Automatic Forecasting: Sales Driven Models Meet Linear Regression

Businesses use forecasting models regularly, especially for their financial statements and operating budgets. Two widely-used approaches are sales-driven forecasting models and linear regression forecasting with independent line-item forecasts. Each has distinct benefits and limitations that influence their suitability depending on business needs and context.

Sales-Driven Forecasting Models

These models begin with a forecast of total revenue (the top line) and then project all other line items, such as cost of goods sold (COGS), variable expenses, and even fixed costs, as a percentage of that revenue.

Usually, the percentages used are either based on a company’s own historical averages or on industry sector comparable figures. This makes it quick to apply and great for benchmarking your company’s performance against peers.

But there is a catch. By its nature, this approach assumes all key expenses and profits scale directly with sales. This can result in overlooked anomalies or non-linear behaviors in costs, such as fixed expenses not rising at the same rate as revenue, or COGS fluctuating due to supply chain changes rather than sales alone.

What does it mean? That sales-driven forecasting models may miss irregularities and company-specific patterns that deviate from industry norms.

Linear Regression Forecasting with Independent Line-Item Forecasts

Here you use linear regression (or other regression techniques) to forecast each line item independently. For example, cost of goods sold (COGS), fixed expenses, and other major accounts are forecasted based on their own historical trends, not as a percent of revenue.

By analyzing each line item separately you can spot anomalies or unusual trends more readily. Significant deviations in, say, utility costs or raw material expenses become evident, as they aren’t masked by a revenue-driven formula that uses an assumed percentage.

Regression models can also help you identify causal relationships unique to the particular company. And this supports more granular, data-driven decision making.

This method requires historical data in order to produce more realistic and potentially more accurate forecasts.

Best Practice: Use Both Approaches

Combining both models is often the best practice. Sales-driven models ground forecasts in industry averages and facilitate comparisons. On the other hand, line-item regression forecasts focus on actual trends and can reveal unique risks or opportunities within the business.

Running both in parallel allows you to cross-validate results, spot historic trends, monitor for unexpected shifts, and ensure forecasts are both consistent and supportable.

You will find this dual-modelling approach especially valuable for business valuation. For the best results, consider including both models in your forecasting toolkit.

See Valuation Tour