What it means
Every business that handles money loses some of it to fraud, whether that is a customer using a stolen card, a supplier submitting an inflated invoice, or an employee filing expenses for trips they never took. The detection rate answers a narrow but important question: of the fraud that got attempted, what proportion did we spot?
The rest, sometimes called leakage, either shows up much later or never surfaces at all. The metric matters because anti-fraud work is expensive and easy to over-fund or under-fund without evidence.
A finance director who can show that the detection rate rose from 72% to 89% after a new screening tool went live has a concrete argument for keeping the budget. A flat detection rate alongside rising fraud losses is an equally clear signal that the current controls have been outgrown.
Calculating it is straightforward arithmetic, but sourcing the denominator is the hard part, because undetected fraud is by definition unknown. Practical teams estimate the true total by adding confirmed detections to fraud that was found afterwards through chargebacks, customer complaints, audits or whistleblower reports.
The rate is then recalculated for prior periods once that late evidence arrives, so recent figures are always provisional. There are two common versions and they can differ sharply.
Count-based detection measures the proportion of fraudulent cases caught, while value-based detection measures the proportion of fraudulent dollars stopped. A team that catches many small card tests but misses one large wire diversion can report a high count-based rate and a poor value-based one, which is why serious reporting shows both.
The nuance most people miss is that a detection rate on its own can be gamed by tightening thresholds until almost everything is flagged. That catches more fraud but also blocks genuine customers, so the number should always be read next to the false positive rate and the revenue lost to declined legitimate orders.
Good anti-fraud teams optimise the pair, not the single figure.
In practice
Real-world examples.
Example
A subscription streaming service tracks fraudulent sign-ups made with stolen cards. Of 4,000 fraudulent accounts created over six months, automated screening caught 3,400 at registration and the remaining 600 surfaced through chargebacks, giving a detection rate of 85%. The product team uses the gap to justify adding device fingerprinting at checkout.
Example
A regional insurer measures how many exaggerated claims its assessors identify before payout. Internal audit sampling suggests the detection rate sits near 60%, meaning roughly two in five inflated claims are paid in full. The claims director uses that figure to build a business case for a specialist investigations unit.
Example
A logistics firm applies the metric internally to expense fraud rather than customer fraud. After introducing automated receipt matching, the proportion of falsified expense claims caught before reimbursement rose from about half to more than four fifths, and the finance team reports the change monthly to the audit committee.
Think of it
“Fraud detection shows how often you catch fraud-your fraud-catching success rate.
Formula
Calculation
Fraud Detection Rate = (Fraud cases detected by controls / Total fraud cases that occurred) x 100
An online electronics retailer reviews the quarter just ended. Its screening system flagged and confirmed 1,020 fraudulent orders. A further 180 fraudulent orders were only identified afterwards through card issuer chargebacks and customer disputes, giving a total of 1,200 fraudulent orders that actually occurred.
Fraud Detection Rate = (1,020 / 1,200) x 100 = 85%
The team then repeats the calculation by value. The 1,200 fraudulent orders were worth $600,000 in total, of which $480,000 was blocked before shipping.
Value-based Fraud Detection Rate = ($480,000 / $600,000) x 100 = 80%
So the retailer caught 85% of fraud attempts but only 80% of the fraudulent value, telling the team that the cases slipping through are larger than average and deserve a targeted rule rather than a blanket tightening.Case study
Seen in the real world.
Northvale Marketplace is a fictional online homeware platform used here purely as an illustrative example. In its first two years the company treated fraud as a cost of doing business, writing off chargebacks without measuring them, and the finance team could not say whether losses were growing because fraud was rising or because controls were weakening.
The new financial controller began reconstructing a true fraud population each quarter by combining orders blocked at checkout with chargebacks received later. The first calculation was uncomfortable: of 900 fraudulent orders in the quarter, only 495 had been caught, a detection rate of 55%. Worse, the value-based rate was lower still, because the largest fraudulent baskets sailed through untouched.
Twelve months and one new screening rule set later, the illustrative detection rate had reached 88% by count and 91% by value, while declines of genuine orders had risen only slightly. The controller now reports both numbers side by side, and the board treats the pair as a single measure of how well the platform is protecting itself without turning away real buyers.
Watch out
Common mistakes.
- Treating the detection rate as a measure of how much fraud is happening. It measures the effectiveness of your controls, and total fraud losses are a separate figure that must be tracked alongside it.
- Using only the fraud you caught as the denominator, which mathematically forces the answer to 100%. The denominator has to include fraud discovered later through chargebacks, audits and complaints.
- Chasing a higher rate without watching false positives, so genuine customers are blocked and the revenue lost quietly exceeds the fraud prevented.
Questions
People also ask.
How often should the fraud detection rate be recalculated?
Monthly for operational review, but each period should be restated for a few months afterwards as late chargebacks and audit findings reveal fraud that was originally missed.
Is a detection rate of 100% realistic?
Almost never, because some fraud is never identified at all; consistently reported perfect scores usually mean the denominator is being measured incorrectly.
Does this metric apply to internal fraud as well as customer fraud?
Yes, and many organisations run separate rates for payment fraud, expense fraud and procurement fraud because the controls and the realistic benchmarks differ for each.
From the founder's library
Take it further with the book.
Build your financial confidence beyond this definition. Shihan's full-length guide, Accounting Fundamentals, takes the same plain-English approach and turns it into a complete, practical playbook for non-finance managers, business owners and students - with chapter-end quiz answers and presentation slides included.
25% off with code MMHQ25, applied at checkout. Priced in USD - checkout may show the equivalent in your local currency.
View the book and save 25%