Screen loan leads for fraud and fit before they reach you

CYBERTECH applies layered checks before handoff, against the criteria agreed for your market and product. No single signal decides whether an applicant moves forward, and a stopped applicant is not delivered.

Updated 29 October 2026

Screening happens before the handoff

CYBERTECH screens applicants before the handoff. The lender defines the market, the product criteria and the capacity. Applicants are checked against those rules and the four integrity checks, and only suitable applicants are routed to the lender, who keeps the final credit decision.

Applicants arrive from paid media, search, mobile products and app stores, CRM and referral traffic. Which checks apply in full depends on the channel. Each applicant carries the source it arrived from, which is what allows outcomes to be compared by source later.

The patterns the checks look for are explained in the fraud guide. This page covers what is checked, what is stopped and what stays with you.

Five checks cover integrity and fit

The first four checks each read a different layer of the application, and none of them is a verdict on its own. The fifth applies the lender's rules.

CheckWhat it readsWhat it surfaces
Device integrityWhether the application comes from a physical device in a normal state, as far as the channel exposes it: emulator signatures and tamper or compromise indicators.Emulators, device farms and modified clients. See emulators and device farms.
NetworkWhere the connection comes from, such as hosting, data-centre or anonymising services, and whether it fits the location the applicant states.Hosted or anonymised traffic, and location claims that the connection contradicts.
IdentityConsistency between the details an applicant gives, and duplicate applications.Recycled and synthetic identities, and duplicate applications. See recycled identities.
Behaviour and velocityHow the application session runs, and how many applications come from one device, identity or source in a short window, a velocity check.Scripted sessions, repeat applications and bursts from a single source. See velocity and behaviour signals.
Fit to your criteriaThe applicant's stated profile and chosen product against the rules the lender has set for that market and product.Genuine applicants who sit outside your rules. They are not delivered to you.

Two patterns cut across the checks. Loan stacking, where one borrower takes out several loans from different lenders in a short window, is only partly visible before handoff. Repeat applications can show in the velocity check, while loans already granted elsewhere stay with your underwriting and credit checks. Incentivised traffic, where people are rewarded for installing or filling in a form, is judged by source, using the outcomes described below.

No single integrity signal decides

Integrity signals are read together. A mismatch between the IP location and the stated city can have an ordinary explanation. The same mismatch next to an emulator signature and a scripted session does not.

Take a fictional applicant who reports a Pixel 7 in Delhi. The connection comes from a data centre far from Delhi and the session runs like a script. Either could have an innocent cause on its own, such as a corporate VPN or browser autofill. Add an emulator signature and the applicant is stopped. A second fictional applicant has a clean device and a residential connection in Almaty, but the same phone number arrives from several other devices in a short window, in what looks like one scripted burst. Behaviour and velocity together stop that one.

A scripted application is valid input, which is why OWASP describes automated threats as often being misuse of valid functionality rather than exploitation of a vulnerability. What gives it away is how it arrives and how often. The evidence behind each signal is in the guides on emulators and device farms and velocity and behaviour signals.

Criteria are agreed per market and product

Before launch, the lender and CYBERTECH agree the rules each applicant is checked against. They are set per market and per product, because eligibility, serviceable areas and handoff differ between, for instance, a personal loan in India and a consumer loan in Kazakhstan, and the same holds in Armenia, Nigeria and Peru. Supply varies by market and product, and each launch is confirmed market by market. What is agreed at onboarding:

  • Eligibility: the profile limits the product accepts, such as an age band or serviceable area.
  • Product fit: which loan product the applicant is applying for, and where it is offered.
  • Capacity: the daily or monthly cap per market and product, after which delivery is throttled. See pricing.
  • Handoff: referral or direct API, and the fields passed with each qualified application. See integration.
  • Feedback: which outcome states the lender returns by postback.

The criteria are the lender's and work as hard rules. An applicant outside the agreed profile limits, serviceable area or product definition is not delivered, however clean the other signals are. If your policy changes, the rules change with it. CYBERTECH does not set them.

A stopped applicant is not delivered

An applicant is stopped when a hard rule fails, or when the integrity signals, read together, point to automation, duplication or an inconsistent application. A stopped applicant is not delivered to the lender. It is not a qualified application, so it cannot reach the billable event agreed before launch.

There are two kinds of stop. A fit stop applies to a genuine applicant who is outside your rules and says nothing about integrity. An integrity stop applies whatever the fit. Neither reaches your decision flow.

A qualified application takes one of two routes. By referral, through a branded redirect or hosted portal, the click and lead parameters are signed. By direct API, the application is posted in real time as JSON, mapped to the lender's schema. Parameters, field mapping and postback setup are agreed before launch, and the routes are described in the integration overview.

Postbacks feed back into the checks

Where the integration allows it, partner postbacks return application and outcome states. Compared by source, those outcomes help tune the checks over time. The mechanics are in the postback guide.

Which outcome states the lender returns, such as approval or disbursal, and at what level of detail, is agreed at onboarding, within what the lender's own borrower consent and confidentiality obligations allow.

You can check the screening against your own data

Whatever a provider reports, the test is how delivered applicants behave in your own systems. Four measures cover it:

  • Duplicates: the share of delivered applicants whose phone, email or device already exists in your database.
  • Funnel by sub-id: approval rate and disbursal rate for each source, set against your other acquisition channels.
  • First-payment default by source, which can only be read once the first payments fall due.
  • Mismatch with your rules: delivered applicants that your own flow rejects on age, location or product, which points to a gap in the criteria rather than fraud.

Lead quality metrics explains how to read the funnel and first-payment default by source; duplicate rules are covered under recycled identities.

Screening does not replace underwriting, KYC or the credit decision

  • It does not make a credit decision. The lender approves or declines, and a delivered applicant is a qualified application, not an approved loan.
  • It is not underwriting. Creditworthiness, bureau checks and affordability stay with your own process.
  • It does not replace your KYC or compliance. The identity checks here are integrity checks, not a regulated KYC process, and your own obligations still apply to every delivered applicant.
  • It does not promise a rate. No fraud-catch, approval or conversion rate is implied, and outcomes depend on the market, the product and your own decision flow.

Sources

  1. OWASP: Automated Threats to Web Applications (2026)

Agree the criteria for your market

Tell us the product, geography and handoff model. We will agree the criteria, the handoff and the required fields with your team.

Discuss an anti-fraud setup