CashoutGuard: fraud prevention for rewards, GPT, faucet, PTC and offerwall sites.

Glossary of rewards, GPT and offerwall fraud terms

The rewards industry has its own vocabulary, and fraud prevention adds another. This glossary explains the terms you meet when you run a rewards, GPT or offerwall business, in plain words and with the reason each one matters.

Short answer

A GPT (get-paid-to) site pays users small rewards for completing offers, surveys, games or tasks, and an offerwall is the list of paid offers a network supplies to it. Fraud on these sites is mostly multi-accounting, VPN and residential proxy traffic, emulators, phone farms and bots, which advertisers take back weeks later as reversals.

Postback
The server-to-server call from the network that says a user completed an offer.
Reversal or chargeback
A conversion the advertiser or network takes back after validating it.
Shadow mode
Running the fraud checks without acting on them, to measure before blocking.
Strongest single signal
A payout address shared by several accounts.

The business

  • GPT site. "Get paid to" site. A website or app that pays users small rewards for completing offers, surveys, games or tasks, and earns from the advertisers behind them.
  • Rewards app. The mobile version of a GPT site: users earn points by playing games or trying apps and cash out to PayPal, gift cards or crypto.
  • Offerwall. A list of paid offers supplied by a network and shown inside a rewards site or app. The network pays the site for each completed offer.
  • Offerwall network. The company that sells offers to publishers and gets paid by advertisers. It sits between the two and absorbs reversals when fraud is found.
  • Publisher. The site or app that shows an offerwall to its users. A network has many publishers.
  • PTC site. "Paid to click." A site that pays users to view ads for a set number of seconds.
  • Faucet. A site that pays tiny amounts of cryptocurrency for claims made at regular intervals.
  • Playtime offer. An offer that pays users for reaching milestones or time played in a mobile game.
  • Survey router. A service that sends users to market-research surveys and pays for completes that the panel accepts.
  • Geo-targeting. Offers that pay only for users in certain countries. The main reason VPN and proxy traffic gets reversed: the advertiser paid for a country the user was never in.

How the money moves

  • Postback. A server-to-server call from the network to the site saying that a user completed an offer and how much it pays. The site credits the user when it arrives.
  • Click ID. A unique identifier attached to an offer click and returned in the postback, so the conversion can be matched to the click.
  • Conversion. A completed offer confirmed by a postback.
  • Reversal or chargeback. A conversion the advertiser or network takes back after validating it, usually because it was fraud or did not meet the offer rules. It arrives days or weeks later.
  • Cashout or withdrawal. The moment a user converts points into money or a gift card. The last point where the site can stop paying a fraudster.
  • Hold. Keeping a reward or a withdrawal pending until it is reviewed, instead of paying it immediately.
  • Payout address. Where the user is paid: a PayPal email, a crypto wallet, a gift card email. Shared payout addresses are one of the strongest fraud signals.

Types of fraud

  • Multi-accounting. One person running many accounts to collect bonuses, referral rewards or offers several times.
  • Farm. An organised operation that runs many accounts at once, on real phones, emulators or scripts, to extract rewards at scale.
  • Phone farm. A rack of real mobile devices used to complete app installs and offers for many accounts.
  • Bonus abuse. Creating accounts to claim sign-up or referral bonuses without any intention of using the service.
  • Self-referral. Referring your own accounts to collect referral commissions.
  • Incentive abuse. Completing offers only as far as needed to get paid, such as installing and immediately uninstalling an app.
  • Forged postback. A fake conversion call sent to the site, pretending to come from the network.

Signals and techniques

  • Device fingerprint. A set of characteristics of a browser or phone (screen, graphics, fonts, audio and more) combined into an identifier that survives cookie deletion.
  • Persistent device ID. An identifier stored on the device so the same browser is recognised on the next visit. Clearing it and coming back is itself a signal.
  • Linked accounts. Accounts connected by a shared device, payout address, email or network pattern. The basis of catching multi-accounting.
  • VPN. A service that routes traffic through another server, hiding the real IP and country.
  • Residential proxy. A proxy that uses real home connections, so the IP looks like an ordinary user. Invisible to IP blocklists; exposed by timezone, language and WebRTC mismatches.
  • Datacenter IP. An address that belongs to a hosting provider rather than a home or mobile network. Real users rarely browse from one.
  • Tor. An anonymity network. Traffic exits from known Tor exit nodes.
  • WebRTC leak. A browser feature that can reveal the real IP behind a VPN or proxy.
  • Emulator. Software that imitates a phone on a computer. Used to run many app accounts without physical devices.
  • Automation. Scripts and headless browsers that click, tap and fill forms without a human.
  • Offer speed. The time between an offer click and its conversion. Much faster than the offer allows means a bot or a shortcut.

Metrics

  • Points per dollar. How many site points equal one dollar of network revenue. It sets how much of each payout the site shares with its users.
  • eCPM. Effective revenue per thousand users or impressions. The usual way to compare offerwalls and ad formats.
  • Validation window. The period during which an advertiser can still reverse a conversion. Holding rewards until it closes removes most chargeback risk.
  • Reversal rate. The share of conversions taken back after validation. Networks watch it per publisher; advertisers watch it per network.
  • Flagged share. The share of events or accounts a fraud check marks for review or block. Useful to compare publishers, countries or traffic sources.
  • Monthly active users. Distinct accounts seen in a month. Fraud tools usually price by it.
  • Velocity. How many actions an account or device performs in a short time: sign-ups per hour from one network, conversions per hour, cashouts per day.

Decisions and modes

  • Risk score. A number, usually 0 to 100, that sums the weight of the fraud signals found on an event or account.
  • Allow, review, block. The three decisions a fraud check returns: pay normally, hold for a person to look, or stop.
  • Shadow mode. Running the fraud checks without acting on them, to measure the problem and tune the rules before blocking anyone.
  • Enforce mode. Acting on the decisions: refusing, holding or stopping what the check flags.
  • Fail-open. When the fraud check is slow or unreachable, letting the user through instead of blocking everyone.
  • False positive. An honest user flagged as fraud. The cost that makes aggressive rules expensive.
  • Allowlist and blocklist. Lists of accounts, devices, IPs or payout addresses that are always allowed or always blocked, whatever the score.

Where to go next

The multi-accounts guide and the VPN guide go deeper into the signals, the chargebacks guide explains reversals, and the fraud cost calculator puts a number on what it costs your site.

Frequently asked questions

What is the difference between a GPT site and an offerwall?

The GPT site is the business that has the users and pays them. The offerwall is the list of paid offers a network supplies to that site.

Why do reversals arrive so late?

Advertisers validate installs and sign-ups after the fact, often at the end of a billing period, so fraud is found days or weeks after the site already credited the user.

Is a VPN user always a fraudster?

No. Many people use VPNs for privacy. A VPN is one signal; it becomes strong evidence together with shared devices, shared payout addresses or geo-targeted offers.

What is the strongest single fraud signal?

A payout address shared by several accounts. It is rarely innocent and it points directly at the money.

Keep reading