High Buyback Customers Agent (Reference)

The High Buyback Customers Agent continuously scans transactional data (with returns captured as negative quantities) over a rolling time window to compute return rate per customer and identify those with abnormally high buyback behavior or return rates that negatively impact profitability.

List of Required Fields

The following table lists the set of transaction fields required for the High Buyback Customers Agent to compute its metrics and detection rules. Ensure these fields or equivalent are available and consistently populated in your source Transaction Datamart before deploying the Agent. Those exact field names are not required, but a similar fields with that data is required.

Definition

Series 1 (Transactions)

Group By

Measures

Filters

Filters are criteria applied to transactional data to ensure only valid records are analyzed, commonly excluding zero-value transactions and constraining invoice dates using relative ranges computed from a configurable anchor date, specified as N months ago.

Example

Evaluate transactions from the last three months up to today, with returns and buybacks recorded as negative quantities.

Pricing Date between (inclusive) custom N months ago 3 (1/20/2026) AND custom Today (4/20/2026)

Summary Table

• Customer ID – The top-level grouping dimension selected in the Group By definition. Rows are organized by Customer ID.

• Customer Name – The Customer Name used as the next grouping level within the table hierarchy.

• Product Group – Optional grouping level that narrows the analysis to specific product families, as referenced in the action description.

• Revenue – Total invoiced value of transactions in the analysis period, including the effect of returns and buybacks recorded as negative amounts.

• Quantity – Net invoiced quantity over the analysis period, with returns and buybacks stored as negative quantities.

• Return % – The calculated return or buyback rate per customer (and product group) over the selected three-month horizon.

• Potential Revenue – Estimated additional revenue that could be achieved if the identified high-buyback behavior is corrected according to the agent’s methodology.

Detection Rules

Detection Rules defines the Agent’s alert conditions and includes scheduling. When conditions are met, actions are triggered during the next Agent run.

Default thresholds are flexible, adjust as necessary.

Example

This rule flags customers whose realized return / buyback rate is higher than 1%, marking them as high‑buyback customers.

Return % > 0.01

Schedule

Set the preferred start date and frequency that you want the Agent to run.

Example

Start Date – The date when the scheduled task will run for the first time.
Period – Period which represents the offset between each run.
Interval – Interval which represents the number of repetitions in a selected period. Allowed characters are 0-9. 0 means one-off run.

Action Definition

Example

Similar Case Handling

• Period – Defines the time period between potential similar action. Similar case will not be recreated before the defined period. Time unit for the duplicate-prevention window.

• Interval – Number of periods between similar actions. Similar case will not be recreated before the defined interval. Prevents creating a very similar case for the same context within the defined interval.

Example

Prevents creating a very similar case for the same context within a 3-month period.

Impact Calculation

You can define specific metrics to compute the foreseen impact of the actions. Please only use total absolute value (and not relative values) as those metrics will be aggregated.

Impact Definition

• Potential Revenue – Estimated additional revenue that could be realized if the high buyback behavior is corrected, with a 50% realization rate used to keep expectations conservative.

• Potential Profit – Estimated profit improvement from reduced returns and buybacks, also modeled with a 50% realization rate.

• Order – Defines the display order of these impact metrics when summarizing the actions created by the agent.

Summary

Review the setup of the Agent, here you can see all the set parameters in one place.

If you are happy with the setup, click Submit for Approval. Once the Agent is approved, it becomes active and starts monitoring your data based on the schedule.