A mid-sized NBFC in Rajasthan was growing at 40% year-over-year. Its credit underwriting was sound. Its investor deck was sharp. Its cost of capital was competitive. Its founder — twelve years in financial services, a previous stint at a large bank — understood credit risk well. And then the collections started slipping. 

It was not that borrowers stopped paying. The repayment intent was there. What was missing was the operational infrastructure to collect. Field agents were visiting borrowers but not recording the visits in any system. Promise-to-pay commitments were made and forgotten because they lived in individual WhatsApp conversations rather than a shared platform. Overdue accounts were escalated based on who shouted loudest rather than which account was most recoverable. The collections team lead was spending half his day reconciling the previous day’s cash receipts across three different cash books, two spreadsheets, and a bundle of handwritten receipts. 

The portfolio was growing. The GNPA was growing faster. 

This is the lending execution problem. It is not a credit problem. It is not a strategy problem. It is the gap between what a lending institution decides to do and what actually gets done at the ground level, every day, across hundreds of field agents and thousands of active loans. 

It is the most common cause of NBFC underperformance in India. It is almost never what NBFC founders believe their problem is. And it is almost entirely fixable — if the right operational infrastructure is in place. 

Lending Does Not Fail on Strategy. It Fails in Execution. 

The phrase on the LoanWise platform is precise: “lending doesn’t fail on strategy, it fails in execution.” It deserves unpacking, because the temptation for most NBFC founders who are experiencing performance problems is to look upward — toward credit policy, toward pricing, toward portfolio mix — rather than downward, toward the specific operational failures happening at the field level every day. 

Credit risk is real. Regulatory compliance is real. Cost of capital is real. None of these are the subject of this article, because none of them is where most NBFC operational losses actually occur. The losses occur in execution — in the gap between what the operations head decided should happen and what the field agent actually did, recorded, and reported. 

What Execution Failure Actually Looks Like — Six Ground-Level Scenarios 

Operational failure in lending is not dramatic. It does not announce itself. It accumulates in the ordinary, unremarkable texture of a working day — in a message left unread, a visit recorded nowhere, a promise neither system captured nor anyone followed. 

These six scenarios are drawn from patterns that appear consistently in NBFC operations that are running on informal systems. Each one is financially quantifiable. Together they represent the operational profile of a lending institution that is working hard and delivering less than it should. 

Scenario 1: The Unrecorded Field Visit  

A field agent drives 45 minutes to visit a borrower who is 30 days past due. The borrower is home. The conversation happens. The borrower explains a temporary cash flow problem and commits to paying within a week. The agent drives away, calls his supervisor on WhatsApp to give a verbal update, and moves on to the next visit. 

Nothing of this visit exists in any system. The date is not recorded. The borrower’s statement is not documented. The commitment is not captured. The supervisor’s WhatsApp message disappears down the thread within hours. A week later, nobody follows up on the commitment because nobody remembers it exists. Three weeks later, the account is at DPD 60 and the cost of collection has increased substantially. The initial visit cost the organisation ₹800 in agent time and fuel. The unrecorded commitment cost it several multiples of that in follow-up expense and delinquency. 

Scenario 2: The Promise-to-Pay That Nobody Tracked  

Collection efficiency in most NBFCs running manual operations is not measured by whether a promise-to-pay was made and kept. It is measured by whether cash was received. This means that the single most valuable leading indicator of collection performance — the ratio of promises kept to promises made — is invisible to operations management. 

A field collection team making 200 visits per week might be extracting 50 payment commitments. If the promise-to-pay follow-up system lives in individual agent WhatsApp messages, the operations head has no visibility into which 50 commitments are outstanding, which have been kept, and which have been missed. She is managing her collections team by outcome (cash received) rather than by process (promises made and followed up). Outcome management in collections is like steering a car by watching the rear-view mirror. You know where you have been. You have no idea where you are going. 

Scenario 3: The KYC Document That Was Never Linked  

A loan application is processed. The field executive collects the KYC documents — Aadhaar, PAN, bank statements — and delivers them to the branch office. They are scanned and uploaded, but the upload is to a shared drive folder rather than to the loan record in the system. When the credit team needs to verify the application, they cannot find the documents. They call the field executive. The field executive is on his next visit and does not respond immediately. The application sits in a pending queue. 

The borrower, who needed the funds for a time-sensitive business payment, begins calling. His relationship with the dealer who referred him begins to sour. The NBFC loses a day in disbursement. Depending on the interest rate and loan size, one day of delay per loan across 1,000 active disbursements represents a material cost — both in direct interest foregone and in the borrower relationship damage that is harder to measure but equally real. 

Scenario 4: The Approval That Required a Physical Signature  

A loan application has cleared credit assessment. It needs disbursal approval from a senior credit officer who is visiting a branch 200 kilometres away. The approval cannot be given digitally because the approval workflow is not configured in a system — it exists as an email convention and a physical stamp on a printed form. The credit officer returns to the main office on Thursday. The loan is disbursed on Friday. The borrower needed the funds by Wednesday. 

This scenario sounds like an edge case. In NBFCs running on informal approval workflows, it is a routine occurrence. The delay is invisible in any reporting because no system records the gap between credit clearance and disbursal instruction. The operations head sees only the disbursal date, not the readiness date. The bottleneck is silent. 

Scenario 5: The Collections Reconciliation That Ate the Afternoon  

At 5pm, the collections team lead begins reconciling the day’s collections. Cash collected by field agents is recorded in individual agent diaries. UPI payments are visible in the company’s payment gateway dashboard. NACH debits are recorded in the bank statement. These three sources need to be reconciled against the loan ledger and against the day’s collection target by account. 

This reconciliation — which in a well-structured system takes minutes — takes two hours in an NBFC running on spreadsheets. Because the cash records, the UPI dashboard, and the bank statement are in different formats, the reconciliation is manual. Errors are common. When an error is found, tracing it requires going back to the field agent who made the collection, who is no longer in the office, who responds slowly on WhatsApp. 

The direct cost of this daily two-hour reconciliation across a team of two people is approximately ₹15,000 per month in staff time. The indirect cost is the decisions not made because the reconciliation was not complete — overdues not escalated because the data was not clean, accounts not prioritised because the status was uncertain. 

Scenario 6: The Dealer Who Had Three Good Months and One Catastrophic Quarter  

An NBFC with a dealer-led lending model — consumer durables, electric vehicles, agricultural equipment — relies on dealers to source borrowers and maintain relationships throughout the loan tenure. A dealer in Tier-2 Maharashtra sourced ₹3 crore in loans in six months. The loans performed well. The dealer got a larger allocation. 

What the credit and operations team did not see — because the visibility into dealer-level performance was not structured into any monitoring dashboard — was that the dealer had changed his business practices. He had begun sourcing borrowers aggressively in a geography where repayment culture was weaker, motivated by the origination fee rather than the loan quality. By the time the GNPA from his portfolio surfaced in the monthly review, it represented a six-month accumulation of risk that real-time dealer monitoring would have surfaced in week three. 

The lending execution gap does not announce itself. It accumulates in small, unremarkable operational failures — an unrecorded visit, a missed follow-up, a delayed approval — until the portfolio data finally makes it visible six weeks after the damage was done.

The Financial Cost of Operational Execution Gaps 

Each of the six scenarios above has a financial cost. The aggregate cost — across a portfolio of meaningful size — is not marginal. It is material, and it is recoverable if the right operational infrastructure is in place. 

The figures above are conservative and calculated independently for each failure mode. An NBFC with ₹100 crore portfolio running on informal operations is typically experiencing several of these simultaneously. The aggregate operational cost — including missed collections, excess back-office headcount, disbursement delays, and undetected dealer risk — routinely exceeds ₹1.5–2 crore annually. At a 2% net interest margin, that is a material drag on profitability that appears nowhere in the credit risk analysis but shows up clearly in EBITDA underperformance. 

Sources: LoanWise product data (40% faster processing, 30% quicker disbursement, 2x collection efficiency). NBFC operational cost modelling based on published benchmarks. 

Why Spreadsheets and WhatsApp Work at ₹10 Crore and Fail at ₹100 Crore 

The transition from informal operations to structured systems is not a technology decision. It is a portfolio mathematics problem. And the mathematics has a fairly consistent breakpoint. 

At ₹10 Crore Portfolio  

The same NBFC now has 3,000 active loans. There are 25 field agents across five districts. The collections team has seven people. The credit team has four underwriters and two analysts. The dealer network has 18 active dealers. The operations head receives updates from 25 agents, seven collections staff, and 18 dealers — some on WhatsApp, some by phone, some in person, some not at all. 

At this scale, the human bandwidth that made informal operations workable at ₹10 crore is no longer sufficient. Not because the people are less capable — they are the same people, often more experienced. The problem is combinatorial: 3,000 active loans × 25 agents × 18 dealers × daily EMI cycles × weekly field visits creates a management complexity that no human working memory can track. The WhatsApp groups that were manageable at 300 loans are now noise. The Excel reconciliation that took 20 minutes at 300 loans takes 3 hours at 3,000 loans and is riddled with errors. 

The breakpoint in our observation tends to occur at portfolio sizes between ₹25 crore and ₹75 crore. Below that range, informal systems function with strain. Above ₹75 crore, they function only by shedding visibility — by accepting that the operations head cannot see everything, that some field visits are not recorded, that some follow-ups do not happen. That accepted invisibility is the operational execution gap. And it costs money every day it remains. 

At ₹100 Crore Portfolio  

The same NBFC now has 3,000 active loans. There are 25 field agents across five districts. The collections team has seven people. The credit team has four underwriters and two analysts. The dealer network has 18 active dealers. The operations head receives updates from 25 agents, seven collections staff, and 18 dealers — some on WhatsApp, some by phone, some in person, some not at all. 

At this scale, the human bandwidth that made informal operations workable at ₹10 crore is no longer sufficient. Not because the people are less capable — they are the same people, often more experienced. The problem is combinatorial: 3,000 active loans × 25 agents × 18 dealers × daily EMI cycles × weekly field visits creates a management complexity that no human working memory can track. The WhatsApp groups that were manageable at 300 loans are now noise. The Excel reconciliation that took 20 minutes at 300 loans takes 3 hours at 3,000 loans and is riddled with errors. 

The breakpoint in our observation tends to occur at portfolio sizes between ₹25 crore and ₹75 crore. Below that range, informal systems function with strain. Above ₹75 crore, they function only by shedding visibility — by accepting that the operations head cannot see everything, that some field visits are not recorded, that some follow-ups do not happen. That accepted invisibility is the operational execution gap. And it costs money every day it remains. 

The Five Operational Capabilities That Separate Lenders Who Scale from Those Who Plateau 

Across lending operations at different scales and in different product categories, we consistently observe that the organisations that maintain profitability as they grow share five specific operational capabilities. These are not the capabilities that appear in investor presentations. They are the ground-level execution capabilities that prevent the Lending Execution Gap from compounding as the portfolio grows. 

Capability 1: Digital Application Flow with KYC Linked at Point of Capture  

The application process is the first point where operational discipline either exists or does not. A digital application flow — where KYC documents are captured, verified, and linked to the loan record at the moment of collection — eliminates the document-chasing problem at source. When the credit team opens a loan file, every document is there, linked, and verifiable. The delay between application and credit assessment is technical, not logistical. 

This sounds like a minimum standard. For the majority of NBFCs operating below ₹200 crore portfolio, it is not yet standard. Applications are collected on paper forms. KYC is collected separately. The linking happens manually, at the branch, by someone who may not have the right training or sufficient time. 

Capability 2: Real-Time Field Visibility with Proof-Backed Activity Recording  

Field operations in lending are invisible in most NBFCs. An agent leaves the office, visits accounts, and returns. Whether the visit happened, what was discussed, what commitment was extracted, and what the next action should be is a reconstruction exercise that happens at the end of the day, imperfectly, from memory. 

Real-time field visibility — GPS-tracked visits recorded in a system, with the ability for agents to log notes, commitments, and outcomes at the point of the visit — changes this completely. The operations head can see where every agent is, which accounts have been visited, and what the outcome of each visit was. When a borrower makes a promise-to-pay, it is recorded in a system that automatically triggers a follow-up on the committed date. The institutional knowledge of the field operation is no longer locked in individual agent memories and WhatsApp threads. 

Source: RBI FSR June 2025 flagged that 31–180 DPD microfinance stress rose from 4.7% to 6.5% between September 2024 and March 2025, indicating that early-stage delinquency management — exactly the stage where field visit recording matters most — is where the battle is won or lost. 

Capability 3: Proof-Backed Collection Recording with Reconciliation Workflows  

A collection that is not recorded with proof is a collection that can be disputed, duplicated, or lost. Proof-backed collection recording means that every cash, UPI, or NACH payment is captured in the system at the moment of collection, with the agent, date, account, and mode of payment recorded. Reconciliation against the loan ledger happens automatically rather than manually. 

The benefit is not just accuracy — though accuracy matters. It is the speed and completeness of the end-of-day position. When an operations head can see the day’s collections against the day’s target, by account, by agent, and by overdue bucket, at 5pm without waiting for manual reconciliation, she can make the same-day decision to escalate accounts that missed collection rather than discovering it the next morning. 

Capability 4: Structured Dealer Management with Performance Monitoring  

For lenders operating dealer-led models — consumer durables, electric vehicles, agricultural equipment, gold — the dealer relationship is both the primary origination channel and the primary early-warning system for portfolio quality. A well-performing dealer surfaces good credit profiles. A deteriorating dealer — one whose referral quality is dropping, whose borrowers’ repayment rates are falling — is the earliest indicator of a portfolio quality problem, well before the GNPA data makes it visible. 

Structured dealer management means tracking each dealer’s portfolio on an ongoing basis: origination volume, disbursal rates, 30-day and 60-day DPD rates, collection efficiency by geography. When a dealer’s portfolio shows early stress indicators, the response can happen at month two rather than quarter three. That is the difference between a correctable allocation decision and a credit loss. 

Capability 5: Connected Monitoring Across the Loan Lifecycle  

The operations head of a ₹100 crore NBFC needs to see, in one view, the current state of the portfolio: active loans by stage (disbursed, current, 0–30 DPD, 30–60 DPD, 60–90 DPD, 90+), field agent performance by collection efficiency, dealer performance by portfolio quality, and processing pipeline by application stage and aging. Without a connected monitoring layer, this view is assembled manually — if it is assembled at all — from multiple sources that describe different time periods and use inconsistent definitions. 

With a connected monitoring layer, it exists automatically, updated with each day’s activity, and visible to the right people at the right level of detail. The operations head sees the portfolio. The area manager sees her territory. The field agent sees his accounts. The credit officer sees the application pipeline. Nobody needs to call anybody else to understand the current state. 

The NBFCs that scale profitably are not the ones that hired smarter people or underwrote better credits. They are the ones that built the operational infrastructure that makes their field teams systematically effective rather than heroically effective.

Why Operational Infrastructure Is No Longer Optional in 2026 

The case for structured lending operations has always been financial. In 2026, it is also regulatory. 

The RBI Digital Lending Directions 2025 — which replaced and materially strengthened the 2022 Digital Lending Guidelines — extended regulatory obligations explicitly to post-disbursement loan management, including recovery. Every collections interaction an NBFC initiates now carries documentation, disclosure, and audit trail requirements. The 2025 regulatory cycle has added reporting, documentation, and audit trail requirements that manual workflows cannot satisfy at any meaningful portfolio scale. 

The cost of non-compliance is quantified and recent. <cite index=”8-1″>RBI levied ₹48 crore in aggregate penalties on NBFCs for collection-related Fair Practices Code violations in FY 2024-25 alone. Individual penalties for FPC violations range from ₹5 lakh to ₹2 crore per instance, with repeat offenders facing the possibility of license-level action.</cite> 

Beyond direct penalties, a regulatory show-cause notice diverts management attention for months, mandates third-party audits as a remediation condition, and in cases involving large borrower populations, triggers consumer protection proceedings that run separately from the regulatory process. 

Manual collection workflows — WhatsApp instructions, handwritten cash receipts, verbal field reporting — cannot produce the audit trail that RBI now requires. An NBFC that has been operating informally for three years and is now facing a regulatory examination needs to produce records of every collection attempt, every agent interaction, and every borrower communication across a portfolio of thousands of accounts. That is not possible from a folder of WhatsApp backups. 

The Right Implementation Sequence: Why KPI Definition Comes Before Technology 

The most common implementation mistake is starting with the technology. An organisation decides to deploy Tableau, or to build a custom dashboard, and the first call is with a vendor or a development team. Requirements are gathered. Data sources are identified. The build begins. 

Six months later, the stakeholders who were not in the requirements meeting are looking at metrics they did not ask for, defined in ways that do not reflect how they actually measure performance, connected to data sources that capture 60% of what matters. The organisation has built exactly what was specified in the requirements document. The requirements document did not capture what was actually needed. 

The correct sequence is specific and it inverts the common approach. 

Phase 1: Discovery and KPI Definition (4–6 weeks) 

Before any data is touched, conduct structured conversations with each role that will use the system. Not about what charts they want, but about what decisions they make, what information they need to make those decisions well, and what currently prevents them from having that information. From these conversations, derive the specific metrics that matter — defined precisely, with calculation methodologies, targets, and thresholds. This is the intellectual foundation of the entire system. Time spent here directly determines how useful the finished system is. 

Phase 2: Data Architecture and Integration (6–12 weeks, depending on complexity) 

With the KPIs defined, the data architecture work begins. This means auditing every source system that contains data relevant to the defined metrics, building the integration layer that extracts, normalises, and reconciles that data, and creating the data quality framework that catches and handles inconsistencies before they propagate into the reporting layer. This phase is unglamorous. It is also determinative. The data architecture phase is where the 67% data trust problem is solved — or left unaddressed. 

Phase 3: Visualisation and Alerting Layer (4–6 weeks) 

Only after the KPIs are defined and the data architecture is solid does the visualisation work begin. By this point, the visualisation decisions are relatively straightforward — the metrics are defined, the data is trusted, the roles are understood. The dashboards are built role by role, each reflecting the specific decision needs of that stakeholder. The alerting thresholds, already defined in Phase 1, are implemented as automated notifications. The connection between metric breach and response is built into the system architecture. 

What Operational Intelligence Looks Like in Practice — Three Engagements 

The following examples are drawn from Mind IT Systems engagements across healthcare, financial services, and business services. They illustrate what operational intelligence delivers when the implementation sequence is followed correctly. 

Healthcare: Real-Time Patient Flow and Department Performance  

A healthcare provider needed visibility into patient flow across departments — OPD throughput, consultation wait times, bed availability, and discharge processing time. The existing reporting was a monthly MIS report compiled manually from three systems. Leadership was making operational decisions on data that was 30 to 45 days old. 

The operational intelligence implementation began with a KPI definition phase that produced 24 metrics across patient experience and operational efficiency dimensions, each with explicit targets and alert thresholds. The data architecture phase connected the HMS, the scheduling system, and the billing system into a unified layer with T-1 refresh. The visualisation phase produced role-specific dashboards for the Medical Superintendent, the OPD Manager, and the Billing Head — each showing only the metrics relevant to their decisions. 

The operational change: department heads began their mornings with yesterday’s truth rather than last month’s report. When OPD waiting time exceeded its defined threshold, the OPD Manager received an alert within minutes rather than discovering the problem in a month-end review. The system did not tell the manager what to do — it told her that attention was needed, while there was still time to act. 

Source: Mind IT Systems Operational Intelligence engagement, healthcare sector. minditsystems.com/case-study/healthcare-ux-design-case-study 

Financial Services: Clarity-Driven Insights for Financial Planning and Reporting  

A financial services organisation needed to replace a manual financial planning and reporting process that was consuming significant analyst time and producing results too slowly to be useful for decision-making. The existing process involved data from multiple accounting and operations systems being manually consolidated in Excel before distribution. 

The KPI definition work identified 15 financial performance metrics, an asset-liability monitoring framework, and an exception reporting structure. The data architecture phase integrated the core banking system, the accounting platform, and the transaction monitoring system. The visualisation layer produced a live financial intelligence dashboard for senior management, with automated exception alerts for any metric that crossed regulatory or internal threshold. 

The outcome: analyst time previously spent on manual compilation was redirected to interpretation and response. Reports that previously took three days to compile were available within hours of the period close. Exceptions that previously surfaced in the monthly review were detected and escalated within the day. 

Source: Mind IT Systems Operational Intelligence engagement, financial services sector. minditsystems.com/case-study/financial-reporting-software-case-study

HealthTech: Centralised Workflow Intelligence for Medical Coding Operations  

A HealthTech company managing medical coding operations across multiple client hospitals needed visibility into workflow performance — coding throughput, query rates, turnaround times, and quality metrics across a distributed team. The existing visibility was a combination of team leader reports and periodic audits that captured performance too late to manage it. 

The operational intelligence system connected the case management platform, the QA tracking system, and the client SLA management system into a unified operational view. Role-specific dashboards gave operations managers visibility into team performance, client managers visibility into SLA compliance by account, and the leadership team visibility into overall operational health and financial performance. Exception alerts flagged throughput drops and SLA risks before they became client-facing problems. 

Source: Mind IT Systems Operational Intelligence engagement, HealthTech sector. minditsystems.com/case-study/coding-workflow-software-case-study

Frequently Asked Questions

The Clear Takeaway 

The dashboard problem is not a tool problem. The organisations that have invested significantly in Power BI, Tableau, or custom-built dashboards and found that decisions did not change are not experiencing a failure of the technology. They are experiencing a failure to build the infrastructure that makes data actionable — the KPI definition work, the data integration work, the alerting architecture, and the role-appropriate view design that collectively determine whether the visualisation layer produces decisions or produces colourful observations that nobody acts on. 

Operational intelligence is the difference between the two. It is not a product you can buy and deploy in a week. It is a system that requires the right implementation sequence — definition before architecture, architecture before visualisation — and the discipline to do that work thoroughly before the charting begins. 

The practical starting point is honest: how does your organisation currently make the decisions that most significantly affect performance? Is there one decision — a resource allocation decision, a customer retention decision, a supply chain decision — where better information, available sooner, would produce a meaningfully different outcome? That specific decision is the right starting point for an operational intelligence conversation. Not “we need better dashboards.” A specific decision, and the specific information that would improve it. 

Mind IT Systems offers a free operational data assessment — a structured conversation about where your data is, how you currently track performance, and what the gap between your current reporting and your decision-making needs actually looks like. No commitment required. If that conversation produces a clear picture of what to build and whether we are the right people to build it, we proceed. If it reveals that your current system is closer than you thought, we say so. 

References 

Shailendra Gupta
(Co-Founder and CEO of Mind IT Systems)

Shailendra Gupta co-founded Mind IT Systems in 2014. Over eleven years the company has modernised and rebuilt software for businesses across fintech, healthcare, supply chain, and business services — in India, the UAE, New Zealand, the UK, and the US. The decision between modernising and rebuilding comes up in almost every legacy engagement we handle, and the right answer is rarely obvious at the outset.