The High Rates Small Loan Model

At a Glance
Fraud and eligibility controls determine whether the Lender can lend. During the pilot, the IBV Analytics Scorecard recommends whether the first-loan offer should be $500, $400, or $300. It is a loan-sizing input, not a stand-alone approval or decline engine. Clean files move directly to funding after reachability, account, and repayment-method controls pass. Human review is reserved for defined exceptions.
Speed-First
Fraud Is the Hard Gate
$500 / $400 / $300
Humans Handle Exceptions
Scorecard Requires Validation

Core Operating Rule
Fraud and eligibility determine whether the Lender lends. A validated IBV Analytics Scorecard helps determine how much the Lender lends. Reachability, account, and repayment-method controls determine whether a file can fund straight through. Operations determine how quickly it funds. Collection Multiple determines whether the model works.

Policy Foundation
Risk the Lender can price
Repayment volatility and ordinary credit loss are expected in a high-APR, low-principal product. The interest rate prices the frequency of repayment losses.
Risk the Lender cannot simply price
Fraud, identity mismatch, unauthorized payment methods, legal ineligibility, and an unusable repayment account are hard failures.
Exposure controls loss severity
The interest rate prices how often borrowers may pay poorly. The $500, $400, and $300 tiers control how much principal is exposed when they do.
What the current evidence supports
Pre-funding bank activity has shown moderate value for loan sizing, particularly recent income, direct-deposit dollars, and financial-obligation payments. It has not been validated as a stand-alone approval or decline gate. The IBV Analytics Scorecard must prove that it improves the Lender’s own seasoned economic results before it receives decision authority.

Three Conditions Before Full Production
1. Validate the scorecard
Confirm the score direction, range, reason codes, model version, and proposed amount bands. Run the scorecard in shadow mode before it independently changes approval or decline decisions.
2. Replace the reachability function of the live call
A clean file must still confirm real-time reachability through a one-time passcode, verified borrower response, automated verification call, or brief human verification touch. A full underwriting call is not required, but reachability cannot be silently removed.
3. Make the $400 tier earn its place
Do not assume $400 is automatically the correct midpoint between $300 and $500. Track the $400 band separately. If its seasoned Collection Multiple underperforms both adjacent tiers, narrow it, redesign it, or remove it.

Decision Flow

1
Automated Fraud, Eligibility, and Reachability Controls
These checks determine whether the application can continue and whether it can proceed without a full live underwriting call. Complete them in seconds wherever possible.
  • Identity and application consistency
  • Phone, email, device, and duplicate checks
  • Real-time reachability through OTP, verified response, automated verification call, or brief verification touch
  • State, product, employer, and legal eligibility
  • Bank-account ownership or appropriate association
  • Repayment-method ownership and authorization
  • Internal fraud or confirmed bad-actor history
Fraud or eligibility failure: hold for resolution or decline. A reachability failure by itself routes the file to exception review unless another policy requires decline.

2
Run the IBV Analytics Scorecard
The scorecard is a standard loan-sizing step, not merely a resolution tool. During shadow mode and the initial pilot, it recommends exposure but does not independently approve or decline. Preserve the exact response that existed when the decision was made.
  • Store the original scorecard and bank-data response
  • Store the score, score band, reason codes, and model version
  • Store the exact repayment checking account
  • Store application and decision timestamps in UTC
  • Store the amount eligible, recommended, offered, accepted, and funded
  • Record straight-through decisions, human reviews, and manual overrides
Current retrospective evidence supports the scorecard’s use as a sizing and confidence layer. Production authority must be earned through the prospective pilot.

3
Assign the Proposed $500, $400, or $300 Offer
Every eligible first-time applicant may qualify for up to $500. During the pilot, the IBV Analytics Scorecard recommends the appropriate exposure tier. The Lender retains the exception and override controls below.
Scorecard ResultProposed OfferTreatment
Tier A or strongest acceptable band$500Straight-through candidate when all controls pass
Tier B or middle band$400Provisional tier; validate separately and retain only if its economics justify it
Tier C or acceptable lower band$300Fast offer for a legitimate borderline borrower
Score unavailable, incomplete, or timed out$300 maximumException review when needed; never fail open to $500
Below the proposed minimum bandManual review during pilotDo not automate a score-only decline until the scorecard is prospectively validated and the reason is specific and documented
Numerical cutoffs should be added only after the provider confirms the score direction, range, recommended bands, and reason codes. The Lender should then recalibrate every tier using its own seasoned, cost-adjusted Collection Multiple. The $400 tier must be evaluated as its own product treatment rather than assumed to be a linear midpoint.

4
Fund the Clean File
  • Confirm real-time reachability through OTP, verified response, automated verification call, or brief verification touch
  • Present the final $500, $400, or $300 offer
  • Complete the agreement and required disclosures
  • Validate and authorize the repayment method safely
  • Fund immediately when ready-to-fund status is reached
A full live underwriting call is not required for a clean file, but reachability must be affirmatively confirmed.

5
Use a Human Only for a Defined Exception
Human review should resolve the named exception. It should not restart the entire application. Any concern about inconsistent or coached answers must identify the specific contradiction rather than rely on an undocumented impression.

Human Review Rules
Require human review when
  • The score is missing, incomplete, unavailable, or materially inconsistent with the underlying data
  • Real-time reachability cannot be confirmed automatically
  • An identity, phone, device, duplicate, or fraud alert appears
  • Application information materially conflicts with bank or employment information
  • A brief verification produces a specific unresolved inconsistency
  • The linked account appears secondary or unusable
  • Income or recurring inflows cannot be reasonably identified
  • Repayment-method ownership does not match
  • The borrower requests assistance
  • An employee requests an amount override
Do not require a full call merely because
  • The borrower uses Cash App, Venmo, PayPal, or Zelle
  • The account contains fees or an occasional NSF
  • The borrower is paying other lenders
  • The profile is risky but appears legitimate
  • The applicant qualifies only for $300
  • An employee wants to inspect every transaction manually

Safe Defaults and Overrides
Safe defaults
  • The maximum potential first-loan amount is $500
  • A missing score never produces an automatic $500 offer
  • A legitimate borderline borrower defaults toward $300
  • A score below the proposed minimum routes to review during the pilot, not an automatic score-only decline
  • A reachability failure routes to exception review unless another policy requires decline
  • A fraud or legal failure is held or declined, not moved into a lower tier
Override policy
  • An employee may reduce the offer by one tier with a documented reason
  • Increasing above the assigned tier requires manager approval
  • Every override must use a standardized, specific reason code
  • “Gut feeling,” “felt coached,” or another undocumented impression is not an acceptable reason
  • Performance must be reported with and without overrides

KPI Scorecard
Funnel and Speed
  • Raised hands answered within 15 minutes
  • Real-time reachability confirmation rate
  • OTP or automated verification completion rate
  • Approved-to-funded rate
  • Scorecard-to-offer time
  • Application-to-funding time
  • Straight-through funding rate
  • Human-intervention rate
  • Applicant abandonment rate
Fraud and Early Risk
  • Confirmed fraud loss per funded loan
  • Identity or account mismatch rate
  • Repayment-method validation failure
  • ACH and debit return rate
  • Payment revocation rate
  • FPD by score and amount band
  • Missing-score rate
  • Exception rate by reason
Economic Outcomes
  • 60-day Collection Multiple
  • 90-day Collection Multiple
  • 120-day Collection Multiple
  • Delta greater than 1.00 Good Outcome Rate
  • Gross collection surplus or deficit
  • Cost per Good Outcome
  • Performance by $500, $400, and $300 amount band
Scorecard Validation
  • Good Outcome Rate and Collection Multiple by score band
  • Incremental lift compared with the control process
  • Calibration: whether observed performance matches the predicted ordering
  • Separate $400-tier performance versus both $300 and $500
  • Manual-override performance versus scorecard recommendation
  • Percentage of decisions supported by a complete frozen decision-time payload
Primary economic definition
Collection Multiple equals Total Collected divided by Loan Amount. Delta greater than 1.00 is principal-positive, but it is not necessarily net profitable after funded-lead cost, payment expense, payroll, and collections expense.

Launch and Validation Plan
  1. Confirm what the IBV Analytics Scorecard is measuring, its direction, range, reason codes, model version, and proposed $500, $400, and $300 cutoffs.
  2. Implement the scorecard in shadow mode and preserve every decision-time response.
  3. Implement the real-time reachability control that replaces the mandatory full live call.
  4. Pilot straight-through processing on a controlled share of clean applications.
  5. Use the current underwriting process as the comparison group during the pilot.
  6. Lock the cutoffs and policy rules before economic outcomes become visible.
  7. Do not use the scorecard as an automatic decline gate during shadow mode or the initial pilot.
  8. Evaluate the $400 tier separately. If it does not justify its exposure, narrow it, redesign it, or remove it.
  9. Review the first 300 funded loans after consistent 60-, 90-, and 120-day seasoning.
  10. Adjust the amount bands using cost-adjusted Collection Multiple, not FPD alone.
  11. Expand straight-through funding only when conversion improves without an unacceptable increase in fraud loss, payment failures, or economic losses.

Final Policy Statement
The Lender will use an automated, score-tiered, exception-based underwriting model. Every eligible applicant may qualify for up to $500. Fraud and eligibility controls determine whether the Lender can lend. During the pilot, the IBV Analytics Scorecard recommends whether the first-loan offer should be $500, $400, or $300, but it does not independently approve or decline. Clean files with confirmed reachability, a validated repayment account, and an authorized payment method may fund straight through. Human review is limited to specific exceptions. Borderline but legitimate borrowers should generally receive a fast $300 offer rather than be lost to unnecessary friction. The $400 tier must independently justify its place. Performance will be managed using seasoned, cost-adjusted Collection Multiple, Good Outcome Rate, fraud loss, and cost per Good Outcome.

This operating model remains subject to applicable lending, eligibility, ability-to-pay, adverse-action, disclosure, privacy, payment-authorization, and fair-lending requirements. Numerical score cutoffs must be documented, consistently applied, and validated against the Lender’s own seasoned economic results. Any decline or amount reduction must use a specific reason supported by the actual information and policy applied.

Leave A Comment?

This site uses Akismet to reduce spam. Learn how your comment data is processed.