputs are directly tied to business outcomes, transforming raw predictions into clear, actionable management strategies. ## The Role of Problem Framing Effective decision scientists act like doctors, diagnosing the root cause behind managerial symptoms. Before building any model, they must frame the problem correctly. For instance, forecasting demand is useless if it does not inform specific inventory production levels. Proper framing ensures the model addresses the actual constraints and business rules involved. ## Avoiding Convenient Data Traps Teams often use easily accessible data instead of identifying what is actually needed to solve a problem. In customer churn scenarios, predicting who will leave is insufficient. Managers need to know which customers are worth retaining and what actions work. Modeling the decision first prevents reliance on convenient but ultimately useless data sets. ## Aligning Metrics with Financial Goals Statistical metrics like goodness of fit rarely interest executives. Decision modeling translates predictive inputs into financial consequences, such as profit or risk. By capturing the asymmetry of outcomes—such as the differing costs of approving a bad loan versus rejecting a good one—models become relevant tools for high-stakes financial decision-making processes. ## Implementing Decision-First Briefs Organizations should replace initial data requests with formal decision briefs. Similar to Amazon’s working backwards process, this requires defining the decision-maker, alternatives, constraints, and success metrics upfront. This discipline prevents technical teams from building impressive but irrelevant models, ensuring resources are only committed to projects with clear, measurable business value and impact. ## Lessons from HanesBrands Ben Martin, VP of data at HanesBrands, mandates that no project proceeds without answering critical questions about the business problem, the required decisions, and the anticipated value. His guiding principle is that only decision analytics drive ROI. This rigorous process forces alignment between technical teams and leadership before any coding begins. ## Building an Enterprise Digital Twin Bipin Chadha at CSAA Insurance Group created an enterprise digital twin to bridge the gap between data science and strategy. This simulation serves as a learning lab, allowing data scientists to understand business rules and test the impact of their models. It transforms theoretical data science into practical, business-focused decision support. ## The Necessity of Domain Knowledge Technical proficiency in machine learning is insufficient without deep business domain knowledge. Data scientists must learn the operational realities of their company to provide value. By embedding analysts within strategy teams or using simulation labs, organizations ensure that models reflect the complex, real-world trade-offs that managers face in their daily operations. ## Optimization Over Mere Prediction Moving beyond prediction, decision models often use optimization and simulation to handle uncertainty. Instead of providing a single forecast, they offer alternative recommendations based on risk-versus-reward trade-offs. This allows managers to choose plans that maximize profit while adhering to capacity, material, and labor constraints, turning data into a strategic competitive advantage. ## Final Strategic Recommendation To maximize ROI, leaders must shift their focus from modeling data to modeling decisions. By fostering communication, prioritizing problem framing, and aligning metrics with financial outcomes, companies can ensure their analytics investments deliver real value. Data science provides the inputs, but decision science provides the path to profitable, evidence-based organizational action. ## Sources covering this story - https://sloanreview.mit.edu/article/why-data-modeling-is-not-enough/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._