Skip to content

Optimizations

Hand-tuning simulations and predictions for maximum effectiveness can be tedious and time-consuming. Instead, Alviss AI provides an optimization feature, allowing you to achieve optimal outcomes efficiently by setting constraints and letting the system do the heavy lifting.

Mode

Dynamic Optimization

Dynamic optimization seeks to find the best possible value given the constraints set. It automatically adjusts the input variables to maximize or minimize the outcome based on your specified goals and constraints.

Goal-Driven Optimization

In goal-driven optimization, you set a specific target to achieve. This can be an absolute value (e.g., exactly X number of sales) or a value relative to the baseline (e.g., Y more sales than the baseline).

Date Range

Choose the date range for the input you want to optimize. You can also decide whether to include pre and post periods in your analysis. Depending on your model, some variables, like media, can have a carry-over effect on your KPIs. This means the investment made in a specific period might affect your KPIs in later periods.

  • Pre-period: Including the pre-period allows you to consider input data before the optimization period, resulting in more accurate predictions.
  • Post-period: Including the post-period makes the optimization more accurate by accounting for the effects on KPIs that your variables might have after the optimization period.

Target

Specify the KPIs for optimization.

  • Dynamic Optimization: Choose whether you want your target to be maximized or minimized.
  • Goal-Driven Optimization: Set a goal you want to achieve and decide if you want it to be relative (the increase or decrease from the observed value) or absolute.

Optimization Variables

Select the input variables that should be optimized. Variables not selected for optimization will retain their values from the dataset.

Constraints

Set conditions for the sum of a variable over time. Specify whether the variable should be greater than, less than, or equal to a specific value. If multiple variables are selected, their values will be summed together.

  • Use Case: Setting a total media budget and combining it with a maximum spend for individual media channels across all time points.

Boundary Constraints

Boundary constraints are set for each individual time point and include the following types:

  • Max: The maximum value the variable can become.
  • Threshold Max: If the value exceeds the Threshold Max, it will jump to the Max value. Note that Max must be set, and Threshold Max must be less than the Max value.
  • Threshold Min: If the value falls below the Threshold Min, it will jump to the Min value. Note that Min must be set, and Threshold Min must be greater than the Min value.
  • Min: The minimum value the variable can become.

Directions

For goal-driven optimizations, you can set directions for variables to specify the desired trend while achieving the set target. This includes:

  • Minimize: Make the variable as small as possible.

  • Maximize: Make the variable as large as possible.

  • Neutral: No fixed direction, or simply do not set a direction for the variable.

  • Use Case: Setting a media variable to Minimize would mean trying to hit your target while spending as little as possible on media.

Running an Optimization

  1. Navigate to Optimization: Go to the Optimization section in the Alviss AI platform.
  2. Select Mode: Choose between dynamic optimization and goal-driven optimization.
  3. Set Variables and Constraints: Select the input variables to be optimized and apply any equality, inequality, or boundary constraints.
  4. Configure Directions (if applicable): For goal-driven optimization, set the desired directions for the variables.
  5. Run Optimization: Click on the “Submit” button to execute the optimization process.

Understanding the Results

When the optimization completes, Alviss AI provides detailed results that include:

  • Optimized Values: The best possible values for the input variables given the constraints and goals.
  • Outcome Analysis: A comparison of the optimized scenario against the baseline, highlighting the improvements or changes achieved through optimization.