Finance ops · Production

Stripe's Adaptive Acceptance AI upgrade recovers $6 billion in falsely declined transactions in 2024

The problem

False declines cost US online retailers an estimated $81 billion in lost sales in 2023, affecting more than half of US customers and considered a major issue by 43% of retailers — with revenue losses from false declines often exceeding those from actual fraud.

First attempt

Stripe's previous gradient-boosted tree model (XGBoost) saved billions of dollars in revenue but was limited in modeling the complex interactions among hundreds of factors influencing transaction success, and recent AI advances offered potential for improvement.

Workflow diagram · grounded in source
1
Issuer rejects legitimate payment
trigger
“recognize complex patterns in transaction data that indicate a legitimate payment was mistakenly rejected by issuers as suspected fraud”
2
TabTransformer+ detects false decline
ai_action
“We transitioned to a TabTransformer-based deep neural network, which we call TabTransformer+. This system excels at modeling complex interactions among hundreds of factors that influence transaction success. A key enhancement in this new…”
3
Automatic retry in real time
output
“automatically identify and retry false declines in real time—without the end customer ever seeing the initial decline”
4
Continuous model retraining
feedback_loop
“This new system reduced our model training time from days to hours, and it allowed us to use a larger dataset—helping our model better understand the complexities of different transaction types and payment behaviors. We can now train and…”
Reported outcome

In 2024, Adaptive Acceptance recovered $6 billion in falsely declined transactions — a 60% year-over-year increase in retry success rate — while achieving 70% greater precision and reducing retry attempts by 35%.
Model training time was cut from days to hours, enabling multiple deployments per week.

Reported metrics
Falsely declined transactions recovered (2024)$6 billion
Retry success rate year-over-year increase60%
Precision improvement in identifying false declines70%
Retry attempts reduced35%
Show all 8 reported metrics
falsely declined transactions recovered (2024)$6 billion
retry success rate year-over-year increase60%
precision improvement in identifying false declines70%
retry attempts reduced35%
model training timefrom days to hours
US online retailer false decline losses (2023, industry)$81 billion
retailers rating false declines as major problem43%
revenue saved by previous XGBoost modelbillions of dollars
Reported stack
Adaptive AcceptanceXGBoostTabTransformer+
Source
https://stripe.com/blog/ai-enhancements-to-adaptive-acceptance
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

In 2024, Adaptive Acceptance recovered $6 billion in falsely declined transactions — a 60% year-over-year increase in retry success rate — while achieving 70% greater precision and reducing retry attempts by 35%.

What tools did this team use?

Adaptive Acceptance, XGBoost, TabTransformer+.

What results were reported?

Falsely declined transactions recovered (2024): $6 billion; Retry success rate year-over-year increase: 60%; Precision improvement in identifying false declines: 70%; Retry attempts reduced: 35% (source-reported, not independently verified).

What failed first in this deployment?

Stripe's previous gradient-boosted tree model (XGBoost) saved billions of dollars in revenue but was limited in modeling the complex interactions among hundreds of factors influencing transaction success, and recent A…

How is this finance ops AI workflow structured?

Issuer rejects legitimate payment → TabTransformer+ detects false decline → Automatic retry in real time → Continuous model retraining.

Related finance ops cases