We Saw It Coming: What Byju’s and Go First Taught Us About the Future of Credit Monitoring

Table of Contents:

• The Two Collapses Nobody Saw Coming
• The Crisis of Credit Monitoring: We’re Still in the 1990s
• What Byju’s Revealed: The Gaps in Governance Monitoring
• What a Modern Monitoring System Would Have Caught
• What Go First Revealed: The Hidden Stress in Operational Metrics
• The Common Thread: We Don’t Monitor the Right Things
• What the RBI Learned: The Early Warning System Framework
• How Accumn’s Early Warning System Works
• The Signals Are Knowable, But Not by Hand
• What an AI/ML model like Accumn’s does instead
• Conclusion

The Two Collapses Nobody Saw Coming

In March 2023, Byju’s- India’s most valuable edtech unicorn- suddenly became a cautionary tale. Months of aggressive fundraising, sky-high valuations, and bullish founder messaging masked a reality that was screaming in the numbers: the company was burning cash at an unsustainable rate, couldn’t meet payroll, and had no clear path to profitability.

Lenders, investors, and financial institutions that had extended credit to Byju’s or to its ecosystem of borrowing subsidiaries and related entities, suddenly faced a painful question: How did we miss this?

A few months later, Go First, India’s budget airline operator, filed for insolvency. Again, the collapse appeared sudden. Operational stress signals had been mounting for months- fuel price volatility, pilot shortages, capacity cuts- but the severity of the financial distress wasn’t apparent until it was too late. Lenders to Go First faced the same reckoning.

Here’s what makes these two cases remarkable: The warning signals were visible months- in some cases, a year- before the collapse. They weren’t hidden in complex derivatives or off-balance-sheet accounting tricks. They were embedded in government filings, regulatory records, transaction patterns, and corporate governance data that is publicly available to anyone disciplined enough to look for them.

The question is: Why didn’t we look?

The Crisis of Credit Monitoring: We’re Still in the 1990s

Most lending institutions- banks, NBFCs, fintech lenders- use credit monitoring frameworks built for a different era.

The traditional model works like this:

• Loan is disbursed
• Lender waits for quarterly or annual borrower statements
• Statements are reviewed (weeks or months later)
• If numbers look bad, lender reaches out to borrower
• By then, default is imminent or already happened

This reactive, backward-looking approach worked when:

• Corporate governance was simpler
• Company structures were less complex
• Information moved slowly
• Fraud detection was less sophisticated

Today, none of these conditions hold true. Yet most Indian lenders still operate monitoring frameworks from the 1990s playbook.

The cost of this gap is enormous. According to RBI data, banks detected 45% of NPAs only after they became overdue by 90+ days. For NBFCs, the rate was even worse. By the time a lender detects financial stress through traditional channels, the borrower’s options have narrowed. Loan restructuring becomes harder. Recovery probability drops. Loss severity increases.

What Byju’s Revealed: The Gaps in Governance Monitoring

Let’s examine what Byju’s signals looked like, viewed through the lens of government data and corporate filings.

The Public Trail of Distress

Timeline of Observable Signals (That Were Publicly Available):

• Byju’s announced ₹500 crore in salary cuts
• This announcement, if captured in regulatory filings or board minutes, was a clear stress signal
• Yet most lenders’ monitoring systems would flag this only when contractual borrowing covenants were breached (usually after 90+ days of missed payments)

• Byju’s subsidiary, Aakash Educational Services (acquired for ₹1,000 crore), showed signs of distress
• The parent company’s acquisition debt was now a liability on an underperforming asset
• Company filings would show: (a) goodwill impairment charges, (b) related-party liabilities increasing, (c) diminished asset quality
• A lender monitoring related-party transactions and asset quality would have flagged this as a red signal

• Byju’s missed payments to lenders
• Founder’s credibility, which had been the key basis for credit extension, suddenly evaporated when facts contradicted public messaging
• The divergence between founder claims (“We’re on track to profitability”) and observable financial reality should have been detected months earlier

What a Modern Monitoring System Would Have Caught

A forward-looking, multi-source credit monitoring system would have:

• Tracked related-party transactions: Byju’s had extended credit to related entities (subsidiaries, founder-linked companies). As these deteriorated, the parent company’s cash flow and credit quality worsened.
• Monitored governance quality: Board meeting minutes (when disclosed), director changes, audit qualifications—these are early warning signals. If directors are resigning, if audit reports include “going concern” notes, if board independence is declining, these are stress indicators.
• Analyzed cash burn patterns: By looking at monthly operating metrics (if disclosed), revenue recognition patterns (did they decelerate?), and operating expense trends, a lender could have calculated the company’s runway. Byju’s was burning cash; the runway was 6–9 months at the rate of burn observed in late 2022.
• Benchmarked against cohort: How was Byju’s performing relative to other edtech companies? If Byju’s was losing market share, if user growth was slowing while burning rates accelerated, that divergence was a critical signal.

What Go First Revealed: The Hidden Stress in Operational Metrics

Go First’s collapse teaches a different lesson: Sometimes, the stress signals are in operational data, not financial data.

Observable Pre-Collapse Signals

• Fuel hedging losses: Go First had hedged fuel prices assuming ₹80–90/bbl oil. When Brent crude spiked to $120+/bbl, the hedges became worthless, exposing the airline to unbudgeted fuel costs.
• Capacity cuts: Go First suddenly reduced flights and network coverage. This signal—visible in flight schedules and route announcements—indicated cash flow stress.
• Pilot shortages and attrition: High turnover in critical roles signals operational distress or compensation stress (i.e., pilots are leaving because the company is in trouble, not because of job market conditions).
• Default on supplier payments: Before Go First filed for insolvency, it was rumored to be delaying payments to fuel suppliers and maintenance vendors. Payment delays are observable through credit bureau records and supplier networks.

What a Modern Monitoring System Would Have Caught

A comprehensive monitoring system would have:

• Tracked operational KPIs: If Go First’s capacity utilization was declining, if average ticket prices were being discounted heavily, if load factors were dropping- these operational signals precede financial collapse.
• Monitored credit bureau activity: Supplier payment delays show up in credit bureau records. If Go First was being reported as a slow payer, this was a red flag months before insolvency.
• Analyzed sector-level stress: Energy prices, labor cost inflation, regulatory changes- these are macro signals that affect airlines differently. A lender should have been monitoring sector-specific stress indicators.
• Connected to regulatory filings: If Go First’s regulatory filings (with DGCA, credit rating agencies, stock exchange) showed deteriorating maintenance standards, regulatory action, or financial disclosures, these were stress signals observable before the insolvency filing.

The Common Thread: We Don’t Monitor the Right Things

Both Byju’s and Go First share a fundamental lesson: Lenders were monitoring financial statements, but not monitoring the conditions that make financial statements deteriorate.

Traditional credit monitoring focuses on:

• Quarterly balance sheets (backward-looking, stale data)
• Annual audit reports (arrived 3–4 months after fiscal year end)
• Covenant breaches (by which point, the company is already in distress)

What they should monitor:

• Governance quality and director networks
• Related-party transactions and group-level stress
• Regulatory and compliance behavior
• Operational metrics (capacity, utilization, turnover)
• Cash burn and runway (if available)
• Market share and competitive position
• Litigation and regulatory action

The gap between what we monitor and what we should monitor is the gap where crises hide.

What the RBI Learned: The Early Warning System Framework

The Reserve Bank of India took note of cases like Byju’s and Go First. In 2021, the RBI issued guidelines on Early Warning Systems (EWS) for credit risk monitoring.

The RBI’s EWS framework mandates that lenders monitor:

• Financial indicators: Profitability decline, increasing leverage, cash flow stress
• Non-financial indicators: Regulatory violations, litigation, governance changes, management turnover
• Macro indicators: Sector stress, competitive disruption, regulatory changes

This framework is a step forward. But implementation remains a challenge. Most lenders struggle with:

• Data fragmentation: Financial data comes from one source (borrower statements), governance data from another (regulatory filings), operational data from a third (borrower disclosure or industry reports). Bringing these together requires manual work.
• Timeliness: Annual financial statements arrive months after the fiscal year ends. By then, the crisis has often deepened.
• Coverage: Small and mid-sized borrowers don’t disclose operational metrics or governance data. Monitoring is limited to financial metrics alone.

How Accumn’s Early Warning System Works

Knowing what to monitor is the easy part. The hard part is doing it on every borrower, every day, and turning hundreds of scattered signals into one decision a risk team can act on. That is the specific job Accumn’s Early Warning System (EWS) does. Here is the mechanism, step by step.

It ingests the signals humans can’t track by hand

The EWS pulls continuously from the same public and transactional sources a diligent analyst would check- MCA filings, GST records, court and litigation databases, credit bureau activity, and alternative data- but it does it across the entire book at once. In practice that means tracking 100+ KPIs drawn from 500+ data points across 100+ sources, refreshed on a rolling basis rather than at quarter-end. The director’s resignation, the two-month GST delay, the slipping payment record, the distress at a related group company- these arrive as they happen, not three months after a balance sheet is filed.

The RisQ Engine turns the pattern into a single score

A folder of red flags is not a decision. Accumn’s proprietary RisQ Engine, a machine-learning model back-tested on over 1 lakh entities, weighs every signal together and converts the borrower’s state into one tracked Risk Score. This is where the convergence logic matters: the model reads a delayed filing, a governance change, and a payment slip as a trajectory, benchmarks the borrower against its cohort to catch outliers, and accounts for group-level contagion when a controlling director’s other companies start to weaken.

Alerts route themselves to the right person

A score nobody sees is worthless, so the EWS escalates on its own. When a borrower crosses a policy threshold, the system tiers the response rather than waiting for a covenant breach:

Every score change explains itself

The RBI’s EWS framework and your own audit trail requires a reason, not just a number. The RisQ Engine attaches the reason to every movement: “Risk score increased because GST filing was delayed two months, related-party director disqualified, litigation filed.” Risk managers see the drivers behind a flag, which makes intervention faster and the decision defensible when an auditor asks how it was reached.

The takeaway from Byju’s and Go First was never that lenders weren’t trying. It was that the reading had outgrown what people can do by hand. Accumn’s EWS is the layer that does that reading continuously, scores it consistently, and surfaces it while there is still time to act.

Explore Accumn’s Credit Monitoring & EWS →

The Future of Credit Monitoring: What We Need to Build

The Byju’s and Go First cases point to four imperatives for the future of credit monitoring:

1. Real-Time Data Integration

Instead of waiting for quarterly statements, lenders need real-time access to:

• Corporate governance data: Director changes, board minutes, regulatory filings (from MCA, stock exchanges, regulatory bodies)
• Compliance data: GST filings, tax payment status, license renewals, court cases
• Operational data: (For applicable sectors) capacity utilization, payment patterns, transaction volumes
• Market data: Share price movements, industry rankings, competitive position

2. Predictive, Not Reactive Monitoring

Instead of flagging breaches after they happen, lenders need to:

• Forecast cash runway: If a company is burning cash, how many months of runway does it have at the current burn rate?
• Benchmark against cohort: Is this company performing worse than peers in the same sector/region? By how much?
• Detect anomalies: Is director behavior changing? Are related companies deteriorating? Is the company’s competitive position weakening?

3. Connected Monitoring of Groups and Related Entities

Many corporate borrowers are part of larger groups. Distress at one entity often spreads to related entities. Yet most lenders monitor each loan in isolation.

Future monitoring should:

• Map director networks: If a borrower’s director controls 5 other companies, and 2 of them are in distress, the borrower’s credit quality is at risk.
• Identify related-party exposures: If the borrower has significant payables to a related entity that is now insolvent, the borrower’s cash flow is at risk.
• Monitor group-level stress: Contagion effects- when one group entity fails, others are dragged down- are predictable if you track group-level metrics.

4. Intervention Protocols, Not Just Monitoring

Monitoring without intervention is theater. The future of credit risk management needs:

• Automated alerts: When a borrower hits a stress threshold, alerts should be triggered to relationship managers and risk committees immediately (not weekly or monthly).
• Tiered response protocols: Amber alert → relationship manager contact; Red alert → formal review meeting; Critical alert → recovery proceedings. Each tier has defined actions.
• Scenario planning: If a borrower is at risk, what are the intervention options? Can the loan be restructured? Can collateral be enhanced? What is the recovery probability under each scenario?

What This Means for Lenders Today

If you’re a bank, NBFC, or fintech lender reading this, here’s the uncomfortable truth: Your monitoring framework is probably still designed for the 1990s, but your borrowers are operating in 2024.

The Byju’s and Go First cases are not anomalies. They’re not black swans. They’re the new normal- complex corporate structures, rapid growth, governance stress, related-party exposure, operational volatility. And the signals of distress are there, embedded in public data, waiting to be detected.

The lenders who win in the next decade won’t be those with the fastest loan approvals. They’ll be those with the best risk management- which means the best credit monitoring.

The Signals Are Knowable, But Not by Hand

Everything above leads to one uncomfortable conclusion: the grounds for predicting both collapses existed in public data. What was missing was any practical way to read them across a live portfolio. That gap is not a discipline problem or an effort problem. It is a maths problem, and it is why manual credit monitoring no longer works at scale.

On what grounds a default actually becomes visible

A default is rarely signalled by one dramatic number. It shows up as a convergence of weak signals that only mean something together.

• Pattern, not a single metric: A delayed GST filing, a director resignation, and a slipping payment record are each forgivable in isolation. Stacked on the same borrower in the same quarter, they describe a trajectory.
• Deviation from the cohort: A figure that looks acceptable on its own is often an outlier once you see where it sits against similar borrowers in the same sector and region.
• Group and related-party contagion: When a director controls five companies and two are in distress, the borrower’s risk is already rising, regardless of what its own statements say.
• Lead time in non-financial data: Governance, compliance, and operational signals move months before the financials catch up. That lead time is the entire opportunity to intervene.

The grounds, in other words, are knowable. The hard part is reading them on every borrower, continuously, and weighing them consistently.

Why this cannot be done manually

Run the numbers. A mid-sized lender monitoring a few thousand borrowers, each with hundreds of relevant signals across MCA, GST, courts, bureaus, and alternative data, refreshed continuously, is looking at millions of data checks a month. A team reviewing quarterly statements cannot come close, for three structural reasons.

• Latency: Quarterly and annual statements arrive months stale. By the time a human reads them, the lead time the non-financial signals offered is already gone.
• Scale: No analyst can hold hundreds of signals across thousands of borrowers in their head, every day. Coverage collapses to a handful of large accounts while the long tail goes unwatched.
• Consistency: Two analysts weigh the same red flags differently. Manual monitoring is only as reliable as which file a person happened to open that week.

This is the gap the cases keep pointing to. It does not close by trying harder. It closes with a model.

What an AI/ML model like Accumn’s does instead

This is precisely the work Accumn’s monitoring model is built for. It tracks 100+ KPIs drawn from 500+ data points across 100+ sources, financial and non-financial, refreshed continuously rather than at quarter-end. A proprietary ML model converts those patterns into a single Risk Score, so a borrower’s stress becomes one tracked number instead of a folder a human has to interpret. The model has been back-tested on over 1 lakh entities, predicts default with 90%+ accuracy, and forecasts portfolio risk with 90 to 360 days of lead time, the same window that would have changed the Byju’s and Go First outcomes.

Most importantly, it runs across the whole portfolio at once.

• Continuous tracking, not periodic review: every borrower scored on every refresh, not just the accounts someone remembered to check.
• Automated Early Warning Signals and RBI-mandated RFA (Red Flagged Account) alerts: when stress crosses a threshold, the alert routes itself instead of waiting for a covenant breach.
• Group-aware monitoring: consolidated alerts across every entity in the portfolio, including group companies, so related-party contagion shows up before it spreads.
• Explainability: each score change carries its reason (“GST filing delayed two months, related director disqualified”), which is what the RBI’s EWS framework and your own audit trail require.

The lesson from Byju’s and Go First is not that lenders need to look harder. It is that the looking has outgrown what humans can do by hand. The grounds for foresight are sitting in the data. An AI/ML model is what turns those grounds into a score you can act on, across every loan in your book, while there is still time to do something about it.

Explore how Accumn’s real-time credit monitoring and EWS and its AI-based credit decision models track portfolio risk at scale.

Conclusion:

The title of this article- “We Saw It Coming”- refers not to actual foresight, but to hindsight. We could have seen Byju’s collapse coming. We could have seen Go First’s distress coming. The signals were there.

The real question is: What are we not seeing today? What crises are brewing in your loan portfolio right now, hidden in plain sight?

The future of credit monitoring isn’t about smarter underwriting or faster approvals. It’s about smarter risk management- the ability to see stress signals months before they become defaults, and to intervene while options still exist.

The lenders who embrace this shift will reduce NPAs. They’ll recover more from troubled loans. And they’ll sleep better knowing that the next Byju’s collapse, the next Go First insolvency- they saw it coming, and they managed it.