Projection Models | DLH.io Documentation

Overview

Projection models determine how UnitOps AI calculates sales forecasts from historical data. Each location can have a default projection model, and users can switch between models when generating or recalculating a forecast.

Available Projection Models

Trailing Average

Calculates the projection based on the average of recent weeks' actual sales for the same day of week.

Parameter	Description	Default
Window Size	Number of trailing weeks to average	6 weeks
Metric	Which value to project	Sales Dollars

Calculation:

Projected Sales (Monday) = Average of last N Mondays' actual sales

Best for: Locations with consistent weekly patterns and no strong seasonal trends.

Weighted Average

Similar to trailing average, but applies higher weight to more recent weeks.

Parameter	Description	Default
Window Size	Number of trailing weeks	6 weeks
Decay Factor	How quickly older weeks lose influence	0.85
Metric	Which value to project	Sales Dollars

Calculation:

Weight for week N = decay_factor ^ (weeks_ago - 1)
Projected Sales = Weighted sum / Sum of weights

Best for: Locations experiencing growth or decline trends where recent data should carry more influence.

Same Week Last Year (SWLY)

Uses the same calendar week from the previous year as the projection basis.

Parameter	Description	Default
Year Offset	How many years back to reference	1 year
Growth Factor	Year-over-year adjustment multiplier	1.0 (no adjustment)

Calculation:

Projected Sales (Mon, Week 18, 2026) = Actual Sales (Mon, Week 18, 2025) × Growth Factor

Best for: Locations with strong seasonal patterns (tourism, college towns, holiday-heavy areas).

Blended Model

Combines multiple projection methods with configurable weights.

Parameter	Description	Default
Trailing Weight	Weight for trailing average component	0.5
SWLY Weight	Weight for same-week-last-year component	0.5

Calculation:

Projected Sales = (Trailing Avg × trailing_weight) + (SWLY × swly_weight)

Best for: Locations that benefit from both recent trends and seasonal patterns.

Forecast Metrics

The projection model can target different metrics:

Metric	Description
Sales Dollars	Total gross sales revenue
Guest Count	Number of transactions/covers
Average Check	Revenue per transaction

When projecting by Guest Count or Average Check, the system derives total sales by multiplying the two projected components.

Configuration

Projection models are configured per location in Forecasting > Settings > Forecast Configuration:

Navigate to Forecasting in the sidebar
Open Settings (gear icon)
Select the Forecast Configuration tab
Choose the location to configure
Set the Default Projection Model and its parameters
Click Save

Recalculating a Forecast

Once a draft forecast exists, you can recalculate it with a different projection model:

Open the weekly forecast
Click Recalculate in the toolbar
Select the desired projection model
Review the updated projections
Optionally adjust individual days

Recalculation replaces all system-generated values but preserves any manual overrides (with a confirmation prompt).

Page Contents