Refund Strategy (Operator Field Manual)
- Before a formal chargeback is filed, refunding versus standing behind the sale is a math decision.
- Set your refund cutoff from actual fees and evidence work. The ticket-size bands below are examples to adapt.
- Fast refunds can prevent chargebacks only before one is filed. A refund returns the sale; a chargeback can add dispute fees and response work. Once a formal card chargeback is open, respond through the dispute case.
- Customer support is your best fraud prevention tool. Empower agents to refund without escalation.
- Track refund-to-chargeback ratio. Keep both counts beside it; check whether earlier refunds reduce disputes.
Cold reality: refund fast when it's cheaper than a chargeback, fight when evidence is strong, and route edge cases by ticket size and business model.
The Math: Refund vs. Fight
Compare the two routes on the same basis. The examples below count cash returned from the original sale, original processing fees and staff time. Product and shipping costs already incurred are common to both routes unless one action recovers them; don't add them again on only one side.
Cost of refunding = Transaction amount returned
+ Original processing fee retained
+ Refund handling time
Expected cost of a later chargeback = (1 - Win rate) x Transaction amount
+ Original processing fee retained
+ Expected dispute fees after any fee refunds
+ Evidence work
This is a pre-filing decision model. Once a formal chargeback is open, use its case workflow; don't send a separate refund. Keep monitoring-program consequences separate from the dollar calculation.
Worked Examples
These are assumed inputs, not benchmarks: the original processing fee is 3%, refund handling is $3, and the fixed dispute fee is $15. Win rates and evidence-work costs are shown in each row. Replace them with your own fees and results.
$30 order, fraud claim, no 3DS:
| Action | Math | Total Cost |
|---|---|---|
| Refund | $30 + ($30 x 3%) + $3 staff time | $33.90 |
| Fight (25% win rate) | 75% x $30 + ($30 x 3%) + $15 fee + $25 evidence work | $63.40 |
Refund: $63.40 - $33.90 = $29.50 less expected cost.
These examples assume a fixed $15 dispute fee. Real schedules differ and may refund a fee on a win. Use your provider's actual terms from the fee table.
$200 order, "not received," tracking confirms delivery:
| Action | Math | Total Cost |
|---|---|---|
| Refund | $200 + ($200 x 3%) + $3 staff time | $209 |
| Fight (70% win rate) | 30% x $200 + ($200 x 3%) + $15 fee + $30 evidence work | $111 |
Fight on these assumptions: $209 - $111 = $98 less expected cost. The expected recovery before common processing fees is (0.70 x $200) - $45 = $95.
$150 order, "not as described," no photos:
| Action | Math | Total Cost |
|---|---|---|
| Refund | $150 + ($150 x 3%) + $3 staff time | $157.50 |
| Fight (20% win rate) | 80% x $150 + ($150 x 3%) + $15 fee + $30 evidence work | $169.50 |
Refund on these assumptions: $169.50 - $157.50 = $12 less expected cost. Expected recovery before common processing fees is (0.20 x $150) - $45 = -$15.
Run your own numbers
The calculator on this page compares refunding now with fighting a chargeback later, before any dispute is filed. Here every dispute fee counts, because without a dispute there's no fee, and fighting avoids the refund handling cost. Expected saving from fighting = win rate x amount, minus expected dispute fees after any fee refund on a win, minus evidence work, plus refund handling. Above zero, fighting a later chargeback is expected to cost less than refunding now. Below zero, refunding costs less. Processing fees are left out because both routes keep them. The $200 example above: 0.70 x $200 - $15 - $30 + $3 = $98, the same difference the table shows.
Already Refunded? Did the Customer Cancel or Withdraw the Chargeback?
Get this wrong and you pay the same customer twice. Find your row before you touch anything.
| What's happened | Separate refund? | Do this |
|---|---|---|
| Customer contacted you; no bank case exists | Available | Use the grid below. If you approve the request, refund through the normal payment flow. See handle a refund request |
| Manual Ethoca or CDRN alert | Available if no credit exists | Check transaction and refund status, then void or refund within the provider deadline and report the outcome |
| Visa RDR shows resolved | No | RDR already created the credit or liability shift. Don't issue another refund |
| Inquiry or retrieval | Provider-specific | Reply inside the case. Satisfactory evidence or a full refund may resolve an inquiry; a partial refund can still escalate |
| Formal card chargeback is open | No separate refund | Use the dispute workflow. Accept or return the funds, or challenge or defend if allowed |
| Refund succeeded before the chargeback | No | Submit proof showing succeeded status, amount and date. Add the ARN, STAN, RRN or processor reference if available |
| Refund is pending when the chargeback opens | No second refund | Check whether the pending refund failed or was cancelled, then respond through the dispute case |
| Customer says they withdrew it | Not while the case remains open | Submit available withdrawal evidence and your response by the deadline. Treat it as closed only when the provider marks it resolved |
The disputed funds are already held or debited. A separate refund can give the customer a second credit.
Stripe and Shopify Payments block separate refunds while a formal card chargeback is open. Resolve the payment through the dispute case. A partial refund during an inquiry can still escalate, and an ACH refund can race with a bank return.
The processor initiated a refund or reversal? Identify the event and ask for its basis.
Decision Flowchart
It excludes RDR-resolved cases and inquiries or retrievals, which have their own provider workflow. The dollar bands are example settings. Recalculate them from the cost model above before using them. No chargeback has been filed. Here, "don't refund" means preserve evidence for a possible later challenge. It weighs dollar amount and evidence strength. For the reason-code decision, see Chargebacks Overview. For 3DS liability considerations, see Representment.
Refund vs. Fight Grid
Same scope as the flowchart: nothing has been filed yet. Here, "fight" means don't refund and keep your evidence for a possible challenge. Once a formal card chargeback is open, "refund" leaves the grid: use the formal card dispute workflow, and accept or return the funds, or challenge or defend if allowed. The economics below still tell you which one.
By Ticket Size
These are example routing bands, not expected win rates or industry cutoffs. Replace them using the cost model above and your comparable final outcomes. No amount makes weak evidence strong, and no band overrides a valid refund obligation.
| Amount | Default | Exception | Reasoning |
|---|---|---|---|
| Under $25 | Consider refund | Check actual fees and handling cost | A $0 or $15 fee doesn't always exceed the sale |
| $25-$50 | Refund | Fight only with delivery proof + fraud claim | Staff time to fight exceeds potential recovery. |
| $50-$100 | Usually refund | Fight with strong evidence AND high win-rate reason code | Break-even zone. Evidence quality decides. |
| $100-$500 | Evaluate evidence | Fight with strong evidence; refund if evidence is weak | Worth the effort when evidence is solid. |
| Over $500 | Fight with evidence | Refund only if zero evidence exists | Almost always worth attempting. |
By Business Model
| Type | Refund Bias | Why | What Changes the Calculus |
|---|---|---|---|
| Digital goods | High | No physical delivery proof; hard to prove customer received it | Device fingerprint matching (CE 3.0) or download logs |
| Physical goods | Low | Shipping confirmation and tracking are strong evidence | Missing or ambiguous tracking weakens your case |
| Services | Medium | Subjective "quality" claims are hard to disprove | Signed contracts, completion photos, or time logs help |
| Subscriptions | High | Cancellation-related disputes are hard to win without clear proof | Timestamped cancellation logs, confirmation emails, renewal notices |
By Dispute Reason
| Reason | Default | Notes |
|---|---|---|
| "I don't recognize this" | Refund | This is a billing descriptor problem. Fix the descriptor; don't waste time fighting. |
| "I cancelled" | Refund unless proof exists | Cancellation timestamp + confirmation email wins this. Without those, refund. |
| "Product not as described" | Evaluate | Product photos, specs, and terms help. Subjective claims are hard. |
| "Never received" | Fight with tracking | Tracking to billing address is strong evidence. This is your most winnable category. |
| "Duplicate charge" | Verify and refund if true | Easy to confirm. If it's actually a duplicate, refund immediately. |
| "Fraud" | Fight with CE 3.0 data | Device match, 3DS authentication, or Visa CE 3.0 data makes this winnable. |
Track your own win rate by reason code rather than relying on a general figure.
By Customer History
| History | Approach | Reasoning |
|---|---|---|
| First-time customer | Lean toward refund | Preserve the relationship. A refund might earn a repeat customer. |
| Repeat customer, first dispute | Refund + benefit of doubt | Good customers have bad days. Don't punish loyalty for one incident. |
| Customer with 2+ prior disputes | Evaluate carefully, document pattern | Possible friendly fraud abuser. |
| Repeated disputes with evidence of abuse | Review the claim and document the pattern | A dispute count alone doesn't prove abuse. Apply any future purchase restriction under your policy. |
Support as Risk Control
Your support team prevents more chargebacks than any fraud tool. A customer who can reach you doesn't need to call their bank.
Empowered Refund Authority
Give support agents authority to refund up to a threshold without manager approval. The table gives example starting limits, not limits justified by revenue alone. Set yours from ticket size, margin and escalation cost; review mistakes and delays after a month:
| Volume | Suggested Threshold | Why |
|---|---|---|
| Under $100K/month | $50 per refund | Covers most small-ticket issues instantly |
| $100K-$500K/month | $100 per refund | Matches typical order value |
| $500K-$1M/month | $150 per refund | Reduces escalation queue |
| Over $1M/month | $200+ per refund | Speed at scale |
The math: A $75 refund returns the sale. If it replaces a chargeback you would have lost, the possible saving is the avoided dispute fee and response work, less refund handling and alert fees. Original processing fees and shipped goods may be unrecovered either way. Use your own costs; not every refund prevents a chargeback.
Response Time SLAs
These are starting service goals. Staff them against your opening hours and request volume, publish what you can meet, and track repeat contacts and disputes while you adjust them.
| Channel | Target | Why This Matters |
|---|---|---|
| Phone | Under 2 minutes hold | Give customers a usable route to resolve the charge with you. |
| Live chat | Under 1 minute | Same urgency as phone. |
| Under 4 hours | Before the customer gives up and disputes. | |
| Social media | Under 2 hours | Public complaints escalate fast. |
If your response time is measured in days, customers go to their bank instead. Every day of delay increases the probability of a chargeback.
Scripts That Prevent Chargebacks
"I don't recognize this charge":
"I can see the charge you're asking about. It's from your order on [date] for [item], and it shows on your statement as [descriptor]. I can send you a copy of the receipt. Would you like a refund, or does that clear things up?"
"I want to cancel":
"Done. I've cancelled your subscription effective immediately. You won't be charged again. Your confirmation number is [number]. Is there anything else?"
"This isn't what I expected":
"I'm sorry to hear that. I can process a full refund right now, or if you'd prefer, I can send a replacement. What works best for you?"
The pattern: acknowledge the problem, offer a solution, execute immediately. No transfers, no hold music, no "let me check with my manager."
Seasonal Adjustments
Refund strategy isn't static. Adjust thresholds based on seasonal patterns:
| Season | Adjustment | Why |
|---|---|---|
| Nov-Dec (holiday) | Test a higher authority limit if past requests justify it | Review gift-related returns, staffing and abuse cost |
| Jan (post-holiday) | Review the temporary limit as those orders mature | Use your own return and dispute timing before extending it |
| Major sale events | Temporarily raise refund authority | Volume spike + impulse purchases = more buyer's remorse |
| Subscription renewal dates | Staff up support | Renewal batch = spike in "I cancelled" complaints |
Test to Run
Refund threshold experiment:
- Pick eligible direct refund requests with no open chargeback and no existing credit. Set approval limits from your own margins, handling cost and policy.
- Record refunds, dispute outcomes, all service fees and support time for comparable orders before and after the change.
- Allow each sales cohort the same time for disputes to arrive; two weeks alone may miss later filings.
- Compare total dollars retained after refunds, disputes and handling costs. Separate product recovery from costs already incurred.
Success criteria: The changed policy improves retained dollars on comparable orders without increasing abuse beyond your agreed limit. A falling chargeback count alone doesn't prove a saving.
Metrics to Track
| Metric | What It Tells You | Target | Red Flag |
|---|---|---|---|
| Refund rate | Refunded orders / orders in the same mature sales cohort | Compare the same products and channels | Rising losses or a new refund reason on comparable sales |
| Refund-to-CB ratio | Are refunds resolving problems before disputes? | Track, don't target | Check refund delays and disputes together; see Benchmarks |
| Time to refund | Approval to refund submission; bank posting is separate | Meet your published promise and applicable deadline | Approved refunds not submitted by that date |
| Win rate by reason code | Where fighting pays off | Varies by code | Costs more to fight than you expect to recover |
| Cost per chargeback | True all-in cost including staff time | Track, don't target | Use to calibrate your refund threshold |
The Monthly Refund Review
Once a month, spend 15 minutes on this:
- Pull your refund-to-chargeback ratio. Keep both counts for the same period. Check whether refund delays or a hard-to-find refund path are driving disputes.
- Check win rate by reason code. Stop fighting categories where the evidence work costs more than you expect to recover.
- Review your top 5 refund reasons. Are they product issues, shipping issues, or billing confusion? Each has a different fix.
- Spot-check 5 chargebacks that were previously refund requests. These are failures - the customer asked for a refund, didn't get one (or got it too slowly), and disputed instead. Fix the process.
- Update your refund threshold. If your average chargeback cost has changed (new processor fees, different win rates), recalculate the break-even point.
Scale Callout
| Volume | Focus |
|---|---|
| Under $100K/month | Set refund authority using your fees, margin and evidence work. Document cancellations. Name one person to resolve requests promptly. |
| $100K-$500K/month | Add reason-code routing. Implement renewal reminder emails for subscriptions. Capture delivery proof on all physical goods. Track refund-to-CB ratio monthly. |
| $500K-$1M/month | Dedicated dispute owner. Evidence packs by reason code. Alert before your processor's written ratio limit. Seasonal threshold adjustments. |
| Over $1M/month | Automated refund processing for low-ticket items. Abuse detection on serial refunders. Tiered customer treatment based on history. Real-time chargeback ratio monitoring. |
Where This Breaks
- No cancellation proof for subscriptions. Add renewal reminders, cancellation confirmation emails, and timestamp logs. Without these, every "I cancelled" dispute is a loss.
- Partial shipments or backorders without customer communication. If the customer ordered 3 items and you shipped 2, tell them before they assume you shorted them.
- Refund promises without follow-through. Support says "we'll refund you" but finance never processes it. The customer waits a week, then disputes. Track refund fulfillment rate.
- International shipping with weak delivery confirmation. Domestic tracking that shows "delivered" wins cases. International tracking that shows "in transit to destination country" does not. Use a carrier with delivery confirmation in the destination country, or accept higher refund rates on international orders.
- "No refund" policies on digital goods. A strict policy doesn't remove dispute rights. Set and disclose a window that fits access, delivery and applicable law. Compare refund costs and later disputes on the same products before widening it.
Next Steps
Setting up triage grid?
- Apply the decision flowchart - Route by amount and evidence
- Implement by scale - Right-sized for your volume
- Track monthly metrics - 15-minute monthly check
Empowering support?
- Set refund authority thresholds - Agents refund without escalation
- Set response time SLAs - Beat the bank
- Use the scripts - Consistent, fast resolution
Fighting specific disputes?
- Check reason codes - What evidence defeats each
- Build evidence packs - By reason code
- Follow representment guide - Full workflow
Related
- Refund Policy Design - Policy design and customer-facing language
- Refund Fraud - Abuse patterns and prevention
- Chargeback Prevention - Reducing disputes
- Winning Evidence - Fighting chargebacks
- Compelling Evidence - Evidence types by reason code
- Representment - Full representment guide
- Chargeback Metrics - Tracking dispute rates
- Fraud Metrics - Fraud rate tracking
- Subscriptions & Recurring - Subscription refund scenarios
- Friendly Fraud - First-party abuse patterns
- Reason Codes - Understanding dispute types
- Experimentation - Testing approaches
- Alerts Configuration - Setting up alerts
- Running Fraud Operations - Operational cadence