Why Video KYC Sessions Fail: The Most Common Causes
Video KYC has transformed how Indian financial institutions verify customer identities, but session failure rates remain a persistent challenge across the industry. Data from multiple V-CIP deployments across banks, NBFCs, and insurance companies indicates that the average first-attempt failure rate for Video KYC sessions in India ranges from 15 to 35 percent, depending on the platform, customer segment, and implementation quality. Each failed session represents a direct cost (wasted agent time, platform fees for incomplete sessions) and an indirect cost (customer frustration, potential drop-off, delayed onboarding revenue).
The RBI's V-CIP guidelines, as outlined in the Master Direction on KYC (updated January 2023) and subsequent circulars, set specific requirements for what constitutes a valid Video KYC session: live video interaction with a trained official, capture of the customer's live photograph, verification of original OVDs (Officially Valid Documents), real-time face matching, geo-tagging, and a complete recorded audit trail. Failure to meet any of these requirements means the session must be repeated, adding cost and friction to the onboarding process.
Understanding why sessions fail is the first step toward systematic improvement. Through analysis of hundreds of thousands of Video KYC sessions across multiple institutions, we have identified the ten most common failure causes, grouped into four categories: network and device issues, document and OCR problems, liveness and biometric failures, and human factors (both agent-side and customer-side). Each category requires different technical and operational interventions.
Network and Device Compatibility Issues
Network instability is the single largest cause of Video KYC failures in India, accounting for 25 to 35 percent of all failed sessions. India's mobile internet infrastructure, while rapidly improving, still has significant coverage gaps and quality inconsistencies. A Video KYC session requires a stable connection of at least 1 Mbps upload and 1.5 Mbps download speed for acceptable video quality. However, TRAI data shows that average 4G speeds in many tier-2 and tier-3 cities drop below 5 Mbps during peak hours, with frequent micro-outages lasting 2 to 10 seconds. These interruptions cause video freezes, audio drops, and session disconnections that force restarts or rescheduling.
The technical root cause often lies in WebRTC connection handling. Most Video KYC platforms use WebRTC for real-time video communication, which relies on ICE (Interactive Connectivity Establishment) to negotiate the optimal connection path between the customer's device and the media server. On congested mobile networks, ICE negotiation can fail, TURN server relay becomes necessary (adding latency), and packet loss exceeds the threshold where video codecs can compensate. Platforms that do not implement adaptive bitrate streaming compound this problem by maintaining high-resolution video even when bandwidth cannot support it, leading to buffer overflow and connection drops.
Device compatibility is the second major technical failure category, responsible for 10 to 15 percent of session failures. India's smartphone market is enormously fragmented: customers use devices ranging from entry-level smartphones with 1 to 2 GB RAM running Android 9 to the latest flagships. Older devices struggle with the simultaneous demands of video encoding, OCR processing, liveness detection, and UI rendering that modern Video KYC platforms require. Browser compatibility adds another layer of complexity -- not all browsers on all Android versions support the WebRTC APIs and camera permissions required for Video KYC.
The fix for network issues involves three strategies. First, implement adaptive bitrate streaming that automatically reduces video resolution (from 720p to 480p to 360p) as bandwidth decreases, maintaining session continuity at the expense of video quality. Second, build robust reconnection logic that allows sessions to resume within 30 to 60 seconds of a disconnection without requiring the customer to restart. Third, provide a network quality pre-check before the session begins, warning customers with insufficient bandwidth and suggesting they switch to Wi-Fi or move to an area with better coverage. For device compatibility, maintain a tested device matrix covering the top 50 smartphone models in India (which account for over 80 percent of the market) and implement graceful degradation that disables resource-intensive features on low-end devices.
Document Quality and OCR Rejection Problems
Document-related failures account for 15 to 25 percent of Video KYC session issues. The most common problem is poor document image quality when customers hold their PAN card, Aadhaar, or other OVDs in front of the camera. Glare from overhead lighting, motion blur from shaky hands, partial obstruction by fingers, and low ambient light all produce images that OCR engines cannot reliably process. When the platform's OCR fails to extract the document number, name, date of birth, or address, the agent must manually enter this information -- slowing the session and increasing error risk -- or the session is flagged for re-capture.
Laminated documents pose a particular challenge. Many Indian customers laminate their PAN cards and Aadhaar letters for protection, but the lamination creates reflective surfaces that cause severe glare under indoor lighting. The customer may need to tilt the document at multiple angles to find a glare-free position, extending the session time and frustrating both parties. Faded or damaged documents are another common issue -- PAN cards issued before 2010 often have worn text and degraded photographs that challenge both OCR engines and visual verification by agents.
A subtler document-related failure occurs when the OCR correctly extracts text but the extracted data does not match the institution's existing records. Name spelling variations between Aadhaar (which uses the UIDAI transliteration standard) and PAN (which uses the Income Tax Department's records) are surprisingly common. "Rajesh Kumar" on PAN versus "Rajesh Kumarr" on Aadhaar, or "Mohammed" versus "Mohammad," triggers a mismatch flag that requires manual agent review and sometimes session escalation.
Solutions for document quality issues include implementing real-time document capture guidance that overlays a frame on the camera view showing the customer exactly where to position their document, with live feedback on focus quality, glare detection, and edge visibility. Advanced OCR engines trained specifically on Indian identity documents (with their unique fonts, layouts, and security features) achieve 95 to 98 percent extraction accuracy compared to 80 to 85 percent for general-purpose OCR. For name matching, implement fuzzy matching algorithms with configurable thresholds (Levenshtein distance, Soundex, or Jaro-Winkler similarity) that can flag probable matches for agent confirmation rather than outright rejection.
Liveness Check Failures: Lighting, Angle, Spoofing Flags
Liveness detection is a critical component of Video KYC that confirms the person on camera is a real, physically present human being rather than a photograph, video replay, or deepfake. RBI's V-CIP guidelines mandate that the verification process must ensure the customer is alive and present during the session. However, liveness checks account for 10 to 20 percent of Video KYC failures, often due to environmental factors rather than actual fraud attempts.
Poor lighting is the most common liveness failure trigger. Customers conducting Video KYC from their homes often sit with a window behind them (creating a backlit silhouette), under dim room lighting (producing noisy, low-contrast video), or under harsh fluorescent lights (creating unnatural skin tone and hard shadows). Active liveness systems that require the customer to perform actions like turning their head, blinking, or following a moving dot on screen are particularly sensitive to lighting -- head-turn detection algorithms need clear visibility of facial landmarks, which degrade rapidly in low-light conditions.
Camera angle and distance create another category of liveness failures. Customers frequently hold their phone too close (cutting off the top of their head or chin), at an extreme angle (looking down at a phone on a desk), or too far away (making facial features too small for reliable analysis). Passive liveness systems that analyze texture, depth cues, and micro-expressions from the video stream require a frontal face view occupying at least 30 to 40 percent of the frame for accurate assessment. Deviations from this optimal positioning generate low confidence scores that either fail the liveness check or require the agent to intervene with positioning instructions.
False positive spoofing flags are the most frustrating category of liveness failures because they flag legitimate customers as potential fraudsters. Customers wearing thick glasses with reflective coatings, heavy makeup that smooths skin texture, or face masks (still common in some settings) can trigger spoofing alerts in systems trained to detect unnatural facial characteristics. Similarly, customers with certain skin conditions or facial scarring may produce texture patterns that confuse liveness algorithms. The solution is a multi-layered liveness approach that combines passive analysis (texture, depth, reflection patterns) with active challenges (random head movements, expression changes) and allows agent override with documented justification when the automated system flags a legitimate customer.
Agent-Side Errors and Training Gaps
While technical failures get the most attention, agent-side errors contribute to 10 to 15 percent of Video KYC session failures. The most common agent error is procedural non-compliance: failing to follow the prescribed V-CIP sequence (identity confirmation, document verification, live photo capture, face matching) in the correct order, skipping mandatory verification steps under time pressure, or not reading the required disclosure statements to the customer. When quality audits identify these procedural gaps, the entire session must be repeated even if the customer's identity was correctly verified.
Technical errors by agents include capturing blurry screenshots of documents (by clicking the capture button while the customer is still adjusting the document position), accepting documents that are partially obscured or have visible tampering, and failing to notice mismatches between the customer's stated details and the information on their documents. These errors often stem from inadequate training: many institutions provide only 2 to 3 days of initial agent training, with minimal ongoing quality coaching. The RBI expects V-CIP agents to be "trained officials" with thorough knowledge of KYC requirements, document security features, and fraud indicators -- a standard that requires continuous training investment.
The fix for agent-side errors is threefold. First, implement guided workflows that enforce the correct V-CIP sequence through the platform interface, preventing agents from advancing to the next step until the current step is properly completed. Second, invest in comprehensive initial training (minimum 5 days covering regulatory requirements, document verification, fraud detection, and platform usage) supplemented by weekly quality review sessions where supervisors review a sample of completed sessions with each agent. Third, use AI-assisted quality checks that automatically flag potential issues (blurry captures, incomplete steps, low face-match scores) for supervisor review before the session is marked as complete.
Customer Drop-off Mid-Session: UX Friction Points
Customer abandonment during an active Video KYC session accounts for 10 to 20 percent of failures and represents the most preventable category. The primary UX friction points that drive mid-session drop-off include excessive wait times in the agent queue (customers who wait more than 3 minutes are 60 percent more likely to abandon), confusing permission prompts (camera, microphone, and location permissions on mobile browsers often require multiple taps and can appear intimidating to less tech-savvy users), and lack of progress indication (customers who do not know how many steps remain or how long the process will take are more likely to abandon).
Language barriers are a significant but often overlooked friction point. Video KYC platforms that conduct sessions only in English or Hindi exclude a substantial portion of India's population. A customer who does not fully understand the agent's instructions may struggle with document positioning, fail to respond appropriately to liveness challenges, or become anxious and disconnect. The RBI's emphasis on financial inclusion means that V-CIP implementations must accommodate India's linguistic diversity -- particularly for institutions serving rural and semi-urban markets where Hindi and English literacy may be limited.
Reducing customer drop-off requires a UX-first approach to Video KYC design. Pre-session preparation is critical: send customers a clear checklist of what they need (specific documents, well-lit room, stable internet) via SMS before the scheduled session. Implement a pre-session technical check that verifies camera, microphone, and network quality before connecting to an agent, so issues are resolved before the session clock starts. Display a clear progress indicator showing the current step and estimated remaining time. Offer multilingual agent support and UI text in regional languages. Most importantly, minimize the total number of steps and interactions required -- every additional screen, prompt, or instruction is an opportunity for the customer to abandon.
How to Reduce Video KYC Failure Rates Below 5%
Achieving a sub-5-percent failure rate requires a systematic approach that addresses all four failure categories simultaneously. On the technical front, invest in a platform with adaptive bitrate video, robust reconnection logic, and broad device compatibility. Ensure your OCR engine is trained specifically on Indian identity documents and implements real-time capture guidance. Use multi-layered liveness detection that combines passive and active methods with agent override capability.
On the operational front, implement rigorous agent training with a minimum initial certification program and ongoing quality reviews. Use guided workflows that enforce procedural compliance through platform design rather than relying solely on agent discipline. Establish a quality assurance team that reviews at least 10 to 15 percent of completed sessions weekly and provides individual feedback to agents. Track failure reasons by category and agent, identifying patterns that indicate training needs or platform improvements.
Co-browsing technology is arguably the single most impactful feature for reducing failure rates. When an agent can see the customer's screen in real time and provide guided assistance -- drawing attention to the camera permission prompt, showing where to position the document, adjusting camera settings -- the resolution rate for technical issues increases dramatically. Institutions using co-browsing-enabled Video KYC platforms consistently report first-attempt completion rates of 80 to 90 percent, compared to 60 to 70 percent without co-browsing.
BaseKYC is engineered to minimize failure rates across every dimension. The platform's adaptive video engine maintains session stability on connections as low as 500 Kbps. Built-in co-browsing allows agents to guide customers through every step in real time. The OCR engine, trained on millions of Indian identity documents, achieves 97 percent first-attempt extraction accuracy. The multi-modal liveness system combines passive texture analysis with randomized active challenges, maintaining a false rejection rate below 2 percent. And the guided workflow engine ensures that agents follow the correct V-CIP sequence every time, with automated quality checks before session completion. Institutions using BaseKYC consistently achieve first-attempt completion rates above 85 percent and overall failure rates below 5 percent.