Payment Agent

The Payment Agent works behind the scenes to recover more revenue from failed payments. It continuously tests new retry strategies against your current one, rolls out the ones that recover more, and shows you the results along the way. This article explains how the Payment Agent works, how to turn it on, and how to check on its progress.

 


What is the Payment Agent?

The Payment Agent is an AI specialist that handles payment-recovery optimization for you. When a subscription payment fails, the timing and number of retries affect whether you get that revenue back, and finding the best retry strategy normally takes ongoing manual testing and analysis. The Payment Agent does that work for you, automatically and around the clock.

The Payment Agent focuses on soft declines: payments declined for a temporary reason, such as insufficient funds, that a well-timed retry can often recover.

What does the Payment Agent do?

Once it's on, the Payment Agent is always working to:

  • Analyze your payment data—declines, retries, and recoveries—to understand what's working.
  • Design new retry strategies worth testing.
  • Test each strategy against your current one with a controlled A/B test.
  • Roll out the winner automatically once a new strategy proves it recovers more, then look for the next improvement.
  • Report its results, with a record of every test it has run and how each one performed.

How the Payment Agent works

Every test compares two retry strategies on your store's failed payments: your current strategy (the control) and a new strategy designed by the agent (the variant). Here's the cycle:

  1. The agent chooses a new retry strategy to test.
  2. It splits eligible failed payments between your current strategy and the new one, and runs the test until it's ready - at least 30 days, up to 90 days.
  3. If the new strategy proves it recovers more, reaching at least a 95% probability to be better, the agent rolls it out automatically and it becomes your current strategy. If it doesn't, your current strategy stays in place.
  4. The agent starts its next test, using your current strategy as the new baseline.

Because each winning strategy becomes the baseline the next test has to beat, your recovery strategy keeps improving over time instead of standing still.

How the agent decides what to test

The Payment Agent bases each new strategy on:

  • Your store's retry history—how often past retry attempts have succeeded.
  • Network-wide learnings—what's working in payment retry tests across Ordergroove, updated continuously as payment patterns shift.
  • Card network and issuer rules—which limit how many times a card can be retried within a set number of days.

Within those limits, the agent adjusts how many retry attempts to make and when to make them: the time of day, the day of the week, and the day of the month.

Built-in safeguards

The Payment Agent runs on its own. To protect your recovery while it works:

  • It monitors every test as it runs. If it detects that a test is hurting your results, it stops the test, restores your previous strategy, and tells you why.
  • It records every test and its results, so you can review each decision at any time.

Before you turn it on

The Payment Agent needs a few things in place to work. If any of these apply to your store, your CSM can help.

Your decline data needs to be readable. The agent learns from the decline reasons your payment processor sends to Ordergroove when a payment fails. If that data doesn't give the agent enough to work with, the Payment Recovery page shows the message Your data is not currently compatible with what our agent needs instead of the Payment Agent toggle. Common causes include:

  • Failed payments are reported as generic or technical errors (such as status code 999) rather than as payment declines. See Order Placement Status Codes.
  • Decline reasons are missing or don't carry meaningful information, like N/A.
  • Nearly every decline shares the same reason, so there's nothing for the agent to tell apart.

To see the decline reasons your store sends, go to Data > Analytics, switch the view to Orders, and click Explore on the Rejection Rate card. The Rejected Order Reasons table lists each reason and its rejection code. For more detail, see Understanding the Orders Analytics Tab. Your CSM can work with our data team to get your decline data into a format the agent can use.

Ordergroove needs to manage your retries. If an outside system, such as a third-party payment recovery tool, determines your retry cadence, the Payment Agent isn't available for your store and its card won't appear on the Payment Recovery page. Instead, the Strategy section notes that an outside system is determining your retry cadence.

Shorter subscription frequencies need a plan. If many of your subscriptions renew more often than every 30 days (for example, every 7 or 14 days), talk to your CSM before turning on the Payment Agent. With shorter frequencies, a customer can have more than one order in retry at the same time, so we'll help you plan the right setup first.


Turning on the Payment Agent

  1. Go to Settings > Payment Recovery.
  2. In the Payment Agent card, click the toggle to turn it on. The card previews what the agent will test first: your current configuration (control) against a personalized strategy (variant).

That's it; there's no implementation work on your end. The agent starts its first test right away, and the card shows A test is currently running.


Checking on the Payment Agent

Results appear after a test's first week and update weekly after that, so we recommend checking in once a week. Pick a regular day, like Monday, to see how the current test is going and whether any tests have finished.

On the Payment Recovery page

The Payment Agent card shows the status of the current test. For the first week, its metrics read Collecting data; early results swing too much to be meaningful, so the agent waits until it has enough data. After that, the card shows:

  • Recovery Rate Lift: how much more (or less) the new strategy is recovering compared to your current strategy.
  • Prob. to be better: how confident the agent is that the new strategy truly outperforms your current one. A new strategy needs to reach 95% to roll out.
  • Est. days to decision: about how long until the test reaches a decision.

Depending on where the agent is in its cycle, the card shows one of these statuses:

Status What it means
A test is currently running The agent is comparing your current strategy with a new one.
Test ended: variant is the winner The new strategy recovered more. It has been rolled out automatically and is now your current strategy.
Test ended: no conclusive winner The new strategy did not product a statistically conclusive winner.
Test ended: variant lost The new strategy didn't beat your current one, so your current strategy stays in place.
No test is currently running The last test has ended, and the agent will start a new one soon.
Current test was stopped The agent stopped the test early to protect your results. The card shows the reason and confirms your previous strategy has been restored. A new test starts automatically soon.

If a new test starts as soon as the last one ends, the card shows the new test along with a summary of the result that just came in.

Experiment details

Click See Experiment Details on the Payment Agent card to open the full results for a test, including:

  • A description of the control and variant strategies.
  • The test's status, lift, estimated annualized incremental revenue, probability to be better, traffic split, estimated days to decision, and start date.
  • Charts and a metrics table comparing control and variant across recovery rate, order retention rate, recovered revenue, recovered orders, and customers. Order and customer counts are available from day one, before the first weekly results.

Recovery Rate is the ship decision metric: the metric the agent uses to decide whether a new strategy wins. Other metrics, like Order Retention Rate, take longer to mature, so they may not show results until later in the test.

After a test ends, you can switch between two views of its results:

  • Metrics at the moment of test completion: the results the agent's decision was based on.
  • Metrics continued after the test: metrics the agent keeps tracking afterward, like order retention rate, to show how the strategy holds up over time.

All experiments

To see every test the Payment Agent has run, go to Data > Experiments, or click See Past Experiments on the Payment Agent card. The Experiments page lists every experiment in your program, including tests from workflows and cancel flows, with each one's status, lift, and probability to be better. Filter the list by Payment Retries to see only the Payment Agent's tests.


Turning off the Payment Agent

Since each test builds on the last, we recommend leaving the Payment Agent on. If you need to turn it off:

  1. Go to Settings > Payment Recovery.
  2. Click the Payment Agent toggle to turn it off.
  3. Click Turn Agent Off to confirm.

When the Payment Agent is off:

  • A test that's already running continues until it's complete.
  • No new tests start.
  • The agent stops analyzing your data for better retry strategies.

You can turn the Payment Agent back on at any time.


Additional reading

  • Ordergroove Agents — an overview of agents, how they differ from "just AI," and the full agent lineup.
  • Retention Agent — how the Retention Agent spots subscriber drop-off and builds workflows for your approval.
  • Analytics Agent — how to get insights and answers from your program data.
  • Experiments Overview — how A/B testing works in Ordergroove.
  • Payment Recovery — your store's broader payment-recovery settings, where the Payment Agent lives.