Businesses spend significant time analyzing historical performance, such as sales, revenue, and profits—for good reason. Understanding past performance is essential for tracking progress over time. It's also the foundation for forecasting future trends, helping organizations anticipate demand, allocate resources, and make more informed decisions.
Advances in artificial intelligence (AI) and machine learning have made predictive modeling faster, easier to scale, and more accessible. Tasks that once required teams of data scientists and months of analysis—such as forecasting demand or customer behavior—can now often be completed more efficiently.
This guide defines predictive modeling and how AI supports it. It also explores several predictive modeling techniques used for forecasting that business leaders should understand.
Featured Certificate Course
Data Science and AI for Decision Making
Learn how leaders can use AI and data science to improve decision-making, innovation, and business performance.
Start Anytime | 4 Modules | 5-6 Hours per Module | $1,949
What Is Predictive Modeling?
Predictive modeling is a data science technique in which historical data is analyzed to predict future outcomes. It involves building statistical or machine learning models trained on historical data to learn relationships among variables. Once trained, these models can analyze new data and estimate the likelihood of different outcomes.
“Machine learning can cluster data into meaningful groups to build predictive models that classify or forecast outcomes,” says Harvard Business School Professor Karim Lakhani in the online course Data Science and AI for Decision Making, which he co-teaches alongside HBS Professor Iavor Bojinov. “It’s based on a common structure: you give the model labeled or unlabeled data, and it learns patterns that help it sort, group, or predict something specific.”
In business, predictive modeling helps organizations anticipate future outcomes so they can make more informed decisions.
“Forecasting uses data and algorithms to estimate values that haven’t happened yet—but will unfold over time,” Bojinov says in Data Science and AI for Decision Making. “Whether you're projecting revenue for next quarter, anticipating demand for a product, or estimating how many users will log in tomorrow, forecasting helps organizations prepare for what’s likely to come.”
Predictive Modeling Techniques for Forecasting
As you plan a predictive modeling initiative, it’s important to choose a technique that aligns with your data and business objectives. While no single technique is right for every forecasting project, some are better suited to specific use cases than others.
“There are many forecasting algorithms, each suited to different kinds of data and goals,” Bojinov says in Data Science and AI for Decision Making. “Some are simple and easy to interpret; others are designed to capture more complex, dynamic patterns. Choosing the right forecasting approach depends on the structure of your data, the decisions your forecast will inform, and the level of accuracy, flexibility, or transparency you need.”
Time Series Models
Time series models are a type of machine learning algorithm designed to analyze values that are ordered sequentially over time. They're used to forecast future values based on previous values in a sequence.
At their simplest, time series models identify trends over time—whether a metric is increasing, decreasing, or remaining stable—and forecast what may happen next. Given enough historical data, they can also be used to identify longer-term cycles and seasonal trends.
Sample use case: Predicting product demand and inventory requirements.
Regression Models
Regression models estimate the relationship between variables to predict how changes in one or more variables may affect a target outcome. For example, you might use regression analysis to understand how marketing spend influences revenue and then forecast the impact of increasing or decreasing your marketing budget.
“In regression analysis, we’re modeling how changes in one or more inputs relate to changes in an outcome,” Bojinov says in Data Science and AI for Decision Making. “This can involve a single input—simple regression—or multiple inputs—multiple regression. Either way, the goal is the same: understand the relationships in the data, and how much confidence to place in the results.”
Sample use case: Forecasting the return on investment (ROI) of increasing or decreasing marketing spend.
Decision Trees
Decision trees are defined in Data Science and AI for Decision Making as models that use a series of if-then decision rules to work through a problem. Combining multiple decision trees creates an ensemble model, such as a random forest, which can improve predictive performance by capturing more complex relationships in the data.
Because their decision rules are transparent and repeatable, decision trees are well-suited to applications that require auditability.
“The model considers different ways to split the data and chooses the cutoff that best separates the outcomes,” Lakhani says in Data Science and AI for Decision Making. “The model keeps splitting and refining until it reaches a stopping point. These rule-based splits form an interpretable model—one that shows not just what the prediction is, but why the model makes it.”
Sample use case: Predicting the likelihood of customer churn based on changes in spending habits.
Neural Networks
A neural network is a computational model inspired by how neurons communicate in the human brain. It consists of an input layer, an output layer, and one or more hidden layers that process information between them. This architecture enables neural networks to model complex, nonlinear relationships within data and solve highly sophisticated prediction problems. It also underpins many large language models (LLMs) and other forms of generative AI.
“A neural network takes in data, processes it through layers of weighted connections, and ‘learns’ by adjusting those weights over time to improve its accuracy,” Lakhani says in Data Science and AI for Decision Making. “For instance, it can learn to recognize that a certain combination of pixels represents a cat and not a dog.”
Sample use case: Forecasting energy demands and grid load requirements for a utility company.
How Does Predictive Modeling Work for Forecasting?
Predictive modeling follows many of the same steps as other data science and machine learning initiatives. Forecasting projects introduce several unique considerations, however, as outlined below.
1. Define Your Business Objective
The first step in any predictive modeling project is to define your business objective. What question are you trying to answer? What outcome do you need to predict? How will your forecast support broader organizational goals—and does this contribution justify the time and resources required?
2. Collect the Data
Once you’ve defined your objective, you must gather the data to train your model. The data sources you leverage will depend on your use case and the outcome you’re trying to predict.
Potential data sources include your customer relationship management (CRM) system, transaction records, usage data, communications channels, and third-party data providers. Any system that captures data about your customers, products, or operations can provide valuable insights.
3. Complete Exploratory Data Analysis
After collecting your data, the next step is exploratory data analysis (EDA). This process helps you understand your data's quality and underlying structure.
EDA can reveal opportunities to clean, transform, or enrich your dataset to improve forecast accuracy. It also helps identify the target variable you want to forecast and the input variables most likely to influence it.
4. Prepare the Data
The next step is preparing your data for modeling. This typically involves cleaning your data, standardizing formats, and splitting it into training, validation, and test datasets.
For forecasting projects, it’s especially important to remove unrelated variables and eliminate outliers that could skew your results.
5. Train the Model
Once your data has been prepared, it's time to train your model. This involves selecting the algorithm that best fits your forecasting objective and training it on your historical data.
Before deploying the model, you must confirm its performance using validation and testing datasets to ensure it can generate reliable forecasts for new data.
6. Monitor and Fine-Tune the Model
As with any machine learning project, training is never a “one-and-done” process. It’s important to monitor your model's performance over time to ensure it continues to generate reliable forecasts as data patterns change.
Retraining the model with new data can also help maintain its accuracy and relevance.
Free Course Flowchart
Which HBS Online Digital Transformation & AI Program Is Right for You?
This flowchart will help you determine which HBS Online digital transformation and AI course best aligns with your career level and experience, learning objectives, and educational goals.
An Enriched Decision-Making Process
The real power of predictive modeling lies in its ability to surface hidden trends and patterns within your historical data. These insights support more informed, data-driven forecasting that can influence everything from budgeting to resource allocation.
To maximize the value of predictive modeling, it’s important to understand both the data requirements and the practical considerations involved in training and deploying forecasting models. Completing an online course like Data Science and AI for Decision Making is one way to build the knowledge and skills to apply these techniques effectively.
Do you want to improve your ability to leverage data and artificial intelligence in your business? Explore Data Science and AI for Decision Making—one of our online digital transformation and AI courses—or download our free flowchart to choose the right course for your goals.