Almost every fraud statistic you have read shares a hidden assumption: that the frauds which get reported look like the frauds that happen.
They do not, and the gap is structural rather than accidental.
How the numbers are built
National fraud statistics are compiled from reports — to banks, to police, to consumer agencies. That makes them rigorous about what they capture and silent about what they do not. A report happens when someone decides the loss is worth the paperwork.
That decision has a threshold. Below it, people absorb the loss and move on. Which means the dataset is not a sample of fraud — it is a sample of fraud that cleared someone’s reporting bar.
Measuring from the other end
We tried approaching it from the opposite direction. Instead of asking who reported a loss, we asked what people search for. Search happens before the reporting decision, and often instead of it.
Across our own Search Console corpus we isolated every query naming a specific money amount alongside payment or screenshot terms. 60.3% sought sums under ₹1,000, with ₹500 the most common single figure. The full breakdown is here.
If demand concentrates below ₹1,000 and reporting concentrates above it, then the two datasets are describing different populations while appearing to describe the same crime.
What this does and does not prove
It does not prove that most fraud is small. Our corpus measures search demand reaching a single publisher, and we rank for verification-type queries, so our audience is skewed by construction. It also cannot distinguish a worried seller from someone shopping for a forgery tool.
What it does show is a divergence: the amounts people search about are not the amounts that dominate loss-based reporting. At minimum, that should make anyone cautious about treating reported-loss averages as a description of typical risk.
Why this matters practically
If you calibrate your caution to published averages, you will prepare for the fraud that gets reported and stay exposed to the fraud that does not. Reimbursement schemes inherit the same blind spot: a rule built on reported cases protects the visible tail.
The practical response is unglamorous. Verify small payments with the same reflex you apply to large ones, and know whether your payment rail can be reversed at all before you need the answer.
And if you are defrauded of a small amount, report it anyway. The under-counting described here is not a flaw in anyone’s methodology — it is the aggregate of thousands of individually reasonable decisions not to bother.
Check a payment screenshot free →
Methodology and limitations
Dataset: Google Search Console, asquaresolution.com. Period: 28 June – 25 July 2026. Sample: 60 amount-bearing queries, 604 impressions, drawn from a corpus of roughly 103,000 impressions. Confidence: medium — internally consistent and reproducible from the published CSV, but single-publisher and demand-side only.
Limitations: this measures search demand reaching one publisher, not victim losses. It cannot separate people verifying a payment from people seeking to produce forgeries. n=604 is small; treat the figures as directional. The full dataset is free to download and re-analyse with attribution.
Every study we publish, with its full dataset, is free and ungated at asquaresolution.com/research.