Referral Fraud Risk Checker

How the referral fraud risk score works

The Referral Fraud Risk Checker scores your referral program from 0 to 100. Lower is safer. Each question maps to a common way referral programs get abused, and each answer adds risk points:

• Missing controls. Each of the 11 checkbox controls carries a fixed weight. An unchecked or unanswered control counts as not in place, and the tool adds its weight. Self-referral blocking (12 points) and rewarding only on a verified paid order (11 points) carry the most.
• Reward timing. Instant payout adds 16 points. Payout after the friend's purchase adds 9, payout after the return window adds 2, and payout after manual approval adds 0.
• Reward value. Value adds points in tiers: 0 under $25, 5 at $25, 9 at $50, 14 at $100, and 18 at $250 or more.
• Reward type. Cash, gift cards, and PayPal add 8 points. Store credit adds 3, free product 2, and a discount on a future order 1.
• Friend offer delivery. One public, reusable code adds 13 points. A shared link with no unique code adds 7, and a unique single-use code adds 0.

The tool divides your total by the maximum possible total of 131 points and scales the result to 100. Scores of 0–33 are Low, 34–66 are Medium, and 67–100 are High. The same math runs for each of the five categories (reward design, coupon and code sharing, identity and duplication, order verification, limits and review), so the breakdown shows where your risk sits. The recommended rules list only the fixes that match your gaps, high severity first.

The dropdowns start on their riskiest option and the reward value starts at $25, so a blank questionnaire scores 90 (High). The score drops as you record the controls you already have.

Worked example

Take a program that gives advocates $50 in store credit after the friend's purchase and sends each friend a unique single-use code. It blocks self-referrals, matches email and domain, and rewards only verified paid orders. It has no phone, device, IP, region, payment, or shipping checks, no velocity limit, and no manual review.

Timing adds 9 points, the $50 value adds 9, store credit adds 3, the unique code adds 0, and the eight missing controls add 46. That is 67 of 131 points, a score of 51: Medium risk. The breakdown shows limits and review at 100% risk, order verification at 58%, reward design at 50%, identity at 44%, and coupon sharing at 0%.

The checker's top recommended rules for this program:

1. Flag same device, IP, payment, or shipping address (high).
2. Delay the reward until the order is paid and the return window passes (high).
3. Limit referrals per advocate per day or week (medium).
4. Require manual review above high reward thresholds (medium).

Now move payout to after the return window and add device checks, IP checks, and a velocity limit. The total falls to 40 points, a score of 31: Low risk. Advocates still get the same $50 in store credit.

Types of referral fraud

Referral fraud is any use of a referral program to collect rewards without bringing in a genuine new customer. Talkable's features page states that across the industry, 40% of referrals are fraudulent. Most referral program fraud falls into five patterns, and each one maps to questions in the checker:

• Self-referral. One person acts as advocate and friend with a second email, phone number, or address. Identity checks and payment and shipping matching target this.
• Fake or duplicate accounts. Throwaway accounts pose as new customers to collect rewards or stack first-order discounts. Paid-order verification and phone checks raise the cost.
• Code leakage. A reusable code lands on coupon sites and deal forums, and strangers redeem it as referrals. The friend offer delivery question covers this.
• Collusion rings. Real people refer each other in loops. Payment, shipping, and region checks plus manual review catch rings that pass email checks.
• Velocity abuse. Scripts generate referrals far faster than any real advocate. Per-advocate velocity limits stop it.

Referral fraud prevention controls

The checker groups prevention into layers. No single control stops everything, and together they make abuse cost more than it pays:

• Reward design: delayed payout, a reward value sized to your margin, and store credit instead of cash.
• Code control: unique, single-use friend codes.
• Identity matching: self-referral blocking, email and domain matching, and phone, device, and IP checks.
• Order verification: rewards only on paid orders, plus payment method, shipping address, and region matching.
• Limits and review: per-advocate velocity caps and a manual review queue above a dollar threshold.

For step-by-step setup of each control, read Talkable's guide to preventing referral program fraud. Talkable's post on smart strategies for preventing referral fraud puts the average cost at 5–15% of a referral program budget.

Why reward timing matters

An instant reward pays out before a chargeback, refund, or fake order can surface. Once a fraudster has the cash or credit, recovering it is slow and often impossible. If you pay after the return window closes, you can cancel the reward on any order that gets refunded or disputed. Honest advocates wait a bit longer and lose nothing.

The score reflects this. Timing moves your total by up to 16 points, from instant payout to manual approval. It also compounds with value: a $250 reward paid instantly stacks the two largest weights in the model. If you plan to raise the reward, move payout later first.

Signs of referral program abuse

A low score means your controls are in place. You still need to watch for abuse once referrals flow. Investigate when you see:

• A spike in referrals or payouts with no campaign, press, or seasonal cause.
• Advocate and friend pairs that share an IP address, device, card, or shipping address.
• Friends whose only order is the referral order, with no repeat purchase.
• Your referral code on a coupon or deal site, followed by redemptions from unrelated accounts.
• Rising referral volume alongside rising refunds or chargebacks on referred orders.
• Bursts of new accounts from one network, or referral timestamps too regular for a person.

One signal alone proves little. Two or three together justify holding payouts for review. Then rerun the checker to find the control that should have caught it.

What to do next

1. Fix the high-severity rules first, then medium, then low. Give each rule an owner.
2. Copy the checklist or download it as Markdown or text, and share it with whoever monitors fraud.
3. Rerun the checker before launch, when you raise the reward, after an unexplained payout spike, and once a quarter.

These free Talkable tools cover the rest of the program:

• Referral Reward Calculator: size the advocate and friend rewards, then check the fraud risk of that value here.
• Referral ROI Calculator: model revenue, reward cost, and payback before you commit a budget.
• Referral Program Checklist: cover setup, terms, and tracking before go-live.

Talkable includes fraud prevention in its referral platform. Its features page reports that its fraud algorithm has blocked over 5 million fraudulent attempts, which it equates to over $110M in savings for customers.