Predictions

Forecast future outcomes in Alviss AI using Predictions to project KPIs like sales or revenue based on customizable drivers. Set variables for granular control, run multiple scenarios for comparisons, and analyze results with time-series graphs, confidence intervals, and filters. Leverage for demand preparation, strategy optimization, and integration with attributions or optimizations.

Predictions in Alviss AI empower you to forecast future outcomes based on customizable drivers, enabling proactive preparation and informed strategic planning. This feature leverages your trained models to project key performance indicators (KPIs) like sales, demand, revenue, or customer metrics under specified conditions. Ideal for scenarios such as inventory optimization, sales forecasting, or budgeting, Predictions focus on absolute outcomes over time, complementing Simulations, which emphasize comparative "what-if" differences.

By adjusting variables like marketing spend, pricing, or external factors, you can generate time-series forecasts with uncertainty estimates, helping you anticipate trends and mitigate risks.

Setting Up Predictions

Variables

Any variable from your dataset can be controlled and set to custom values to explore their influence on outcomes. Follow these steps to configure:

Navigate to Predictions: Access the Predictions section via the side menu or directly at https://app.alviss.io/predictions.
Add Variables: Select variables of interest (e.g., advertising spend, promotions, or economic indicators).
Set Variable Values: Define values for each time point, providing granular control over inputs.
- Example: To forecast sales, set advertising spend to varying levels across time periods and observe the projected impact.

For a detailed walkthrough of the variable modification interface—including patterns, scaling, quick actions, and granular edits—see [Modifying Scenario and Baseline](Simulations#Modifying Scenario And Baseline). The Prediction interface shares the same intuitive controls.

Multiple Scenarios (Optional)

Predictions support running multiple scenarios in parallel for side-by-side comparison:

Copy an existing scenario to iterate on variations.
Add new scenarios from scratch to test diverse assumptions.

This allows efficient evaluation of alternatives, such as different budget levels or market conditions.

Running Predictions

Configure Inputs: Finalize variable settings and any additional parameters (e.g., date range).
Run Predictions: Click "Run Predictions" to process in the background. Results appear once complete.

Understanding Prediction Results

Results provide a comprehensive view of forecasts:

Predicted Values: Core outcomes based on your inputs, such as projected sales totals.
Time Series Analysis: Graphs showing how predictions evolve over time.
Confidence Intervals: Bands indicating uncertainty, helping assess reliability (e.g., likely ranges for outcomes).

Visualizations include totals, per-period trends, cumulative sums, and toggles for tables, uncertainty display, and Filtering. Tooltips offer additional context on hover.

Utilizing Prediction Insights

Leverage Predictions to drive business value:

Prepare for Future Demand: Forecast inventory needs to reduce stockouts or overstock.
Optimize Sales Strategies: Identify high-impact drivers for targeted interventions.
Strategic Planning: Inform investments, resource allocation, and scenario planning with evidence-based projections.

Integrate with features like Attributions for impact breakdowns, Optimizations for recommendations, and Response Curves for sensitivities. For a hands-on guide, see the Run a Prediction tutorial.