HomePlatformsDocumentsGlossaryBlogContactAbout
Reference dataset

The KYC Rejection Index

A structured classification of every documented reason identity verification fails, showing which stage of the pipeline produces each one, whether retaking the photograph can possibly help, and where the published evidence actually sits.

Version 1.0 · published 30 July 2026 · maintained by KYC Rejected

Key facts

  • Sixteen documented rejection causes, each mapped to one of five pipeline stages.
  • Only nine of the sixteen can be resolved by retaking the photograph. The other seven cannot, which is where most wasted attempts go.
  • Published research puts identity verification abandonment between roughly 25 percent at established banks and over 60 percent at unfamiliar fintech products.
  • Signicat has reported that 63 percent of European consumers abandoned a financial services signup because verification was too cumbersome.
  • This version contains no first party measurements. The methodology for those is published below and the figures will appear here when the measurement pipeline is live.

What this index is, and what it is not

The classification below is our own analytical work, built from the sixteen rejection causes we document in detail. It is a structured taxonomy, not a survey, and every entry is a statement about how verification pipelines behave rather than a measurement of how often something happens.

Where numbers appear, they come from named third parties and are attributed in place. We are not publishing first party frequency data in this version because we do not yet have it, and inventing it would make this document worse than useless. The section on methodology explains exactly what will be measured and how.

Part one: the classification

Verification runs as a pipeline. Each stage tests something different, a failure at any stage ends the run, and the message returned to the applicant rarely identifies which stage produced it. That opacity is the single biggest reason people waste their limited attempts, because the natural response to any rejection is to retake the photograph, and for seven of the sixteen causes below that is guaranteed to fail again.

The column that matters most is the last one.

Rejection causePipeline stageWhat is actually measured or comparedFixable by recapture
Document blurryImage qualityEdge sharpness across the document surfaceYes
Glare detectedImage qualitySpecular highlight coverage over printed fieldsYes
Photo too darkImage qualityExposure histogram and resulting sensor noiseYes
Low resolutionImage qualityEffective pixel density across the documentYes
Hologram reflectionImage qualityWhether a security feature obscures a printed fieldYes
Document croppedGeometryPresence of all four corners for perspective correctionYes
Unable to read IDExtractionOptical character recognition confidence, machine readable zone parseYes
Selfie rejectedBiometricFace image quality against capture constraintsYes
Liveness check failedBiometricEvidence of a physically present, moving personYes
Face not matchingBiometricGeometric comparison against the document portraitPartly
Face verification failedBiometricCombined match score across capture and comparisonPartly
Name mismatchData comparisonProfile name against extracted document fieldsNo
Address mismatchData comparisonProfile address against a proof of address documentNo
Document expiredDocument validityExpiry date against the current date and any bufferNo
Duplicate accountScreening and riskIdentity matched against existing account recordsNo
Sanctions or PEP false matchScreening and riskName against sanctions, PEP and adverse media dataNo

Read down that final column and the practical lesson is immediate. Nine causes respond to a better photograph. Two respond partially, in that better capture technique helps but a genuinely different appearance from the document portrait will not be solved by lighting. Five cannot be touched by the camera at all, and four of those five are not even about the document.

This is the whole argument for diagnosing before resubmitting. Attempts are commonly capped between three and five. Spending one on a recapture when the failure came from a name field or a watchlist is not a small inefficiency, it is a third of your budget gone with the underlying cause untouched.

Stop guessing what went wrong A human specialist corrects the document for you and sends it back ready to submit, usually in under 10 minutes. $1.99, money back if it is not approved. Get my fix

Part two: what the published evidence says

These figures are not ours. Each is attributed to the organisation that published it, and each is included because it speaks to something the taxonomy above cannot: how often people give up, and what the fraud pressure looks like from the other side of the counter.

FindingReported byWhat it tells you
Verification abandonment ranges from roughly 25 percent at established banks to over 60 percent at unfamiliar fintech productsOnboarding research summarised across industry sourcesTrust in the brand changes tolerance for friction more than the checks themselves do
63 percent of European consumers have abandoned a financial services signup because identity verification was too cumbersomeSignicatAbandonment is the majority outcome in some markets, not an edge case
38 percent of people who abandoned did so because they did not have the required documents to hand, and 21 percent because it took too longOnboarding drop off researchA large share of failure is preparation, not capability. Gathering documents first materially changes the outcome
Most users abandon entirely after two or three failed attemptsOnboarding drop off researchThe attempt budget is behavioural as well as technical. People quit before the platform locks them out
Deepfakes account for around 11 percent of first party fraud globally, and 41 percent of attempts in EuropeIndustry fraud reporting, 2026Explains why liveness checks became strict, and why legitimate users now fail them more often
A 700 percent year over year increase in injection attacksJumio, 2026 Online Identity StudyVerification is tightening in response to attack volume, which raises the false rejection rate for ordinary applicants
An identity fraud rate of 1.53 percent in iGaming in Q1 2026, up 18 percent year on year, measured across more than three million verification attemptsSumsub, Identity Fraud Report 2025 to 2026Gives a sense of the true positive base rate the screens are tuned against
74 percent of crypto providers now prioritise verification accuracy over onboarding speed, against 39 percent prioritising speedIndustry survey reporting, 2026The industry has explicitly chosen stricter checks over smoother signup, so expect more friction rather than less

Two things follow from reading those together. First, the dominant failure mode in the industry is not rejection, it is abandonment, and a meaningful slice of that is people who simply were not prepared when they started. Second, the checks are getting stricter for reasons that have nothing to do with you, because attack volume is rising fast and the systems are being tuned to catch it. Legitimate users absorb that tightening as a higher chance of being wrongly refused.

Part three: methodology for first party data

We are publishing the method before the numbers, deliberately, so that anyone citing this later can judge how the figures were produced rather than taking them on trust.

What will be measured. For each document assessed by our checker, five properties computed in the browser: edge sharpness as a proxy for focus, specular highlight coverage as a proxy for glare, exposure distribution as a proxy for lighting, corner detectability as a proxy for framing, and effective resolution across the document. For selfies, face presence and inter pupil distance in pixels.

What will be recorded. Only which thresholds were crossed, as counters. No image, no filename, no document number, no name, no identifier of any kind. The document itself never leaves the device, which is what makes the counters safe to publish.

What will be published. The share of assessed documents crossing each failure threshold, a total sample size, and the collection window. Nothing will be published below a sample size where a single submission could move a figure meaningfully.

What will not be claimed. These counters will describe documents people chose to check with us before submitting them somewhere. That is a self selected group, most likely skewed toward people who already suspect a problem, and it will not be representative of all verification attempts anywhere. We will say so on the figures rather than in a footnote.

Current first party sample size: zero. The measurement pipeline is not live. When it is, this section becomes a table and this sentence disappears.

How to cite this

The classification in part one is original work and may be quoted with attribution to KYC Rejected, citing this page and the version number at the top. If you are quoting a figure from part two, cite the organisation named in the middle column rather than us, because the finding is theirs.

If you find an error in the classification, or you have a documented rejection cause we have not covered, the contact page reaches us and corrections get made in a numbered revision rather than silently.

Frequently asked questions

What is the most common reason KYC verification fails?

Across the sixteen documented causes, image quality problems make up the largest group, seven of them, covering blur, glare, darkness, low resolution, hologram interference, cropped corners and unreadable text. They are also the group that responds to retaking the photograph. The second largest group is data comparison, where a name or address on the profile disagrees with the document, and no photograph affects those at all.

How many KYC rejection causes can be fixed by taking a better photo?

Nine of sixteen respond directly to recapture, two respond partially, and five cannot be affected by the camera at all. The five are name mismatch, address mismatch, expired document, duplicate account, and a sanctions or politically exposed persons false match. Recognising which group you are in before resubmitting is the single highest value thing you can do, because attempts are usually capped between three and five.

How many people abandon identity verification?

Published research puts abandonment between roughly 25 percent at established banks and over 60 percent at unfamiliar fintech products. Signicat has reported that 63 percent of European consumers abandoned a financial services signup because verification was too cumbersome, and drop off research attributes 38 percent of abandonment to not having the required documents to hand and 21 percent to the process taking too long.

Why are liveness checks getting harder to pass?

Because attack volume rose sharply. Industry fraud reporting for 2026 puts deepfakes at around 11 percent of first party fraud globally and 41 percent of attempts in Europe, and Jumio's 2026 Online Identity Study reported a 700 percent year over year increase in injection attacks. Systems tuned to catch that will refuse more legitimate people as a side effect, which is why a genuine user can fail a liveness check several times without doing anything wrong.

Does this index contain your own measurements?

Not in version 1.0. The classification in part one is our own analytical work, and the figures in part two are attributed to the third parties who published them. First party frequency data has a published methodology on this page and a current sample size of zero, because the measurement pipeline is not live yet. We would rather show an empty table than a fabricated one.

Can I reuse this classification?

Yes, with attribution to KYC Rejected and a link to this page including the version number. If you are quoting one of the third party statistics, attribute it to the organisation named beside it rather than to us. Corrections are welcome through the contact page and are issued as numbered revisions.

Out of attempts, or out of patience? Skip the trial and error. A real specialist rebalances the lighting, lifts the glare and squares the edges, then emails it back. Human specialist, under 10 min, $1.99. Get my fix