From error codes to actions
Understand payment declines. Take the right action.
Decline Decoder helps merchants and payments teams understand why a payment failed, whether to retry, what to tell the customer, and how to prevent similar failures.
Example diagnosis
Why it failed
Insufficient funds. The customer's account was short on the debit date.
What to do
Re-present the EMI a few days after the customer's likely salary date, and remind them two days before.
Retry?
Yes — In the next NACH presentation cycle, timed after salary credit rather than straight away.
Prevent it
Move the EMI date close to when the customer is paid.
Customer message
Your EMI payment of ₹7,500 didn't go through this time. We'll try again on [date], so please keep the amount available in your account.
What can you do with Decline Decoder?
Turn payment failures into actionable insights.
Understand the root cause
Decode cryptic error codes across cards, UPI, UPI Autopay and NACH into plain English.
Know whether to retry
Retry advice with timing, based on NPCI, RBI and card-network rules.
Get the right customer message
Ready-to-send messages that never guess or blame the customer.
Find patterns and prevent failures
Spot repeated problems in a batch, like expiring mandates, and fix them at the source.
How it works
From one decline to a clear next step.
Payment details
The decline code, payment method, amount, mandate type and attempt number, plus anything the bank or customer said.
Check the reference
The code is looked up in a reference of 200 codes grouped into 18 root causes. If it isn't there, the tool says so instead of guessing.
AI diagnosis
For known codes, AI picks the right cause from the reference and applies written retry rules to recommend what to do and what to tell the customer.
Patterns and prevention
Upload a batch to see what's recoverable now and which repeated problems to fix.
How it was built
A measured path from reference data to reliable recommendations.
Built the reference
Mapped 200 decline codes from NPCI, ISO 8583, Razorpay, Cashfree and Stripe into 18 root causes.
Wrote the test set first
50 scenarios with expected answers, including trick cases and made-up codes, plus 10 kept aside.
See the test setTested a simple version (v1)
One AI prompt: right cause 82–84% of the time, right retry decision 70–74%, and it invented answers for 4 of 6 codes it should have refused.
Rebuilt it (v2)
Reference lookup first, then AI with written rules: right cause 100%, right retry decision 98%, and no invented answers.
Checked the rules
Audited every retry and compliance rule against current NPCI, RBI and card-network sources, and fixed three that were out of date.