# Trust and Safety at Signup: A Marketplace Playbook

> Marketplaces have a harder signup problem than SaaS, because a bad account damages another customer rather than just costing you margin. Here is how to separate buyer risk from seller risk, and why the same score should not govern both.

**Author:** Matt King | **Published:** August 19, 2026 | **Category:** Fraud Prevention

---

A SaaS product with a fake signup problem loses margin. A marketplace with a fake signup problem loses somebody else's money, and then loses the customer it happened to.

That difference should change your controls, and in most of the marketplaces I have looked at it has not. The same score, the same thresholds, the same checks applied identically to a person buying a used bicycle and a person listing forty of them.

## Two sides, two different risks

Work out what each side can actually do to you before deciding what to check.

**A fraudulent buyer** pays with a stolen card. The chargeback arrives weeks later, and depending on your model it lands on you, on the seller, or on both. The damage is financial, bounded by transaction value, and it surfaces through the payments system where you have some recourse. This is a well-understood problem with well-understood tools.

**A fraudulent seller** takes money for something that does not exist. The buyer is out of pocket, angry in public, and gone. The damage is to trust, essentially unbounded in reputational terms, and it surfaces through support rather than through payments. There is no dispute mechanism that gives you back the reputation.

There is a third case people forget, which is the seller defrauding the platform rather than the buyer: listing goods, buying them with stolen cards through accounts they also control, and collecting the payout before the chargebacks land. That is money laundering with extra steps, and it is why linkage between the two sides of your marketplace is not optional.

So: buyer risk is a payments problem, seller risk is an onboarding problem, and the overlap between them is where the expensive schemes live.

## Do not put the friction at signup

The instinct is to verify sellers hard at registration. It is the wrong instinct, and it is the most common way marketplaces damage their own supply growth.

A seller who has listed nothing has cost you nothing. Requiring identity documents before they can create a listing is friction applied at the moment of maximum uncertainty about whether this person will ever matter to your business, and the drop-off is severe. You lose genuine supply, which is the scarce side of most marketplaces, in exchange for stopping fraud that had not happened yet.

Stage it against value at risk instead:

**At signup**, run the cheap checks that cost the user nothing. Email validation, IP characteristics, device signals. Nothing that requires the user to do anything.

**At first listing**, add velocity checks and cross-account linkage. Is this the fourth seller account from this device this week?

**At first payout**, require real verification. This is the natural checkpoint, because it is the first moment money leaves your platform and the first moment a fraudulent seller must expose a bank account they control. Verification here is also much easier to justify to a genuine user, who is being asked to prove identity in order to receive money, which is exactly what they expect from any financial process.

**At threshold events**, re-verify. Listing above a value band, a sudden change in listing volume, a payout destination change. That last one is worth a hard stop, because payout redirection is the classic account takeover payoff and it is cheap to guard.

## What to check at signup

The checks that cost the user nothing, run on both sides.

```bash
curl -X POST https://fidro.io/api/check \
  -H "Authorization: Bearer $FIDRO_KEY" \
  -d '{
    "email": "new.seller@example.com",
    "ip": "203.0.113.42",
    "account_type": "seller"
  }'
```

```json
{
  "risk_score": 68,
  "signals": {
    "email_disposable": false,
    "email_domain_age_days": 11,
    "datacenter_ip": false,
    "proxy_likelihood": "medium",
    "distinct_accounts_same_ip_30d": 4
  },
  "recommendation": "review"
}
```

The field doing the most work there is `email_domain_age_days`. A disposable address is easy to catch and most operators have moved past them. A freshly registered custom domain is the current shape of the problem: it passes disposability checks because it is a real domain that the fraudster owns, and it is eleven days old. Legitimate businesses do sign up with new domains, which is why this is a scoring signal rather than a block. The broader mechanics are in [how to block disposable emails](/blog/block-disposable-emails-api-guide) and the reasoning behind domain-level scoring in [prevent fake signups without killing conversion](/blog/prevent-fake-signups-without-killing-conversion).

`distinct_accounts_same_ip_30d` needs different thresholds per side. Four buyer accounts from one address is a household or an office. Four seller accounts from one address is worth a human look.

## Linkage is the part that pays

Single-account signals catch opportunists. Marketplace abuse at any scale is operated as a network, and networks are visible only through linkage.

The links worth maintaining:

**Payment instrument to account.** If the same card funds several buyer accounts, or the same bank account receives payouts for several sellers, you have a cluster. The card fingerprint is the strongest version of this signal because cards cost real money to replace, which is the whole argument in [catching cross-account fraud with Stripe card fingerprints](/blog/stripe-card-fingerprint-cross-account-fraud).

**Buyer to seller.** The most important link in a marketplace and the one most often missing, because the two sides are frequently modelled as separate systems with separate tables. If a buyer account and a seller account share a card, a device, or a payout destination, that transaction is not a transaction. It is a wash trade, and the platform is the counterparty.

**Device to account across both sides.** Cheap for a fraudster to rotate, so treat a match as strong evidence and a non-match as no evidence at all.

Store these as relationships rather than as flags on individual accounts. The question you will want to ask during an investigation is "what else is connected to this", and that question has no answer if you only recorded a score.

## Chargebacks are the closing layer

Buyer-side fraud eventually shows up as disputes, and disputes have a threshold with real consequences attached. Under Visa's VAMP programme, dispute rates above 1.5% trigger fees of around 8 dollars per dispute and account reviews, which is the point at which a fraud problem becomes a processing problem. We covered the thresholds in [Visa VAMP 2026](/blog/visa-vamp-2026-chargeback-thresholds).

For a marketplace this arrives with an extra complication: your dispute rate aggregates across sellers, so a handful of bad sellers can push the whole platform over a threshold that then affects everyone. That is a good argument for monitoring dispute rate per seller rather than only in aggregate, and for treating a seller's dispute rate as an onboarding signal for their next listing.

Signup checks reduce how many bad actors get in. Payment-layer checks catch what got through and limit the damage. The order matters: [lead with the check](/vpn-detection-api), close with the payment, and keep the two linked so that a dispute on one side can raise the risk on the other.

[The docs](/docs) cover both integration points, and [the use-case page for SaaS signup protection](/use-cases/saas-signup-protection) has the single-sided version if only one half of this applies to you.

---

## Frequently Asked Questions

### How is marketplace fraud different from SaaS fraud?

In SaaS, a fraudulent account usually costs you margin through free-tier abuse or an eventual chargeback. In a marketplace, a fraudulent account can take money or goods from another user, which converts a cost problem into a trust problem. The damage lands on a customer rather than on your balance sheet, and that difference should shape how much friction you are willing to add and where you place it.

### Should buyers and sellers face the same signup checks?

No. The two sides carry different risk in different directions and at different moments. A fraudulent buyer risks a chargeback against a seller and is best controlled at payment. A fraudulent seller risks taking money for goods that never arrive, and the damage is done before any payment dispute exists, so seller-side controls need to bite earlier and harder.

### How do you stop fake sellers without discouraging real supply?

Stage the verification against value at risk rather than applying it all at signup. Let a new seller list immediately, then require stronger verification before their first payout, before they can raise prices beyond a threshold, or before they can list high-value items. Payout is the natural checkpoint because that is where money actually leaves your platform and where a fraudulent seller must finally reveal a bank account.

### What signals matter most for marketplace signup fraud?

Email disposability and domain age, IP characteristics including datacenter and proxy indicators, whether the same payment instrument or bank account appears across supposedly unrelated accounts, and behavioural velocity such as several seller accounts created from the same fingerprint within a short window. Cross-account linkage tends to matter more here than any single-account signal, because marketplace abuse is usually operated at scale rather than one account at a time.

### When should a marketplace require identity verification?

At the point where the platform's exposure exceeds the cost of the check, which for most marketplaces is the first payout rather than signup. Requiring documents to browse or list kills supply growth for very little benefit, since a fraudulent listing that never results in a payout costs you reputation but not money. Tie the requirement to money movement and the economics work out.