The Top 7 Forecasting Models for Project Managers

In project management, making informed decisions is crucial to achieving success. Accurate forecasts can enhance decision-making, optimise resource allocation and improve project outcomes.

In this article, we explore the top seven forecasting models every project manager should know, leveraging various forecasting methods and algorithms to predict future values based on historical data.

Table of contents

1. Linear regression

2. Moving average

3. Exponential smoothing

4. ARIMA (autoregressive integrated moving average)

5. Neural network models

6. Econometric models

7. Delphi method

1. Linear regression

Linear regression is a foundational quantitative forecasting model that predicts future values based on the relationship between one or more independent variables and a dependent variable. By analysing historical data, linear regression helps project managers anticipate future trends and make data-driven decisions.

Types of linear regression

The two most common types of linear regression are simple linear regression and multiple linear regression. Simple linear regression uses one independent variable to predict a dependent variable, while multiple linear regression involves two or more independent variables.

Applications in project management

Linear regression can be used in various aspects of project management:

• Predicting project costs: Estimating future project expenses based on historical spending patterns.
• Estimating project timelines: Forecasting the duration of future projects by analysing past project timelines.
• Resource allocation: Determining the optimal allocation of resources by predicting future resource needs.
• Risk assessment: Identifying potential risks by analysing the relationships between different project variables.

Steps to implement linear regression

Implementing linear regression involves several steps:

1. Data collection: Gather historical data relevant to the variables you want to analyse.
2. Preprocessing: Clean and prepare the data for analysis, addressing any missing values or outliers.
3. Model building: Use statistical software to build the regression model, specifying the independent and dependent variables.
4. Model evaluation: Assess the model’s performance using metrics such as mean absolute error (MAE), root mean squared error (RMSE), and R-squared.

Advantages and limitations

Linear regression offers several advantages and limitations:

• Advantages:
  • Easy to understand and implement.
  • Provides clear metrics for accuracy assessment.
  • Suitable for both short-term and long-term forecasting.
• Limitations:
  • Assumes a linear relationship between variables, which may not always be the case.
  • Sensitive to outliers, which can affect the accuracy of predictions.

Case study

Consider a project management team that needs to forecast the cost of upcoming projects. By applying linear regression to historical cost data and other relevant variables, they can create a model to predict future costs. This helps them allocate budgets more accurately and plan more effectively.

Tools and software

Several tools and software can assist with linear regression analysis, including:

• R: A programming language and software environment for statistical computing.
• Python: Widely used for data analysis, with libraries such as scikit-learn for implementing regression models.
• Excel: Offers built-in functions for performing linear regression analysis, making it accessible for many users.
• PM3: Integrates historical data analysis and resource forecasting to enhance project cost and timeline predictions.

2. Moving average

The moving average method smooths out short-term fluctuations and highlights longer-term trends or cycles. By averaging data points over a specified time period, it provides a clearer view of trends, making it particularly useful for sales forecasting and demand forecasting.

Types of moving averages

There are three main types of moving averages: simple moving average (SMA), weighted moving average (WMA), and exponential moving average (EMA). SMA assigns equal weight to all data points, WMA assigns more weight to recent data points, and EMA gives exponentially decreasing weights to older data points.

Applications in project management

Moving average can be applied in various aspects of project management:

• Analysing project performance trends against benchmarks: Smoothing out fluctuations to identify performance patterns.
• Predicting future project milestones: Estimating the completion dates of future milestones based on past data.
• Managing inventory levels: Forecasting future inventory needs by analysing historical consumption data.
• Budget planning: Smoothing budget data to identify spending trends and make future financial projections.

Steps to implement moving average

Implementing moving average involves several steps:

1. Data collection: Gather historical data relevant to the variables you want to analyse.
2. Smoothing process: Choose the type of moving average and the period length for averaging.
3. Calculation: Apply the moving average formula to the data points.
4. Interpretation: Analyse the smoothed data to identify trends and make forecasts.

Advantages and limitations

Moving average has several advantages and limitations:

• Advantages:
  • Reduces noise from random fluctuations in data.
  • Easy to implement and understand.
  • Essential for effective time series analysis.
• Limitations:
  • Can lag behind actual data trends.
  • Less effective with highly volatile data.
  • Does not account for seasonal variations unless adjusted.

Case study

Consider a project management team that needs to forecast future inventory levels. By applying a moving average to historical inventory data, they can smooth out irregularities and identify underlying trends. This allows them to predict future inventory needs more accurately, ensuring optimal stock levels and preventing shortages or excesses.

Tools and software

Several tools and software can assist with moving average calculations, including:

• Excel: Offers built-in functions for calculating different types of moving averages.
• Python: Provides libraries such as Pandas for implementing moving average calculations.
• R: A powerful tool for statistical analysis, including moving average computations.
• PM3: Provides built-in features for performance trend analysis and resource management, facilitating accurate moving average calculations.

3. Exponential smoothing

Exponential smoothing is a versatile time series model that weighs past observations differently, giving more importance to recent data points. This approach is effective for capturing trends and seasonal patterns, making it valuable for project managers.

Types of exponential smoothing

There are three main types of exponential smoothing: single exponential smoothing, double exponential smoothing (Holt’s method), and triple exponential smoothing (Holt-Winters method). Single exponential smoothing is used for data without trend or seasonality, double exponential smoothing adds components for trends, and triple exponential smoothing includes components for both trend and seasonality.

Applications in project management

Exponential smoothing can be applied in various aspects of project management:

• Forecasting project timelines: Predicting the duration of future projects based on past performance.
• Budgeting: Estimating future budget needs by analysing historical financial data.
• Demand planning: Forecasting future demand for resources or products to ensure timely availability.
• Workforce planning: Anticipating future staffing needs based on historical workload trends.

Steps to implement exponential smoothing

Implementing exponential smoothing involves several steps:

1. Data collection: Gather historical data relevant to the variables you want to analyse.
2. Parameter selection: Choose the smoothing parameters (alpha, beta, gamma) based on the type of exponential smoothing being used.
3. Model building: Apply the exponential smoothing formula to the data points.
4. Model evaluation: Assess the model’s performance using metrics such as mean absolute error (MAE), root mean squared error (RMSE) and mean absolute percentage error (MAPE).

Advantages and limitations

Exponential smoothing has several advantages and limitations:

• Advantages:
  • Adjusts quickly to changes in data patterns.
  • Suitable for both short-term and medium-term forecasts.
  • Handles seasonality and trends effectively.
  • A reliable method for time series forecasting.
• Limitations:
  • Requires selection of appropriate smoothing parameters.
  • May not perform well with data having sudden changes or outliers.
  • Can be computationally intensive for large datasets.

Case study

Consider a project management team tasked with forecasting future demand for a new product. By applying triple exponential smoothing to historical sales data, they can account for both trends and seasonal variations. This enables them to predict future demand more accurately, ensuring that production and inventory levels are optimised to meet customer needs.

Tools and software

Several tools and software can assist with exponential smoothing, including:

• R: Provides packages such as forecast for implementing exponential smoothing models.
• Python: Offers libraries such as statsmodels and pandas for exponential smoothing.
• Excel: Includes functionalities for basic exponential smoothing calculations.
• PM3: Supports detailed trend analysis and forecasting, making it easy to apply exponential smoothing techniques across project portfolios.

4. ARIMA (autoregressive integrated moving average)

The ARIMA model is a powerful time series model that combines autoregression, differencing and moving averages to handle various types of time series data. This model is particularly effective for forecasting in environments with complex data patterns, such as stock prices, weather forecasting and sales forecasting.

Components of ARIMA

ARIMA has three main components: autoregression (AR), differencing (I) and moving average (MA). Autoregression uses past values of the dependent variable to predict future values. Differencing involves subtracting previous observations from current observations to make the time series stationary. Moving average models the relationship between an observation and a residual error from a moving average model applied to lagged observations.

Applications in project management

ARIMA can be applied in various aspects of project management:

• Forecasting project completion times: Predicting the time required to complete future projects by analysing past project timelines.
• Resource utilisation: Estimating future resource needs based on historical usage patterns.
• Cost projections: Forecasting future project costs by examining past expenditure trends.
• Risk management: Identifying potential risks by analysing historical data patterns and forecasting future uncertainties.

Steps to implement ARIMA

Implementing ARIMA involves several steps:

1. Data collection: Gather historical time series data relevant to the variables you want to analyse.
2. Model identification: Use time series analysis techniques to determine the order of ARIMA components (p, d, q).
3. Parameter estimation: Estimate the parameters of the ARIMA model using statistical software.
4. Model diagnostics: Evaluate the model’s performance using metrics such as Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC).
5. Forecasting: Use the ARIMA model to make forecasts and interpret the results.

Advantages and limitations

ARIMA has several advantages and limitations:

• Advantages:
  • Handles a wide range of time series data.
  • Effective for capturing trends, seasonality and cyclic behaviour.
  • Provides accurate forecasts for complex data patterns.
• Limitations:
  • Requires careful parameter tuning for optimal performance.
  • Can be computationally intensive.
  • Assumes that the underlying time series is linear and stationary.

Case study

Consider a project management team that needs to forecast the completion times of upcoming projects. By applying ARIMA to historical project timelines, they can account for trends and seasonal patterns. This allows them to predict future completion times more accurately, enabling better planning and resource allocation.

Tools and software

Several tools and software can assist with ARIMA modelling, including:

• R: Offers packages such as forecast for implementing ARIMA models.
• Python: Provides libraries such as statsmodels for ARIMA modelling.
• SAS: A comprehensive statistical software suite that supports ARIMA modelling.
• PM3: Integrates complex time series data analysis, supporting ARIMA model implementation for accurate project forecasting.

5. Neural network models

With the advent of machine learning, neural networks have become prominent in forecasting. These models can capture intricate patterns in large datasets, making them suitable for complex forecasting tasks like market research, consumer behaviour analysis and supply chain optimisation.

Types of neural networks

There are several types of neural networks used in forecasting, including feedforward neural networks, recurrent neural networks (RNN), and convolutional neural networks (CNN). Feedforward networks are the simplest type, where information moves in one direction from input to output. RNNs are suitable for sequential data, as they have connections that form directed cycles, allowing them to maintain information across time steps. CNNs are primarily used in image and spatial data but can also be adapted for certain types of forecasting.

Applications in project management

Neural network models can be applied in various aspects of project management:

• Predictive maintenance: Forecasting when equipment will need maintenance or replacement based on historical data.
• Risk management: Identifying potential project risks by analysing patterns in past project data.
• Optimising resource allocation: Predicting future resource needs to ensure efficient project execution.
• Market research: Analysing large datasets to forecast market trends and consumer behaviour.

Steps to implement neural networks

Implementing neural networks involves several steps:

1. Data collection: Gather large and relevant datasets to train the neural network.
2. Data preprocessing: Clean and prepare the data, including normalisation and splitting into training and test sets.
3. Model selection: Choose the appropriate type of neural network for the forecasting task.
4. Model training: Train the neural network using the training data, adjusting weights and biases through backpropagation.
5. Model evaluation: Assess the model’s performance using metrics such as mean absolute error (MAE), root mean squared error (RMSE) and accuracy.

Advantages and limitations

Neural network models have several advantages and limitations:

• Advantages:
  • Can model complex, non-linear relationships.
  • Suitable for large and diverse datasets.
  • Continually improve with more data.
• Limitations:
  • Require large amounts of data and computational power.
  • Can be difficult to interpret the models’ internal workings (black box nature).
  • Require expertise in machine learning to implement effectively.

Case study

Consider a project management team that needs to forecast future demand for a product. By applying a recurrent neural network (RNN) to historical sales data, they can capture complex patterns and trends. This enables them to predict future demand accurately, ensuring optimal production levels and inventory management.

Tools and software

Several tools and software can assist with neural network implementation, including:

• • TensorFlow: An open-source machine learning framework developed by Google.
  • PyTorch: An open-source machine learning library developed by Facebook, known for its flexibility and ease of use.
  • Keras: A high-level neural networks API, written in Python and capable of running on top of TensorFlow or Theano.
  • PM3: Facilitates the integration of advanced machine learning models for predictive maintenance, risk management and resource optimisation.

6. Econometric models

Econometric models are used to forecast economic variables and are particularly useful for project managers working in finance or sectors heavily influenced by economic conditions. These models use statistical methods to quantify the relationship between economic factors and the dependent variable of interest.

Types of econometric models

There are several types of econometric models, including structural models, reduced form models and simultaneous equation models. Structural models are based on economic theory and specify relationships between variables. Reduced form models express one variable directly as a function of other variables. Simultaneous equation models involve multiple interdependent equations representing complex relationships.

Applications in project management

Econometric models can be applied in various aspects of project management:

• Economic impact analysis: Assessing how changes in economic conditions affect project outcomes.
• Policy evaluation: Analysing the impact of policy changes on project performance.
• Strategic alignment: Forecasting long-term trends to inform strategic decisions.
• Budget forecasting: Estimating future budget requirements based on economic indicators.

Steps to implement econometric models

Implementing econometric models involves several steps:

1. Model specification: Define the economic theory and relationships between variables.
2. Data collection: Gather relevant economic data, including time series data.
3. Parameter estimation: Use statistical methods to estimate the model parameters.
4. Model validation: Evaluate the model’s accuracy using metrics such as R-squared, F-statistic and p-values.
5. Forecasting: Use the model to make forecasts and interpret the results in the context of economic conditions.

Advantages and limitations

Econometric models have several advantages and limitations:

• Advantages:
  • Integrate economic theory with statistical techniques.
  • Useful for long-term forecasting.
  • Can incorporate various economic indicators and market conditions.
• Limitations:
  • Require a solid understanding of economic theory.
  • Assume that the relationships between variables remain stable over time.
  • Sensitive to model specification errors and multicollinearity.

Case study

Consider a project management team working on infrastructure development. By applying an econometric model, they can forecast the impact of economic variables such as interest rates, inflation and GDP growth on project costs and timelines. This enables them to make informed decisions about project financing and scheduling.

Tools and software

Several tools and software can assist with econometric analysis, including:

• • STATA: A comprehensive statistical software package used for data analysis and econometric modelling.
  • EViews: A software package for econometric analysis, primarily used in academic and policy-making institutions.
  • R: Provides packages such as plm and vars for econometric modelling and analysis.
  • PM3: Supports economic impact analysis and policy evaluation, integrating econometric models for strategic planning. 

7. Delphi method

The Delphi method is a qualitative forecasting technique that relies on the consensus of experts. This method is particularly useful for new product forecasting or situations where historical data is limited or non-existent. Through multiple rounds of questioning and feedback, it helps in reaching a well-informed forecast.

Steps in Delphi method

Implementing the Delphi method involves several structured steps:

1. Selection of experts: Choose a panel of experts knowledgeable about the forecasting topic.
2. Initial questionnaire: Develop and distribute a questionnaire to gather initial opinions and forecasts.
3. First round of feedback: Collect responses, summarise the results and share them with the panel.
4. Subsequent rounds: Conduct multiple rounds of questionnaires, refining and narrowing down the forecasts based on feedback.
5. Final consensus: Analyse the responses from the final round to establish a consensus forecast.

Applications in project management

The Delphi method can be applied in various aspects of project management:

• Expert consensus for project risk assessment: Identifying and evaluating potential project risks by consulting experts.
• Strategic planning: Gathering expert opinions to inform long-term strategic decisions.
• New product development: Forecasting market potential and customer acceptance for new products.
• Policy formulation: Developing and assessing policies by leveraging expert insights.

Advantages and limitations

The Delphi method has several advantages and limitations:

• Advantages:
  • Leverages expert knowledge and judgement.
  • Useful for long-term and strategic forecasting.
  • Effective in scenarios with limited quantitative data.
  • Anonymity of experts reduces bias and pressure.
• Limitations:
  • Time-consuming due to multiple rounds.
  • Relies heavily on the expertise and willingness of participants.
  • May result in a consensus that is more conservative or risk-averse.

Case study

Consider a project management team tasked with assessing the risks of a large infrastructure project. By applying the Delphi method, they can gather insights from experts in engineering, finance and environmental science. This helps them identify potential risks, such as cost overruns or environmental impacts, and develop strategies to mitigate them.

Tools and software

Several tools and software can assist with implementing the Delphi method, including:

• SurveyMonkey: An online survey tool that can facilitate the distribution and collection of questionnaires.
• Google Forms: A free, user-friendly tool for creating and distributing surveys.
• Delphi Decision Aid Software: Specialised software designed to support the Delphi method by managing rounds of questionnaires and analysing responses.
• PM3: Facilitates expert consensus through integrated collaboration tools and project management features, enhancing the effectiveness of the Delphi method.

—
Main Image: Astrid IQ