Skip to main content

OrderIQ Configuration Parameters

Distributor guide to OrderIQ suggested order settings in Encompass: order_size, round_threshold, product overrides, POS data, layer rounding, loss function and Partition ID. Needs the OrderIQ app and v25.08+.

B
Written by Bailey Naas-Thornton

Minimum Version Required: 25.08

Pre-Conditions

Before configuring OrderIQ, ensure the following are in place to allow for certain features:

  • Install OrderIQ App: OrderIQ App is installed and suggested order configurations are enabled and configured for accounts

  • Retail Insights Data Connection: For leveraging POS data, a connection to Retail Insights must be active. This ensures the Handoff Product Inventory Daily Trace table is populating with data.

  • Inventory Layer Thresholds: To enable rounding to the next layer or pallet, the promptLayerThreshold must be set up for relevant products in the Inventory table.

Steps

OrderIQ is a solution that provides tailored product suggestions to users. It has several configuration parameters that allow for granular control over how these suggestions are generated. These parameters help optimize inventory and streamline the ordering process.

Granular Order Size Control

The order_size parameter allows users to control the size of a suggested order.

How it Works

  • It uses a percentage list (e.g., 0.5;0.4) to distribute a predicted weekly order across multiple delivery stops.

  • The default value is 1, which corresponds to a 100% order.

  • If you set order_size to 1.2 (120%), OrderIQ will recommend an order that is 120% of a standard order.

How to Configure

  1. Navigate to the Suggested Order Configuration dashboard.

  2. Update the order_size parameter.

  3. The value can be a percentage list or a single decimal.

  4. If the number of delivery stops is less than the number of percentages listed, the remaining percentages are summed and applied to the last delivery stop.

Adjustable Rounding Thresholds

The round_threshold parameter dictates how predicted quantities are rounded up or down.

How it Works

  • It is a decimal value between 0 and 1.

  • A higher round_threshold increases the confidence required to add the next incremental case, which is useful if you have a high product removal rate.

  • A lower round_threshold makes it easier to add the next incremental case, which is helpful if you have a high product addition rate.

  • For products designated as 'Sell By Case Only,' predicted quantities are converted to cases, rounded using the specific threshold, and then converted back to units for precise suggestions.

How to Configure

  1. Navigate to the Suggested Order Configuration dashboard.

  2. Update the round_threshold with a decimal value between 0 and 1.

  3. This setting ensures that suggestions align with your desired ordering patterns.

Product Type-Specific Customizations

You can apply unique order_size and round_threshold settings for specific products or product types.

How it Works

  • Specific settings for a product, product type or product type code take precedence over the general configurations.

  • For products designated as 'Sell By Case Only,' predicted quantities are converted to cases, rounded using the specific threshold, and then converted back to units for precise suggestions.

How to Configure

  1. Navigate to the Suggested Order Configurations table.

  2. Select the sub-table Suggested Order Overrides.

  3. Click Add Suggested Order Overrides.

  4. Add a product type, product, or product type code.

  5. Update the order_size or round_threshold parameters for the selected items.

  6. Click Add Record.

POS Data Integration

Integrating Point-of-Sale (POS) data can significantly boost the precision and reliability of order suggestions.

How it Works

  • By including historical POS sales and inventory metrics, the system can deliver more accurate forecasts.

  • This feature is controlled by the EnablePOSData parameter in the Suggested Order Configuration table.

  • The available options are:

    • False: Excludes all POS data.

    • Include POS Sales: Incorporates only retail sales data into the model.

    • Include POS Sales & Inventory: Includes both retail sales and inventory data. Note that this requires accurate inventory data.

How to Configure

  1. Ensure a Retail Insights Data connection is active and the Handoff Product Inventory Daily Trace table is populated.

  2. Navigate to the Suggested Order Configuration table.

  3. Update the Enable POS Data column to one of the three options.

Operational Packing Units

This feature automatically rounds suggested order quantities to the nearest full layer or pallet.

How it Works

  • It leverages existing system configurations for product packing and inventory thresholds.

  • The feature uses the inventory.promptLayerThreshold to determine when to round a suggested value to the next layer.

  • When a product's recommended order is greater than or equal to its prompt order layer threshold, OrderIQ will round the quantity up to the next layer. This ensures that recommended orders align with operational packing requirements.

How to Configure

  1. Go to the Inventory table and set up the prompt layer threshold for the products you wish to round.

  2. Navigate to the Suggested Order Configurations table.

  3. Set Enable Layer Round to True for OrderIQ accounts.

Fields And Definitions

  • order_size - A percentage value or a list of percentages that controls how predicted orders are scaled and distributed across multiple deliveries. A value of 1.2 suggests 120% of a standard order.
    ​
    ​

  • round_threshold - A decimal between 0 and 1 that sets the confidence level for rounding predicted quantities. A higher value reduces the chance of rounding up to the next case, while a lower value increases the likelihood to add the next case.
    ​
    ​

  • Partition ID - This setting determines how many models OrderIQ trains by grouping customer data.
    ​
    ​

  • Loss Function - The mathematical method used by the AI model to calculate prediction errors.

    • Squared Error (Default) - A general-purpose loss function that treats all errors equally. This is what OrderIQ uses unless the Loss Function is changed in the OrderIQ app settings.
      ​
      ​

    • Tweedie Loss - A specialized loss function designed for data with many zero values and occasional large spikes, common in beer distribution. Set the Loss Function to reg:tweedie to use it.
      ​
      ​

  • EnablePOSData - A configuration setting that dictates whether to include POS data in the OrderIQ model.

    • False - Excludes POS data.
      ​
      ​

    • Include POS Sales - Integrates only retail sales data.
      ​
      ​

    • Include POS Sales & Inventory - Includes both sales and inventory data.
      ​
      ​

  • Enable Layer Round - A boolean (True/False) setting that, when enabled, automatically rounds suggested quantities to the nearest full layer or pallet based on the promptLayerThreshold.

Best Practices

  • Start with Defaults: Begin with the default settings and observe performance before making extensive changes. This provides a baseline for evaluating the impact of your customizations.
    ​
    ​

  • Use Tweedie Loss for Specific Patterns: If your business has many products with zero sales on most days and occasional large orders (e.g., craft beers), switch the Loss Function to reg:tweedie. This will lead to more accurate suggestions.
    ​
    ​

  • Tailor with Specificity: Use the Suggested Order Overrides sub-table on Suggested Order Configurations to apply custom order_size and round_threshold settings for high-velocity or unique product types. This granular control prevents irrelevant recommendations and optimizes inventory for key products. Do not use Suggested Order Limitations for this: it is the legacy Retail Insights path and is ignored entirely for accounts with AI Suggested Order set to True.
    ​
    ​

  • Leverage POS Data: If available, enable Include POS Sales to improve prediction accuracy. If you have reliable and accurate inventory data, consider enabling Include POS Sales & Inventory for even greater precision.
    ​
    ​

  • Optimize for Operations: For products that are handled in specific packing units (layers, pallets), enable Enable Layer Round to ensure order suggestions align with operational requirements, reducing the need for manual adjustments. This makes the ordering process more efficient.
    ​
    ​

  • Best Practices for PartitionID Selection: Choosing the right Partition ID is a crucial step in optimizing your OrderIQ models. The best choice depends on the nature of your customer data. Here's a concise guide to help you decide:

    • CustomerID: Choose this for high-volume, consistent accounts where you need maximum personalization. This option trains a unique model for each customer. It's highly accurate for customers with rich historical data but performs poorly for new or low-volume accounts due to data scarcity.
      ​
      ​

    • CustomerTypeID: Use this for low-volume or low-velocity accounts with sparse data. By training a single model for each customer type (e.g., bars, restaurants), it pools data from similar customers. This improves the model's ability to learn general sales trends and prevents it from overfitting to limited data. However, it may overlook individual customer nuances and relies heavily on accurate customer categorization.
      ​
      ​

    • ChainID: Select this when ordering patterns are standardized across a chain. It trains a single model for all locations within a chain, providing consistent suggestions. This is particularly useful for new locations that can benefit from the collective data of the entire chain. A drawback is that it may ignore unique local variations and demands.

Did this answer your question?