Docs
Revive is Chargebee's machine learning–driven recovery engine designed to optimize payment retries and maximize revenue recovery. It intelligently analyzes multiple attributes, including payment data, customer behavior, merchant characteristics, invoice details, and timing, to determine the optimal time and method to retry a failed transaction.
Operating within a 30+ dimensional space, Revive leverages complex ML algorithms developed from Chargebee's extensive payment behavior data to predict the most effective retry strategies, significantly improving recovery rates without manual intervention.
At its core, Revive identifies behavioral patterns that indicate when a payment is most likely to succeed. It continuously learns from outcomes and adapts its retry predictions over time. By doing so, it replaces rigid retry rules with intelligent, data-driven decisions that evolve with your customers and payment environments.
To ensure that the ML model is properly tailored and effective, Chargebee follows a structured enablement process:
Data Access: We secure your permission to use relevant data for training purposes.
Data Curing: Over a 3–4 week period, your data is analyzed and fine-tuned to align the models with your specific business patterns.
Onboarding: Using a no-code onboarding flow on the Revive Dashboard, we align on your current recovery rates and initiate setup.
Early Access Program (EAP): Once enabled, you can log in to the dashboard to track Revive's performance in real time.
There is no coding effort required.
The time it takes to observe results depends on your current dunning window. We recommend measuring recovery performance after at least one full dunning cycle. Beyond that, you can accurately assess Revive's impact on revival rates.
During the Early Access Program, Revive supports:
Gateways: Stripe and Braintree.
Payment Methods: Cards, Apple Pay, and Google Pay.
All other payment methods and invoices continue to be managed by Chargebee Smart Dunning.
Revive includes multiple safeguards to ensure reliability and maintain performance:
Adheres to industry best practices for retry counts and timing.
Responds dynamically to gateway feedback (e.g., avoiding retries when instructed not to).
Performs continuous internal monitoring, with alerts for any performance dips.
Allows merchants to instantly revert to Smart Dunning through a no-code switch if needed.
Revive operates autonomously in a high-dimensional algorithmic environment, which means adding manual retry rules can interfere with its performance. Therefore, custom rules are not supported.
However, Revive will always respect your defined boundary conditions, such as:
Maximum of 12 retries per invoice.
The configured dunning period.
Revive evaluates over 35 parameters to determine the likelihood of a successful retry. While individual factor contributions are not currently exposed, and explainability is hard, merchants can access visual insights on the dashboard, including analytics on revival rates, payment method trends, and gateway performance, to better understand and interpret system outcomes.
No. Revive does not require personally identifiable information (PII) such as customer name, email address, or phone number for payment retry optimization. The models select approximately 30 relevant parameters from a broader set of around 300 data points. These parameters are limited to payment-related information such as BIN, issuer country, card country and similar transaction attributes.
Chargebee Revive includes standard off-the-shelf models. A merchant’s data may be used to fine-tune models specific to that merchant and may also contribute to enhancing Chargebee’s standard Revive models.
The merchant retains ownership of their data. Chargebee is granted a perpetual license to use the data to improve and enhance the quality, performance, and functionality of its Services, including fine-tuning, benchmarking, and developing, training, and enhancing machine learning models.
Chargebee applies its standard tenancy and data-handling safeguards to ensure appropriate separation and protection of customer data.
ML model training may continue beyond an opt-out or cessation of use in accordance with the license granted to Chargebee.
Was this article helpful?