Gazelle Forecasting Optimizer

Forecasts built on what drives demand.

Gazelle Forecasting Optimizer is a fully automated, multi-industry forecasting solution. It tests a wide range of internal and external variables, selects the ones that matter, and validates every forecast it produces.

Thousands
of models tested automatically
Minutes
to a full set of forecasting models
Multi-industry
internal and external drivers

The problem

A forecast that only looks backwards can’t explain what changed.

Many forecasts extend past sales forward. They miss the drivers — price, promotion, seasonality, market conditions — and tuning better models by hand takes analysts weeks.

  • History-only forecasts miss the reasons demand moves.

  • Choosing which variables to include is slow, manual and subjective.

  • Models are rarely re-tested once they are in production.

  • Forecasts are hard to connect to the business decisions behind them.

How it works

Hundreds of models in, one validated forecast out.

Every stage is automated, so the search for the best model happens in minutes instead of weeks.

  1. 01

    Select the causal drivers

    Hundreds of Random Forest models run automatically to identify the most important causal variables.

  2. 02

    Measure their impact

    An automated regression-based algorithm measures how each driver moves demand, running hundreds of iterations to find the best-fit model.

  3. 03

    Optimize the models

    Thousands of models forecast the drivers themselves — chosen to maximize fit and minimize out-of-sample error.

  4. 04

    Calculate the forecast

    The final forecast is calculated from the forecasted drivers, with automated insights on what is behind it.

Forecasting Optimizer · model search

Hundreds Random Forest models

select the causal drivers

Hundreds regression iterations

measure each driver’s impact

Thousands driver forecasts

max fit, min out-of-sample error

1 validated forecast

calculated from the drivers

Fully automated — minutes, not weeks.

What it evaluates

How each forecast is judged

Model fit

How well the model explains the demand you have already seen.

Out-of-sample error

Accuracy on data the model was not trained on — the honest test.

Driver importance

Which internal and external variables matter most, ranked.

Driver impact

How much each variable moves demand, measured by regression.

Why it’s different

What sets Forecasting Optimizer apart.

1

Speed

A full set of forecasting models in minutes.

2

AI automation

Drivers are selected and forecast automatically to arrive at the optimal algorithm.

3

Cloud-based

Scales to businesses of any size, across industries.

4

Self-reliant

Your team runs it — no dependency on external vendors.

See it in action

See Forecasting Optimizer on a category like yours.

We show the real product on real data in a 30-minute walkthrough — not a mock-up.

Data & integrations

Works with the data you already have.

Internal drivers

Sales historyPrice & promotion plansPromotion calendar

External drivers

SeasonalityMarket conditionsOther external series

Systems

SAPTallyUnicommercePower BICSV & Excel

Forecasting Optimizer FAQ

Questions we hear about this.

Something else? info@northlightanalytics.com

What can it forecast?+

Demand and revenue across industries. The level of detail and forecast horizon are set during onboarding to match how your business plans.

Which variables does it use?+

A wide range of internal and external variables can be tested. The AI selects the most important ones automatically, so you don’t have to decide in advance.

How do we know the forecast is accurate?+

Models are chosen to minimize out-of-sample error — accuracy on data the model has not seen — not only to fit the past.

Does it connect to our promotion plans?+

Yes. Price and promotion plans, including the calendar from the Promotion Optimizer, can feed the forecast as drivers.

See it on your own data.

A 30-minute walkthrough of Gazelle on a category like yours.