AI Dunning: Recover 40% More Failed Payments

If I could give every SaaS founder one piece of retention advice, it would be this: fix your involuntary churn first. It's the single highest-ROI retention investment you can make, and AI dunning makes it almost trivially easy.

Here are the numbers. In a typical SaaS business, 20-40% of total churn is involuntary. Credit cards expire, payments get flagged by banks, accounts hit their limits. These customers didn't want to leave. They often don't even know they've been canceled.

A fixed retry schedule (retry on days 1, 3, 5, 7) recovers maybe 50-60% of these. AI dunning recovers 70-85%. That delta is pure found revenue.

How AI dunning works

Traditional dunning retries your failed payment on a fixed schedule and sends a generic "please update your card" email. AI dunning does three things differently:

1. Smart retry timing

Different payment failures have different optimal retry windows. A card that failed because of insufficient funds is more likely to succeed on the 1st or 15th of the month (payday). A card flagged for suspicious activity needs 48-72 hours for the bank to clear the flag. A card that failed because of a temporary network error might succeed if retried within minutes.

AI models learn these patterns across millions of transactions. They know that Visa cards issued by Chase have a different optimal retry window than Mastercard cards issued by Capital One. This level of optimization is impossible to do manually.

2. Personalized outreach

The email you send to a customer whose card expired is different from the one you send to a customer whose payment was declined for insufficient funds. AI dunning segments the outreach:

• Expired card: "Your card ending in 4242 expired last month. Update it here in 30 seconds to keep your account active." Direct, no shame.
• Insufficient funds: A gentler message, sent later. "There was an issue processing your payment. We'll try again in a few days, or you can update your payment method here." No need to embarrass anyone.
• Bank decline: "Your bank flagged a recent charge. This sometimes happens with recurring payments. You may need to call your bank or try a different card."

3. Escalation intelligence

AI decides when to escalate from automated retries to email outreach to in-app notifications to SMS. It also decides when to stop. Sending 12 emails about a failed payment damages your brand. AI optimizes the number of touches based on what's actually working.

The numbers: how much does this actually recover?

Based on data I've seen from multiple SaaS companies:

• No dunning at all: recover 10-20% of failed payments (some succeed on automatic retry)
• Basic fixed dunning: recover 50-60%
• AI-optimized dunning: recover 70-85%

Let's do the math for a company with $500K MRR and 4% involuntary monthly churn:

• Monthly involuntary churn: $20,000 in at-risk MRR
• Basic dunning recovers: $10,000-12,000
• AI dunning recovers: $14,000-17,000
• Additional monthly recovery: $4,000-5,000
• Annual impact: $48,000-60,000 in saved MRR

And that's just the direct recovery. Each saved customer continues paying for months or years, so the lifetime value impact is 5-10x the initial recovery number. Use the MRR simulator to model this for your specific numbers.

Implementation: faster than you think

This is the easiest AI retention strategy to implement because you don't have to build anything from scratch.

Option 1: Stripe Smart Retries. If you're on Stripe, enable Smart Retries in your dashboard. It uses Stripe's network-wide data (billions of transactions) to optimize retry timing. Free with your Stripe account. This alone gets you most of the benefit.

Option 2: Specialized tools. Churnkey, Butter Payments, or Recurly's dunning features. These go beyond Stripe by adding personalized email outreach, in-app notifications, and payment method update flows. Typical cost: $200-500/month depending on your transaction volume.

Option 3: Build your own. Only if you have unusual requirements. You'd need to build the retry logic, email sequences, and ML model. Takes 4-6 weeks of engineering time. Usually not worth it when the tools above exist.

Add proactive card expiration outreach on top of dunning to catch failures before they happen. This combination covers both reactive recovery and proactive prevention.

For the full playbook, see our smart dunning experiment. To understand where dunning fits in the broader AI retention strategy, read the AI churn reduction guide.