Skip to main content
Documentation
Documentation
Administrator Guide
Last Updated: 2026-02-06
FAQ: Labor Demand Forecasting Engine

FAQ: Labor Demand Forecasting Engine

What is a gradient-boosted tree model?
A Gradient-Boosted Tree model is one of the advanced methods our systems use to make highly accurate predictions. You can think of it as a team of many small decision-makers working together. The team of decision-makers consists (the trees) of:
  1. The First Tree (The Rookie): The process starts with a simple "decision tree." This first tree takes your historical data and makes a very basic guess (prediction). It's usually wrong, or only a bit right.
  2. The Next Trees (The Correctors): We then build a second tree. But instead of trying to predict the original data, this new tree focuses specifically on the errors made by the first tree. Its entire job is to learn how to fix the mistakes of the previous one.
  3. Boosting the Team: This process repeats. Each new tree is built to correct the remaining errors of the entire team of trees that came before it. This process of sequentially fixing errors is called "boosting."
  4. Final Prediction: When it's time to make a final forecast (Example: predicting sales for next week), the model passes the data through all the little trees, and their individual predictions are added together to create one highly accurate, final result.
This method is powerful because it's meticulous. By building thousands of small corrections on top of each other, the model can capture extremely subtle and complex patterns in your data—like how demand might spike higher when a major holiday event is occurring. This is why it often generates the most accurate forecasts for complex business metrics.
What is a Trend-Based Forecasting Prediction Model?
A Trend-based Forecasting model prediction model is one of the most straightforward methods our system uses for forecasting, and it's a great baseline for stability. Here's how it works:
Example: You need to predict next Tuesday's sales. The YoY model simply looks back at what actually happened on that exact day one year ago (last Tuesday). It then adjusts that historical number based on a recent, observed trend.
  1. Find the Baseline: The model finds the exact data point from 365 days ago. This captures all the historical patterns you already know about—like the effect of a certain holiday, a weekly promotion, or a seasonal dip.
  2. Calculate the Growth/Decline: It then calculates the overall rate of change your business has experienced since that time (your "growth rate").
  3. Apply the Change: The model applies that growth or decline percentage to the baseline number from last year to get the final forecast for next week.
The YoY model is most valuable for stable metrics or when you need a reliable, conservative forecast. It works well because it assumes that the underlying patterns of your business remain consistent from one year to the next, while simply scaling the results to match your current rate of growth.