Almost every internal conversation about return-fraud budget starts with the same mistake: "We lose X euros a month on fraudulent refunds, so anything that costs less than X euros a month to prevent it is worth doing." The number is wrong on both sides of the equation. The visible refund undercounts the real loss by a factor of two or three, and many of the non-obvious costs are the ones that compound over time.
This article puts a number on what is actually lost when a fraudulent claim succeeds. The numbers are based on pooled data from mid-sized DTC brands we have worked with (€2M–€50M online revenue) and cross-checked against public NRF and bevh benchmarks. Your actual multipliers will differ; the model is the point.
The Iceberg Model
| Cost bucket | Typical amount | % of true cost | Visibility |
|---|---|---|---|
| 1. Direct refund + reverse shipping | €100 + €8 | 54 % / 4 % | Visible — P&L line |
| 2. CS and intake labour | €18 | 9 % | Visible if time-tracked |
| 3. Resale-value write-down (if any product returned) | €22 | 11 % | Usually invisible |
| 4. Brand-trust drag (social / review share) | €15 | 8 % | Invisible |
| 5. Chilling effect on legit returns | €18 | 9 % | Deeply invisible |
| 6. Chargeback fee (when denial escalates) | €15–30 × 10 % | 2 % | Visible but attributed elsewhere |
| True cost | €182 (median) | 100 % | Rarely totalled |
The €100 refund is what ends up in the financial report. The €82 of additional loss is distributed across four teams (customer support, operations, brand, marketing) and rarely reconciled into a single return-fraud line. This is the single biggest reason return-fraud defence is under-funded: the P&L only shows half the problem.
Bucket 1 — Refund and Reverse Shipping
This is the easy number. The refund flows back to the customer at the ticket's recorded amount; the reverse shipping is either paid by the merchant (typical in EU under §357 BGB for goods above €40) or absorbed as "free returns" for brand reasons. Across our dataset, the refund averages the order value and reverse shipping adds 6–10 %.
For a €100 fraudulent claim, bucket 1 is €108. Straightforward.
Bucket 2 — CS and Intake Labour
Every fraudulent claim consumes CS time. The time is distributed across several touchpoints:
The labour profile of a fraudulent damage claim
Typical minutes per touchpoint. Multiply by the agent's loaded hourly rate (salary + overhead, ~€35/hour for EU CS).
- Initial ticket triage3–5 minutes. The CS agent reads the claim, checks the order record, and classifies the ticket. Even a quick triage costs €2.
- Photo review (when fraud is suspected)5–12 minutes. A careful look at metadata, visible signs of manipulation, or cross-checks against the reverse-image search. This is the most variable bucket — software reduces it to 1 minute, manual review takes 15+.
- Customer back-and-forth (stages 1–2 denial)8–20 minutes across two to three emails. The template approach reduces this but rarely below 8 minutes because the agent has to fill placeholders and personalise opening.
- Intake inspection (if physical return)5–10 minutes. A warehouse worker inspects, photographs, and records the item condition. At a loaded €22/hour for warehouse labour, this is €2–4 per case.
- Chargeback defence preparation (10 % of cases)30–45 minutes. If the claim escalates, someone collates evidence, writes the narrative, and uploads to the payment processor's dispute portal.
- Management review (high-value or escalated cases)10–30 minutes of a lead's time. Happens in roughly 20 % of denied claims.
Pooled across these touchpoints, the typical fraudulent claim costs €15–25 in CS and intake labour. For our €100-claim example, the median is €18. The labour cost is higher than merchants assume because the time is distributed — the ticket feels like "only ten minutes" but the downstream photo review, inspection, and occasional chargeback prep are all part of the same claim.
Bucket 3 — Resale-Value Write-Down
Relevant only when a physical product comes back. Two sub-cases:
Item returned in unsellable condition. A wardrobed dress that comes back perfumed, a worn swimsuit, an opened beauty product — these cannot be resold as new. Typical write-down: 60–100 % of retail. You have paid the refund AND lost the inventory.
Item returned in B-grade condition. Electronics opened and tested, apparel with minor handling. Resold through outlet channels at 30–60 % of retail. Write-down: 40–70 %.
For a €100 fashion order where the item was wardrobed and unsellable, bucket 3 is easily €30–40. For an empty-box case where no product comes back at all, bucket 3 is 0 — but then bucket 1 is €100 against 0 recovered inventory. Either way, the total is the same order of magnitude.
Bucket 4 — Brand-Trust Drag
The least tangible bucket, and the hardest to budget against. Two mechanisms:
Public-review blast radius. A customer denied a legitimate claim because your CS team was too aggressive writes a one-star review. A reviewer looking at your Trustpilot sees the review and decides not to buy. The dollar impact is real but diffuse. Studies from the UCLA Anderson School and Harvard Business Review put the typical revenue impact of each one-star review between €60 and €300 in lost first-time-customer revenue, depending on review volume and conversion elasticity.
Social-media propagation of fraud tactics. When TikTok and Reddit posts celebrate a successful claim against your brand, the playbook spreads. Our dataset shows a measurable spike in similar-pattern claims within 72 hours of a viral TikTok, concentrated on the specific brand named in the video.
For our €100 example, bucket 4 averages €12–18 across the dataset. This is the bucket with the widest range; some cases are zero, some are hundreds.
Bucket 5 — The Chilling Effect on Legitimate Returns
This is the least-discussed and the one that compounds worst. Every round of policy tightening aimed at fraudsters also friction-slows legitimate customers. Measurable consequences in our dataset:
| Policy change | Reduction in fraud | Reduction in legitimate repeat purchase | Net effect |
|---|---|---|---|
| Shorten window 30 → 14 days | -15 % fraud | -4 % repeat purchase | Slight positive |
| Add restocking fee (10 %) | -22 % fraud | -11 % repeat purchase | Mildly negative depending on mix |
| Require photo before refund | -35 % fraud | -8 % repeat purchase | Positive |
| Ban returns after one prior refund | -48 % fraud | -19 % repeat purchase | Strongly negative |
| Charge for return shipping | -30 % fraud | -15 % repeat purchase | Usually negative |
The chilling effect is the bucket that surprises CFOs. The finance team approves a policy change that reduces fraudulent claims by 30 %; six months later loyal customers have 15 % lower purchase frequency. The fraud line dropped but so did LTV. Pricing this bucket requires a controlled A/B or cohort analysis — we allocate €15–20 on a €100-claim basis as a rough assumption, but it is the softest number in the model.
Bucket 6 — Chargeback Fees
When a fraudulent claim is denied and the customer escalates to a card dispute, the merchant pays a processor fee regardless of who wins. Stripe, Adyen, and Mollie charge €15–30 per disputed transaction. Winning the dispute returns the transaction amount but not the fee.
Roughly 10 % of denied claims escalate to chargeback. On €100 claim basis: 0.10 × €20 = €2 average. Small per claim, meaningful in aggregate for high-volume merchants.
The Worked Example
Put it together for a €100 fashion order, wardrobed and denied at stage 2 (no chargeback):
| Bucket | Calculation | Amount |
|---|---|---|
| 1. Refund (partial, §357a applied at 50 % deduction) | €100 - €50 refund = €50 refunded | €50 |
| 1b. Reverse shipping | Paid by merchant | €8 |
| 2. CS labour | 22 min × €35/hour loaded | €13 |
| 2b. Intake inspection | 8 min × €22/hour warehouse | €3 |
| 3. Resale write-down | Dress resold at 40 % retail = €40 on €100 item | €60 - €40 recovered = €20 |
| 4. Brand drag | Low — no public dispute | €6 |
| 5. Chilling effect on repeat purchase | Small allocation | €10 |
| 6. Chargeback fee (10 % probability × €20) | Expected value | €2 |
| True cost of the claim (with stage-2 denial) | €112 |
Note how different this looks from a full refund without §357a deduction:
| Bucket | Amount |
|---|---|
| 1. Full refund + shipping | €108 |
| 2. Labour (smaller, approved quickly) | €8 |
| 3. Write-down (full, item unsellable) | €60 |
| 4. Brand drag | €3 |
| 5. Chilling effect | €0 — no policy friction |
| 6. Chargeback | €0 |
| True cost of full-refund approval | €179 |
Partial refund with inspection is not only morally more defensible — it is €67 cheaper per case. Across a mid-sized DTC processing 500 fraudulent-adjacent claims per year, that's €33k of recovered margin.
What This Means for the Defence Budget
Two operational implications.
First, the business case for photo-forensic software or trained CS time is built on the full €182 cost, not the €108 refund. A tool that costs €18k a year and reduces successful fraudulent claims by 200 cases is a €36k saving on refund alone — but €12k+ of CS-labour recovery and €4k+ of write-down recovery on top, for a €36k + €12k + €4k = €52k annual saving against €18k cost. That reframes the ROI conversation.
Second, the chilling effect (bucket 5) puts an upper bound on policy-only defence. Above a certain tightness, each additional unit of fraud prevention costs more in lost loyalty revenue than it saves. Our rough threshold is around 30 % fraud reduction from policy alone; past that, investment has to shift to detection (which reduces fraud without customer friction) rather than policy (which reduces fraud with customer friction).