Every KYC vendor publishes a verification pass rate. The number appears prominently in sales materials, usually framed as something like "98.4% document verification accuracy" or "industry-leading 97.2% auto-approve rate." These numbers are real. They are also carefully defined in ways that make them much less informative than they appear.
The denominator in a pass rate calculation is typically the set of sessions where a user submitted a document for verification. That means the pass rate only measures users who made it to the document submission step. It says nothing about the users who started the onboarding flow, hit the document upload step, and left. It says nothing about users who saw the camera permission request and declined. It says nothing about users who uploaded a document, got an error, and abandoned rather than retrying.
That missing population is where the actual conversion cost of KYC friction lives. And the vendors that most benefit from pass rate framing have no structural incentive to publish data on it.
Why Abandonment Data Stays Private
KYC vendors typically instrument their SDK or hosted flow to measure completion rates within their system. They know what percentage of sessions that start their verification step complete it. Some vendors share this data; most don't publish it as a benchmark. The reason is structural: abandonment rate is primarily a function of user experience design, friction, and platform context, not verification model quality. A vendor whose SDK has a confusing camera interface will show higher abandonment than a vendor whose interface is smooth, even if the underlying verification accuracy is identical.
This means abandonment is a product design metric that reflects on the vendor's UX choices, and it varies substantially by client industry, user demographic, and device type. Publishing an aggregate abandonment number would be misleading in one direction (low abandonment clients would inflate the number) and potentially damaging for vendors whose particular SDK design choices create friction. So it doesn't get published.
The result is that most compliance and product teams making KYC vendor decisions are working with incomplete data. They can compare pass rates and fraud catch rates. They cannot compare abandonment rates, retry rates, or the specific step in the flow where abandonment concentrates. Those metrics require instrumenting the vendor's flow from your own analytics layer, which most teams do not do rigorously before selecting a vendor.
What Step-Level Drop-Off Actually Looks Like
When we instrument a client's existing KYC flow at the step level, the abandonment pattern that emerges is usually not a flat attrition across the flow. It concentrates at specific friction points, and the friction points differ by user segment and device type.
On mobile, the highest abandonment step is almost always the camera permission request or the first document capture attempt. Users on iOS and Android devices who encounter the camera permission prompt for the first time within a web browser flow (not a native app) have a meaningful denial rate. Users who grant permission but then get an error on the first capture attempt (poor lighting, angle error, blurriness) abandon at significantly higher rates than users whose first attempt succeeds. The first attempt success rate on document capture is directly tied to the quality of the capture guidance in the UI, which is entirely a product decision by the vendor, not a reflection of the underlying verification technology.
On desktop, drop-off concentrates differently. Users who are asked to complete a verification step that requires a mobile device (for document scanning or liveness check) and are directed to a QR code handoff to mobile show substantial abandonment at that handoff. The QR code flow introduces a context switch that a meaningful fraction of users either do not understand or do not want to complete.
Retry Abandonment
A particularly invisible drop-off source is retry abandonment: users who complete a first attempt that fails (document not recognized, liveness check failed, data mismatch) and then abandon rather than retrying. This population is almost never counted in vendor-published metrics because they did "attempt" the verification. Whether they successfully navigated a retry path is a separate question.
Retry abandonment rates are highly sensitive to the error message quality. An error message that tells the user what specifically went wrong and how to fix it (wrong document type, wrong side of document, lighting issue) produces meaningfully lower retry abandonment than a generic "we couldn't verify your identity, please try again" message. This is another product design decision that sits with the vendor, and it varies substantially between vendors without appearing in any published benchmark.
The Demographic Distribution Problem
Abandonment rates are not uniform across user demographics, and understanding which segments show elevated abandonment can surface both product problems and potential adverse impact concerns.
Older users consistently show higher abandonment at liveness check steps. Active liveness checks that require a specific gesture (turning head, blinking, speaking a phrase) show higher abandonment among users unfamiliar with the format. Passive liveness checks, which require only a still selfie, perform better on this dimension but have different fraud resistance properties.
Users with non-standard documents (non-US passports, state IDs with older formats, documents from countries with less common document types in the training data) show higher abandonment at the document classification step because error rates are higher for those document types and the retry burden falls on the user. Document coverage breadth is a real vendor differentiator that usually appears in sales conversations but rarely appears as a structured metric in vendor comparison frameworks.
We're not saying that abandonment rates should override fraud prevention requirements. A KYC flow that passes everyone with no friction is not a KYC flow. But the demographic distribution of abandonment has compliance dimensions beyond just conversion: if certain demographic groups abandon at systematically higher rates due to document coverage gaps or liveness bias, that distribution is worth understanding and documenting as part of your model risk management process, particularly for regulated financial services products subject to fair lending or non-discrimination requirements.
How to Measure What Vendors Don't Publish
The practical solution is to instrument the KYC flow from your own analytics layer rather than relying on vendor-provided reporting. This requires event tracking at each step of the flow with user-level session identifiers that allow you to reconstruct the funnel.
The minimum events to track are: flow started, camera/document permission granted or denied, first document submission attempted, first document submission result, retry attempted (if applicable), liveness check started, liveness check completed, overall session decision received, and flow exit without completion. With those events, you can construct step-level conversion rates and identify exactly where your flow loses users.
For vendors who use a hosted flow (user redirects to vendor's URL), step-level instrumentation requires either the vendor's analytics API or tracking redirect parameters back to your analytics. Some vendors support this well; others do not. The degree to which a vendor enables measurement of their own flow is itself a signal about how confident they are in their UX quality.
Abandonment vs. Fraud Catch Rate: The Real Tradeoff
The reason abandonment data matters alongside pass rate and fraud catch rate is that the three metrics form a tradeoff surface. You can reduce abandonment by lowering friction, which typically means either removing verification steps or making them less stringent. That improves conversion and reduces the drop-off problem, but may reduce fraud catch rates. You can increase fraud catch rates by adding checks or raising thresholds, which may increase abandonment for legitimate users who find the additional friction unacceptable.
Making informed decisions about where to sit on that tradeoff surface requires knowing all three numbers. Vendors who publish only pass rate and fraud catch rate, without abandonment data, are implicitly asking their customers to optimize for metrics they can see while accepting an unknown cost on the metric they cannot. That is a poor design for the client's onboarding economics and a convenient structure for vendors whose abandonment numbers would compare unfavorably.
Building visibility into the full tradeoff surface is a primary design goal in how we structure KYC flows at IDPylon. The routing logic adapts to where each step sits on that tradeoff in the context of that session's risk profile, rather than applying a single static configuration that implicitly accepts whatever abandonment cost the vendor's default design produces.