Fraud Loss Rate vs Approval Rate
ARTICLE — PUBLICATION-READY COPY Fraud Loss Rate vs Approval Rate: Finding the Point Where Risk Controls Create More Value Than They Destroy Consider two fraud policies. Policy A produces lower fraud loss but rejects more transactions. Policy B permits somewhat more fraud but app
ARTICLE — PUBLICATION-READY COPY
Fraud Loss Rate vs Approval Rate: Finding the Point Where Risk Controls Create More Value Than They Destroy
Consider two fraud policies.
Policy A produces lower fraud loss but rejects more transactions.
Policy B permits somewhat more fraud but approves substantially more profitable, legitimate customers.
Which policy is better?
Fraud rate cannot answer that question.
Approval rate cannot answer it either.
The correct comparison is economic:
Which policy retains more legitimate contribution after fraud losses, disputes, review costs, fraud-control costs, customer-support costs, and other material consequences of the decision?
That changes how merchants should think about fraud loss rate vs approval rate.
The goal of fraud operations is neither minimum fraud nor maximum approval.
It is:
Maximum sustainable economic value from legitimate commerce after the cost of fraud and the controls used to manage it.
That distinction matters because the easiest way to reduce many forms of payment fraud is also commercially useless: stop accepting uncertain customers.
An ecommerce merchant that approves nothing will have very little approved-payment fraud.
It will also have very little business.
The real work begins between those extremes.
Visa's current merchant fraud-management material explicitly frames the objective in terms of accepting more good orders while managing fraud risk, rather than simply driving fraud toward zero. Mastercard likewise presents reducing false declines, preventing fraudulent transactions, and increasing approvals as related decisioning objectives.
The merchant therefore needs one economic model that Fraud, Payments, Product, and Finance can use together.
The easiest way to reduce fraud is to reject more
Suppose a fraud team tightens a threshold.
Before the change:
more uncertain transactions proceed;
some are legitimate;
some later produce confirmed fraud.
After the change:
more transactions are blocked;
some fraud disappears;
some legitimate customers disappear with it.
If the newly blocked population contains disproportionately fraudulent activity, the merchant's observed fraud loss rate can improve.
That still does not prove the change created value.
The missing question is:
What happened to the legitimate transactions inside the same blocked population?
A fraud control should not receive credit for preventing $1 of expected fraud while escaping scrutiny for blocking $10 of profitable legitimate contribution.
This is the central weakness in optimizing fraud rate by itself.
A lower fraud rate can mean:
better discrimination between legitimate and fraudulent activity
or:
more aggressive rejection of both.
Those are very different accomplishments.
Why fraud loss rate alone fails
A useful merchant fraud-loss metric might be expressed as:
Confirmed fraud loss
÷ defined transaction-value base
× 100
The denominator could be captured value, settled value, completed-order value, or another consistently defined population.
The definition matters.
So does timing.
Fraud outcomes can mature after the transaction through:
disputes;
investigations;
refunds;
confirmed account compromise;
other post-transaction evidence.
A very recent cohort may therefore appear unusually clean simply because its losses are incomplete.
But even a perfectly defined, fully matured fraud-loss rate answers only one side of the commercial question:
How much fraud made it through?
It does not tell leadership:
How much legitimate business did the fraud controls prevent from getting through?
Both belong in the decision.
Why approval rate alone fails too
The opposite mistake is equally dangerous.
A merchant can usually increase acceptance by relaxing controls.
That may produce:
more legitimate sales;
more fraudulent sales;
more disputes;
higher fulfillment loss;
increased support or investigation cost;
greater downstream payment risk.
An approval increase is therefore not automatically evidence of payment optimization.
The merchant needs to know which customers were added at the margin.
Were the incremental approvals profitable legitimate orders?
Or did the looser policy simply admit economically unattractive fraud?
That is why neither team should have unilateral authority to define success as:
Fraud: “Our fraud rate fell.”
or:
Payments: “Our approval rate rose.”
Leadership needs a common economic objective.
First fix the word “approval”
One reason fraud rate vs approval rate debates become confused is that “approval” can refer to several different stages.
A payment can fail before an issuer ever evaluates it.
A useful merchant model separates at least three concepts.
Metric
What it measures
Example denominator
Merchant risk acceptance
Transactions allowed to proceed after merchant-controlled fraud decisions
Eligible transactions evaluated by merchant fraud controls
Issuer authorization rate
Authorization attempts approved by the issuer
Transactions actually sent for authorization
End-to-end payment success
Defined customer payment attempts that ultimately complete successfully
Merchant-defined eligible payment attempts
These are not interchangeable.
A transaction can fail because:
the merchant blocked it;
a fraud provider blocked it;
authentication failed;
PSP configuration prevented processing;
the issuer declined authorization;
processing failed;
the customer abandoned.
The exact payment architecture varies, but the control-boundary principle is constant:
Do not attribute a lost payment to the issuer when the merchant never sent it to the issuer.
This becomes especially important after fraud-policy changes.
If the merchant tightens a pre-authorization rule and overall payment success falls, saying “authorization deteriorated” may hide the fact that fewer legitimate transactions were permitted to reach authorization at all.
Merchant fraud blocks belong in the economic model
A useful operational measure is:
Merchant fraud-block rate
=
transactions rejected by merchant-controlled fraud decisions
÷ eligible transactions evaluated
This is not the same thing as false-positive rate.
Some blocked transactions genuinely are fraudulent.
Some have unresolved outcomes.
Some may be legitimate.
The block rate tells the merchant how aggressively its own controls are acting.
The next question is what those decisions cost.
A false positive has an economic value, not just a count
“False positive” sounds like a model-quality term.
For a merchant, it can also be a commercial event.
Suppose a legitimate customer is incorrectly classified as risky.
The consequence depends on the action.
If the transaction is rejected
The merchant may lose:
current-order contribution;
future customer value;
acquisition spend already incurred;
support cost caused by the rejection.
If the transaction is challenged
The customer may complete successfully.
Or the additional friction may cause abandonment.
If it is sent to manual review
The order may eventually be approved, but the merchant incurs:
reviewer labor;
fulfillment delay;
possible customer dissatisfaction.
So a “false positive” should not automatically be valued as:
full order value lost.
The correct cost depends on the decision and what happens afterward.
Visa's current merchant materials explicitly connect fraud decisioning with order acceptance and false-decline reduction, reinforcing why security outcomes and legitimate-payment outcomes should be evaluated together.
Fraud loss rate is not chargeback rate
This distinction is essential.
A fraud loss rate attempts to measure fraud-related financial loss against a defined merchant transaction base.
A chargeback rate measures chargebacks against whatever transaction population the applicable merchant, PSP, acquirer, or card-network methodology defines.
They overlap.
They are not synonymous.
Some fraudulent activity may never become a chargeback.
And not every chargeback represents stolen-payment fraud. Visa's current merchant guidance includes disputes involving fraud, unrecognized charges, processing issues, and product or service problems.
If leadership uses chargeback rate as a complete fraud-loss metric, the economic model can become distorted.
Keep:
fraud
disputes
and:
chargebacks
separately measurable.
Combine their financial effects only where the decision model requires it.
Gross fraud is not necessarily net financial loss
Suppose a fraudulent transaction for $100 is confirmed.
Calling the merchant's economic loss exactly $100 may be wrong.
The merchant may also have:
fulfilled inventory;
shipping expense;
dispute costs;
investigation cost.
Or it may later recover part of the value.
A useful internal distinction is therefore:
Gross confirmed fraud
The defined transaction value associated with confirmed fraudulent activity.
Net fraud-related financial loss
A merchant-defined economic measure incorporating relevant:
fraud principal
unrecoverable fulfillment/service cost
attributable dispute cost
relevant operating cost
− recoveries
The exact components vary by business.
This is a merchant management framework, not a universal card-industry formula.
The important thing is consistency.
Revenue is not the same thing as economic value
This is where many fraud-versus-approval discussions go wrong.
Suppose relaxing a rule recovers $100,000 of legitimate sales.
Did the merchant create $100,000 of economic value?
Usually not.
That is transaction value or revenue—not necessarily contribution.
If the merchant earns only a fraction of that value after:
cost of goods;
fulfillment;
commissions;
payment costs;
variable servicing cost;
then the fraud decision should be evaluated against that lower economic contribution.
For a high-margin digital product, recovering another legitimate transaction may be highly valuable.
For a low-margin physical product with expensive fulfillment, the same gross payment value can have very different economics.
Therefore:
Risk thresholds should ultimately be judged against the economic value of the incremental legitimate transaction—not simply its face value.
The Net Revenue Risk Model
A practical fraud decision can be framed as:
Net expected value of a policy change
Incremental legitimate contribution
− incremental fraud loss
− incremental dispute cost
− incremental manual-review cost
− incremental fraud/authentication/control cost
− incremental customer-support and other attributable operating cost
=
incremental net economic value
This is the article's secondary framework.
It is intentionally incremental.
The question is not:
“How profitable is our whole business under this threshold?”
It is:
“What additional value or loss occurs if we move from the current policy to the proposed one?”
That is a much better threshold decision.
Average performance can hide the decision that matters
Assume the merchant currently approves 90% of a transaction cohort.
Management is considering changing the threshold so that it approves 92%.
The relevant economic population is not primarily the original 90%.
Those transactions were already accepted under either policy.
The decision lives largely inside the incremental population between the two thresholds.
Ask:
How much incremental legitimate business will be recovered?
How much incremental fraud will enter?
What additional disputes will result?
Does review volume change?
Does customer-support workload change?
What contribution does the recovered legitimate business produce?
This is marginal economics.
And marginal economics—not the average fraud rate across the full portfolio—is what should determine whether the threshold moves.
An illustrative threshold comparison
The following numbers are purely hypothetical.
Suppose two policies operate on the same mature cohort.
Policy A
Policy B
Merchant acceptance
84%
88%
Matured fraud loss rate on accepted value
0.35%
0.50%
Relative posture
Stricter
Looser
Policy A looks better if the only objective is fraud rate.
Policy B looks better if the only objective is acceptance.
Neither conclusion is sufficient.
Suppose Policy B produces, relative to A:
Incremental legitimate contribution: +$12,000
Incremental fraud loss: −$1,460
Incremental dispute, review and support cost: −$1,000
Then:
Incremental net value = +$9,540
Under those assumptions, Policy B creates more economic value despite producing the worse fraud metric.
Now change one business assumption.
Suppose the incremental legitimate transactions produce only $2,000 of contribution because the merchant sells a low-margin product.
The same risk increase now creates:
$2,000 − $1,460 − $1,000 = −$460
Policy A becomes economically preferable.
Nothing about the fraud model had to change.
The merchant economics did.
That is why a universal “optimal fraud threshold” does not exist.
The Fraud–Approval Decision Frontier
The article's primary framework is the Fraud–Approval Decision Frontier.
Imagine a chart.
Horizontal axis
Legitimate acceptance
Moving right means allowing more legitimate commerce through.
Vertical axis
Expected fraud exposure
Moving upward means accepting more expected fraud cost.
Different thresholds, rules, review strategies, authentication policies, and fraud models produce different points on that plane.
A very restrictive policy may sit low and left:
low fraud, but low legitimate acceptance.
A very permissive policy may sit high and right:
high legitimate acceptance, but higher fraud exposure.
The merchant's job is not to find the lowest point.
It is to find the economically acceptable operating territory.
Not every point on the frontier is equally efficient
Suppose Policy C and Policy D produce the same fraud exposure.
But Policy D approves more legitimate customers.
Policy C is economically dominated.
Likewise, if two policies produce the same legitimate acceptance but one produces materially more fraud, the higher-fraud policy has no obvious economic advantage.
Good fraud optimization tries to move the merchant toward a better frontier:
more legitimate acceptance for the same fraud exposure, or less fraud for the same legitimate acceptance.
Technology, data, authentication, better rules, stronger models, and useful human review can sometimes shift the frontier.
Simply loosening or tightening a threshold usually moves the merchant along the existing tradeoff.
That distinction matters.
The economically optimal point can still sit inside hard constraints
Pure expected-value maximization is not the merchant's only concern.
Risk appetite can impose boundaries.
A merchant may decide that certain operating points are unacceptable because they create excessive:
fraud volatility;
financial exposure;
disputes;
operational backlog;
customer harm;
acquirer or provider risk;
other material constraints.
Those become guardrails around the frontier.
So the practical objective is:
Find the highest-value operating point inside the merchant's acceptable risk boundaries.
Risk appetite should therefore be expressed as a set of decision constraints—not merely:
“We are conservative.”
Risk appetite is where Fraud, Payments, and Finance meet
Fraud may understand the loss distribution.
Payments may understand authorization and conversion.
Finance understands margin and realized financial loss.
Product understands customer friction.
Risk appetite converts those perspectives into a common operating policy.
For example:
Fraud may propose tighter controls because incremental fraud has risen.
Finance may discover that the incremental legitimate sales still produce several times the additional loss.
Payments may show that a specific merchant rule—not issuer behavior—is causing the approval deterioration.
Product may identify abandonment created by a new verification step.
The correct decision is not determined by departmental seniority.
It should be determined by the agreed economic model and risk boundaries.
False declines should be treated as a hypothesis to measure
Merchants often know fraud losses better than false-positive losses.
Fraud creates visible downstream evidence.
A legitimate customer who was blocked may simply leave.
That creates an asymmetric dataset:
fraud becomes labelled
while:
lost legitimacy may remain invisible.
So merchants need multiple false-positive indicators.
Useful evidence can include:
transactions manually approved after initial risk escalation;
legitimate customers confirmed through support;
successful permitted retries;
rule reversals;
later legitimate activity;
customer complaints associated with fraud blocks.
No single signal establishes a perfect false-positive rate.
The merchant should define:
population
decision being evaluated
definition of legitimate
observation window
treatment of unresolved cases
before publishing the metric.
Manual review belongs in the economics too
Moving a risk threshold from:
reject
to:
manual review
can recover legitimate approvals.
But review is not free.
The economic comparison becomes:
legitimate contribution recovered through review
− fraud admitted after review
− review labor
− delay/customer cost
− other incremental operational cost
Manual review is valuable when human information can materially improve decisions enough to justify those costs.
If a merchant sends a huge uncertain population to reviewers simply to avoid making a threshold choice, it may replace a false-decline problem with an operational bottleneck.
Detailed queue design belongs in Manual Review Best Practices.
For this article, review cost simply belongs on the same economic ledger as fraud and approvals.
Thresholds should be evaluated at the margin
Consider a fraud score from 0 to 100.
Suppose the merchant currently rejects above 80.
Changing the threshold to 78 does not primarily require re-evaluating transactions scoring 20.
The economically interesting population is near the changing decision boundary.
What happened to transactions around:
78–80?
Among them:
how many became confirmed fraud?
how many were legitimate?
what was their contribution?
what dispute cost followed?
what review or support workload was created?
That is the population that tells the merchant whether the threshold should move.
This principle applies even when the underlying fraud system does not use a single numeric score.
The relevant question is always:
What does the next increment of restriction save, and what legitimate value does it destroy?
Segment economics before changing a global threshold
An average portfolio result can hide very different economics.
A threshold may behave differently for:
new vs returning customers;
high-margin vs low-margin products;
payment methods;
transaction-value bands;
markets;
customer tenure;
subscriptions vs one-time purchases;
PSPs;
authenticated vs non-authenticated flows.
The point is not to create a unique policy for every possible segment.
It is to identify segmentation when the economic or fraud difference is large enough to matter.
A global control can look efficient while:
under-protecting one cohort;
over-blocking another.
The deeper methodology for geographic variation belongs in Risk Thresholds by Market.
Market is context, not an economic shortcut
If one country's fraud rate is higher, that does not prove customers from that country deserve a stricter threshold.
Differences might instead come from:
payment-method mix;
product mix;
customer maturity;
acquiring setup;
authentication;
transaction size;
provider configuration;
data maturity.
Risk thresholds should respond to measured segment economics.
Not geographic stereotype.
PSP differences can distort the frontier
A multi-provider merchant may discover that:
PSP A approves more transactions
than:
PSP B.
That does not by itself prove PSP A creates more value.
The populations might differ.
Routing might differ.
Authentication might differ.
Fraud controls might differ.
Likewise, native provider risk scores should not be treated as interchangeable merely because both providers return numbers.
For an economic comparison, the merchant needs common downstream outcomes:
merchant action;
authorization;
successful completion;
matured fraud loss;
dispute cost;
relevant operating cost.
The detailed normalization problem belongs in Managing Fraud Across Multiple PSPs.
That is relevant to this article because risk controls influence which transactions proceed.
But a fraud product cannot define the merchant's economic optimum by itself.
The merchant still needs to supply the business context:
contribution economics;
fraud tolerance;
fulfillment cost;
customer value;
review cost;
dispute economics;
market strategy.
A useful payment-risk system should support better discrimination.
The merchant's leadership still owns the question:
How much additional risk is economically justified by the legitimate value it permits?
Test economic hypotheses, not just fraud hypotheses
Suppose the merchant proposes relaxing one risk threshold.
The incomplete hypothesis is:
“Fraud will remain acceptable.”
A stronger hypothesis is:
“Relaxing this threshold for this defined cohort will recover enough legitimate contribution to exceed the additional matured fraud, dispute, review, and operating cost while remaining inside our risk guardrails.”
Now the proposed change has:
population
economic outcome
risk outcome
guardrails
observation window
and:
a condition under which it should be reversed.
A backtest can reduce uncertainty.
It cannot prove future fraud behavior.
Controlled rollout, shadow evaluation, cohort analysis, or other validation may be appropriate depending on the control.
The detailed testing mechanics belong elsewhere in the cluster.
What result should cause the merchant to reverse course?
Every material threshold change should be falsifiable.
Suppose a stricter rule was expected to:
reduce matured fraud loss;
with only modest legitimate-customer impact.
After sufficient outcome maturity, the merchant observes:
a small fraud improvement;
substantially lower profitable legitimate acceptance;
higher support contacts;
no compensating reduction in other costs.
That evidence contradicts the original economic hypothesis.
The rule should not remain simply because:
“fraud went down.”
Likewise, a looser threshold should be reconsidered if incremental fraud and related costs exceed the incremental legitimate contribution it recovered.
This is how fraud controls become economic decisions rather than permanent beliefs.
Leadership should review the slope, not just the endpoints
A fraud leader says:
“Fraud is 0.4%.”
A payments leader says:
“Approval is 91%.”
Those numbers are useful.
But the more important leadership question is:
What happens economically if we move the decision boundary?
Specifically:
If we approve the next cohort currently being rejected, how much legitimate contribution do we gain per additional dollar of fraud and operating cost?
Or:
If we reject the next cohort currently being approved, how much fraud do we prevent per dollar of legitimate contribution sacrificed?
That is the slope of the Fraud–Approval Decision Frontier.
It is where the actual threshold decision lives.
A practical leadership decision table
Question
Why it matters
Which control or threshold is changing?
Avoid attributing portfolio movement to the wrong decision
Which population changes outcome?
Focus on marginal, not average, economics
How much legitimate contribution changes?
Quantifies growth impact
How much matured fraud loss changes?
Quantifies risk impact
How do disputes change?
Captures costs not contained in fraud alone
How does review volume/cost change?
Prevents approval gains from becoming queue cost
What customer/support impact appears?
Makes hidden false-positive costs more visible
Are merchant blocks separate from issuer declines?
Assigns the effect to the correct system
Has the fraud cohort matured enough?
Prevents premature conclusions
Is the result consistent across key segments?
Prevents blended averages from hiding damage
What result triggers rollback?
Keeps thresholds from becoming permanent by inertia
This is the level at which Fraud, Payments, Finance, and Product should converge.
The right question is not “How low can fraud go?”
Fraud controls consume economic opportunity.
Sometimes that cost is justified.
Sometimes it is not.
A tighter threshold can prevent expensive fraud.
It can also reject a profitable customer.
A looser threshold can recover legitimate sales.
It can also allow losses that exceed the contribution created.
Manual review can recover uncertain customers.
It can also create an expensive queue.
Authentication can improve confidence in some contexts.
It can also introduce customer friction.
The merchant therefore needs to stop treating fraud loss rate vs approval rate as a contest between two metrics.
The operating question is:
At the margin, does this control save more economic value from fraud than it removes from legitimate commerce?
That answer should be measured against mature outcomes, contribution rather than vanity volume, explicit operating costs, and a defined risk appetite.
The economically strongest fraud operation is not necessarily the one reporting the lowest fraud.
Nor is it the one approving the most transactions.
It is the one that can identify where the next unit of restriction stops paying for itself—and has the evidence and governance to stop tightening when it reaches that point.
🙂 🙂 🙂 🙂 🙂 🙂 🙂 🙂 🙂
EDITORIAL INTERPRETATION — NOT PART OF THE ARTICLE
This article owns the economic objective of fraud operations. It should not become another dashboard, rule-tuning, or market-threshold guide. The controlling brief requires a distinct article/SEO separation and specifically positions the piece around the proposition that merchants can reduce fraud simply by approving less business—which means fraud performance cannot be judged independently of legitimate acceptance.
Unique anchor: a stricter fraud policy can show a better fraud metric while destroying more legitimate contribution than it saves.
Editorial position: fraud and approval performance should be evaluated together using incremental economics, not treated as competing departmental KPIs.
Assumption challenged: “Lower fraud is always better.”
Primary merchant decision: Does the next increment of fraud restriction—or relaxation—create more economic value than it destroys?
Selected format: economic decision model built around the Fraud–Approval Decision Frontier and Net Revenue Risk Model.
Scope boundary: this article determines what economic outcome the merchant should optimize. Detailed dashboards belong to Fraud Performance Reporting; individual control changes belong to How to Tune Fraud Rules; country-specific calibration belongs to Risk Thresholds by Market. The brief explicitly assigns those boundaries.
SEO / EDITORIAL PACKAGE — NOT PART OF THE ARTICLE
Fraud Loss Rate vs Approval Rate: How to Optimize the Tradeoff
Alternative
Fraud Loss Rate vs Approval Rate: Finding the Economic Optimum
The alternative better reflects the article's advanced strategic intent; the first may be more immediately legible in search.
Recommended H1
Fraud Loss Rate vs Approval Rate: Finding the Point Where Risk Controls Create More Value Than They Destroy
Keep the brief's working title.
Suggested URL slug
/solve/fraud-blocking-growth/fraud-operations/fraud-loss-rate-vs-approval-rate/
Meta description
Fraud loss rate and approval rate should not be optimized separately. Learn how to compare legitimate contribution, fraud loss, disputes, review and operating costs to find economically sustainable risk thresholds.
Article excerpt
A lower fraud rate does not automatically mean a better fraud system. Merchants should evaluate risk controls by the incremental legitimate contribution they preserve or recover after fraud losses, disputes, review costs, support costs, and other material risk-control expenses.
fraud loss rate vs approval rate
Secondary keywords
fraud rate vs approval rate
fraud approval tradeoff
fraud false declines
fraud prevention approval rate
risk appetite payments
fraud loss optimization
payment fraud economics
Useful semantic extensions:
fraud approval optimization
fraud threshold economics
false decline cost
merchant fraud block rate
payment approval economics
ecommerce fraud losses
fraud risk appetite
marginal fraud loss
legitimate payment approval
net fraud loss
fraud contribution margin
fraud decision threshold
Relevant entities
Article architecture
Two policies, two apparently conflicting metrics
The easiest way to reduce fraud
Why fraud rate alone fails
Why approval rate alone fails
What “approval” actually means
Merchant blocks vs issuer declines
Fraud loss vs chargeback rate
Gross fraud vs net financial loss
Revenue vs contribution
Net Revenue Risk Model
Average vs marginal economics
Illustrative threshold comparison
Fraud–Approval Decision Frontier
Risk appetite as an operating boundary
False-positive economics
Manual-review cost
Threshold economics
Segment, market and PSP context
Testing/falsification
Leadership decision table
Economic conclusion
This keeps the article firmly in the territory the cluster ownership map assigns it: the economic tradeoff between protecting revenue and allowing legitimate revenue through.
Original framework 1: Fraud–Approval Decision Frontier
Visual design
Use a two-axis scatter/curve.
X-axis: Legitimate acceptance / legitimate contribution retained
Y-axis: Expected fraud exposure / fraud-related cost
Plot several policy configurations:
Very restrictive
Current policy
Candidate threshold A
Candidate threshold B
Very permissive
Connect efficient configurations into a frontier.
Then shade:
Economically acceptable operating territory
bounded by:
merchant risk appetite;
dispute constraints;
operational capacity;
other relevant guardrails.
Show one dominated policy inside the curve:
Same fraud exposure, lower legitimate acceptance.
Core visual annotation
Tightening a threshold usually moves down-left.
Loosening usually moves up-right.
Better decisioning moves the frontier outward.
That last sentence is especially important.
A better fraud system should not merely choose between fraud and approval more aggressively; it should improve the merchant's ability to discriminate between legitimate and fraudulent activity.
Recommended alt text
Fraud–Approval Decision Frontier plotting legitimate customer acceptance against expected fraud exposure, showing restrictive and permissive fraud policies, dominated policies, merchant risk constraints, and the economically acceptable operating region.
Original framework 2: Net Revenue Risk Model
Recommended visual equation:
Incremental legitimate contribution
− Incremental fraud loss
− Incremental dispute cost
− Incremental manual-review cost
− Incremental fraud/control cost
− Incremental support and operating cost
= Incremental net economic value
Underneath:
Change the control when the marginal economics justify it—not because one headline KPI moved.
Use contribution, not GMV, as the primary recommended economic input where the merchant can calculate it.
Metric definitions for publication
Fraud loss rate
Possible merchant definition:
Confirmed fraud loss
÷ defined transaction-value base
× 100
Publication must specify whether the denominator is:
approved;
captured;
settled;
completed-order;
another defined value population.
Also define outcome maturity.
Merchant fraud-block rate
Transactions rejected by merchant-controlled fraud decisions
÷ eligible transactions evaluated
× 100
Do not mix with issuer declines.
Merchant risk acceptance rate
Useful internal complement:
Transactions permitted through merchant fraud controls
÷ eligible transactions evaluated
This is a merchant-management metric, not a universal card-network definition.
Issuer authorization rate
Use the merchant/acquirer/provider's explicitly documented methodology.
At minimum, do not place transactions rejected before authorization into an issuer-decline denominator.
Legitimate volume recovered
For a policy relaxation:
additional transactions ultimately classified as legitimate and accepted under the new policy
Measure by count, value, and preferably economic contribution.
State how legitimacy is established.
Incremental fraud loss
matured fraud loss under candidate policy
− matured fraud loss under baseline
Use comparable populations and observation windows.
Net expected value
Recommended management formula:
Incremental legitimate contribution
− incremental fraud loss
− incremental dispute cost
− incremental review cost
− incremental fraud/control cost
− incremental support/operating cost
State excluded costs.
Measurement plan
For every meaningful policy/threshold comparison, record:
Dimension
Requirement
Baseline policy
Explicit
Candidate policy
Explicit
Changed population
Defined
Observation period
Defined
Outcome maturity
Sufficiently comparable
Merchant blocks
Separate
Issuer declines
Separate
Legitimate approvals
Measured/estimated transparently
Confirmed fraud
Defined
Disputes
Separate
Contribution economics
Defined
Review cost
Included where changed
Support/ops cost
Included if material
Primary segments
Compared where economically relevant
Rollback condition
Defined before change
The full reporting architecture still belongs in Fraud Performance Reporting, not here.
Internal-link recommendations
Fraud Performance Reporting
Anchor:
measure fraud and approval outcomes on comparable populations
Use after metric definitions.
How to Tune Fraud Rules
Anchor:
change thresholds after production economics show the control should move
Use after the marginal-threshold section.
Risk Thresholds by Market
Anchor:
determine when market-specific fraud economics justify different thresholds
Use in the segmentation section.
Fraud Rule Testing
Optional:
test the economic hypothesis before relying broadly on a proposed fraud-control change
Manual Review Best Practices
Optional:
evaluate whether manual review creates enough decision value to justify its cost
Do not reproduce either article's detailed methodology.
Authoritative external-source recommendations
Visa Acceptance Solutions — payment fraud management
Useful current first-party support for the principle that merchants should manage fraud while improving acceptance of good transactions rather than treating fraud reduction as the sole objective.
Visa — enterprise fraud management
Useful for the explicit connection among fraud reduction, false declines, acceptance rates, and operational manual-review costs.
Visa Protect / authorization-risk material
Useful for explaining why fraud detection and legitimate authorization outcomes need to be considered together.
Mastercard — transaction optimization
Current first-party material explicitly presents reducing false declines, fraud prevention, and higher approval rates as connected transaction-decisioning objectives.
Visa — Chargebacks
Useful for verifying that chargeback/dispute measures should not be treated as equivalent to fraud measures because dispute reasons include more than unauthorized-payment fraud.
Current first-party support for custom risk rules, alert/decline actions, and human review positioned around reducing fraud while allowing legitimate transactions through.
No competitor source is necessary.
Proposed author and reviewers
Primary author:
Head of Fraud / Fraud Operations leader with direct experience setting or governing merchant risk thresholds.
Finance reviewer:
Finance or FP&A leader capable of validating contribution-margin and net-loss treatment.
Payments reviewer:
Payments leader capable of separating merchant risk acceptance from issuer authorization and end-to-end payment performance.
Optional data reviewer:
Fraud analytics/data-science specialist for cohort maturity and marginal-increment methodology.
The article brief specifically calls for a Fraud leader plus Finance/Payments contribution.
The ideal genuine first-party contribution is not a generic quote about “balancing fraud and conversion.” It should explain one case where the fraud-optimal and economically optimal thresholds differed, without disclosing exploitable control details.
Fraud fact-check list
Before publication verify:
Definition of confirmed fraud.
Which fraud types enter the financial model.
Treatment of account takeover and non-card fraud where applicable.
Fraud outcome maturation period.
Treatment of fraud recovered later.
Whether disputes are included or separate.
Treatment of refunds.
How merchant blocks are identified.
How false-positive indicators are validated.
Whether manual approvals are treated as false-positive evidence or only a proxy.
How unresolved cases are handled.
Whether the proposed threshold framework maps to actual fraud decision architecture.
Whether risk appetite has explicit financial or operational guardrails.
Metric-definition fact-check list
Confirm:
Fraud loss rate
numerator;
denominator;
outcome window;
gross vs net loss.
Merchant fraud-block rate
eligible population;
whether provider blocks count as merchant controlled;
treatment of challenges/reviews.
Authorization rate
exact denominator;
retries;
technical failures;
authentication failures;
whether merchant-blocked attempts are excluded.
Legitimate volume recovered
legitimacy definition;
count/value/contribution basis;
maturation.
Contribution
gross margin vs contribution margin;
variable fulfillment cost;
payment fees;
business-specific cost treatment.
Dispute cost
principal vs fees vs operating cost.
Review cost
labor;
tooling;
queue overhead if material.
Net expected value
included and excluded cost categories.
PSP/provider fact-check list
Where provider data enters the model, verify:
which system made the fraud decision;
provider vs merchant block;
authentication stage;
authorization stage;
risk-score semantics;
rule or model version;
provider action;
whether native risk information is retained;
routing population;
whether providers receive comparable traffic.
Do not compare PSP approval/fraud economics without accounting for routed population differences.
Market fact-check list
If the article later receives market-specific examples, verify:
payment method;
product;
merchant entity;
customer population;
transaction value;
acquiring configuration;
PSP;
authentication environment;
fraud-outcome maturity;
margin differences.
Do not assume a market's higher average fraud rate means a tighter threshold creates more economic value.
Security / privacy / compliance review
No detailed legal guidance is required for the article.
However, internal review should confirm that any suggested segmentation or data use:
has an appropriate operational basis;
complies with applicable privacy/security requirements;
does not use protected characteristics as fraud shortcuts;
does not expose exploitable fraud-rule details.
The article's economics framework should never be interpreted as permission to ignore legal, contractual, security, card-network, or other applicable constraints simply because a transaction appears profitable.
Before publication, verify internally:
current PASSPORT.RISK product name;
custom-rule functionality;
available risk actions;
human-review workflow;
which transactions the risk team reviews;
how merchant-side blocks are exposed;
issuer-decline reporting;
false-positive measurement capability;
approval reporting;
fraud-loss reporting;
chargeback/dispute cost visibility;
market segmentation;
PSP/acquirer segmentation;
whether control versions and threshold changes are retained;
reporting/API fields available for economic analysis.
population;
denominator;
observation period;
payment methods;
geography;
merchant cohort;
methodology
are approved and relevant to this article.
Amber-flag claims
“Lower fraud means better fraud performance.”
Reject as incomplete.
“Higher approval means better payment performance.”
Reject as incomplete.
“Every blocked transaction is a false decline.”
Incorrect.
“Every issuer decline is caused by fraud.”
Incorrect.
“Approval rate equals conversion.”
Incorrect.
“Chargeback rate equals fraud rate.”
Incorrect.
“A recovered $100 order creates $100 of value.”
Only if that is genuinely the relevant economic measure; normally contribution needs consideration.
“False-positive cost equals order value.”
Too simplistic.
“One global threshold maximizes revenue.”
Unsupported.
“A tighter fraud threshold always lowers total economic loss.”
Unsupported.
“A lower average fraud rate proves the threshold at the margin is correct.”
Incorrect inference.
“Backtesting proves which threshold will perform best.”
Too strong.
Do not publish without approved first-party methodology.
Cluster-wide sameness audit
Field
Fraud Loss Rate vs Approval Rate
Opening device
Comparison between two economically ambiguous policies
Primary reader
Fraud, Payments, Finance, Risk and ecommerce leadership
Primary failure
Optimizing one departmental KPI
Format
Economic decision model
Unique anchor
Merchant can reduce fraud by rejecting more legitimate business
Editorial position
Optimize incremental net economic value
Assumption challenged
“Lower fraud is always better”
Merchant decision
Should the risk boundary tighten or loosen?
Primary framework
Fraud–Approval Decision Frontier
Secondary framework
Net Revenue Risk Model
Primary metric concept
Marginal net economic value
Central distinction
Average vs marginal performance
Payment distinction
Merchant block vs issuer decline
Risk distinction
Fraud rate vs chargeback rate
Finance distinction
Revenue/GMV vs contribution
Customer dimension
Economic cost of false positives
Operational dimension
Review/support/control cost
Market treatment
Segment economics, not geographic stereotype
Closing question
Does the next unit of restriction save more than it destroys?
CTA/internal journey
Reporting → rule tuning → market thresholds
This keeps the article separate from the rest of the Fraud Operations cluster. Its unique contribution is not another exhortation to “balance fraud and conversion.” It gives leadership the specific economic test the brief asks for: compare the marginal legitimate contribution created or destroyed by a risk decision with the marginal fraud and operating costs that same decision produces.
Need help narrowing down payment options?
Answer a few questions about your business and compare the capabilities that matter to you.
Explore matching options ↗