Here are a few frequently used qualitative forecasting approaches:

Delphi method

In the Delphi method, several experts are invited to answer a series of questionnaires seeking their perspectives on the business case or metric to be forecasted. Responses are anonymous, allowing viewpoints to be considered equally. Replies from the previous questionnaire are used to craft the next questionnaire, and this process continues until a consensus is reached on a forecast.

Market research

Enterprises enlist the help of market research firms to conduct customer surveys and ask their opinions about products or services. Data collected from these surveys is then used to inform sales forecasts and product or service improvement initiatives.

Benefits and limitations of qualitative forecasting

Qualitative forecasting has the following advantages:

• It can be used when data is limited, such as when evaluating the market acceptance rate or market penetration rate of new products or technologies.

• It integrates information from experts and people highly knowledgeable about the enterprise and its offerings, which quantitative data might be unlikely to capture.

• It can often consider one-off incidents or atypical scenarios, like a crisis or disaster. This means that qualitative forecasting might be a good fit for situations where conditions are constantly evolving.

But this type of forecasting also has its drawbacks:

• Because it relies on human judgment, qualitative forecasting can be subjective, incorporating bias that leads to either overemphasized or overlooked factors and assumptions.

• Qualitative information might at times consider only the most recent events or first-hand experiences, so long-term trends or patterns from past data might be missed.

Here are some common quantitative forecasting strategies:

Time series forecasting

This quantitative method uses historical data modeled as a time series to project future outcomes. A time series is a series of data points plotted in chronological order.

Time series forecasting models can help reveal predictable trends in the data influenced by cycles, irregular fluctuations, seasonality and other variations.

Time series analysis is frequently mentioned alongside time series forecasting. While time series analysis entails understanding time series data to glean insights from it, time series forecasting moves beyond analysis to predict future values.

Time series forecasting encompasses a number of methods:

Naive

The naive method uses the data point from the previous period as the forecast for the next period. This makes it the simplest time series forecasting method and is often considered a preliminary benchmark.

Simple moving average

The simple moving average technique calculates the average of the data points from the last T periods. That average then serves as the forecast for the next period.

Weighted moving average

This method is based on the simple moving average technique, but with a weight applied to each data point of the last T periods.

Exponential smoothing

Exponential smoothing works by applying an exponentially weighted average to time series data. Weights diminish exponentially as data becomes older—the more recent the data, the more weight it has.

A smoothing coefficient (also called a smoothing factor or smoothing parameter) controls the weights assigned to past and current data. Using these weights, the weighted moving average is then computed and serves as the forecast. This forecast becomes a smoothed version of a time series, eliminating fluctuations, noise, outliers and random variations from the data.

Exponential smoothing doesn’t normally require a huge dataset, which makes it a good forecasting method for short-term projections. And because it gives more weight to current data, exponential smoothing can quickly adapt to new or changing trends.

Seasonal index

A seasonal index can be valuable for businesses whose production or demand of goods or services is dependent on the seasons.

To compute the seasonal index, take the average demand for a particular season and divide it by the average demand across all seasons. These averages are usually calculated using a moving average technique, but exponential smoothing can also be applied using time series data only for that season. A resulting seasonal index less than 1 signifies a lower than average demand, while a value greater than 1 denotes a higher than average demand.

To estimate the forecast for the next season, that season’s projected demand will be multiplied by the corresponding seasonal index.

Causal models

Causal models are a mathematical expression of causal relationships in data. These forecasting models can be suitable for forecasts with a longer time horizon.

Regression models

Regression-based models analyze the relationship between a forecast or dependent variable and one or more predictor or independent variables. An example of a regression model is linear regression, which represents a linear relationship between a forecast variable and a predictor variable.

Econometric models

Econometric models are similar to regression models, but with a focus on economic variables, such as interest rates and inflation, and economic relationships, such as market conditions and asset prices.

Benefits and limitations of quantitative forecasting

Quantitative forecasting offers these advantages:

• It’s grounded on numbers and math, which can result in more objective predictions. 

• It provides consistent, replicable and structured outputs that help streamline analysis across specific time frames.

But this forecasting approach also has some pitfalls:

• It’s difficult to merge expert insights, insider information and other qualitative data into quantitative forecasts.

• It needs sufficient historical data to produce reliable predictions.