Insights
The Payment Is Overdue. The NPL Flag Says No: What Vanke Reveals About Credit Classification State
Reuters reported in September 2026 that Chinese financial regulators had informally asked some major banks not to classify overdue loans to China Vanke as non-performing, while also encouraging maturity extensions and, in some cases, deferred interest collection. The reported guidance is not a published regulatory rule, but it exposes an important private-credit data problem: contractual payment state, days past due, forbearance, restructuring, regulatory classification, accounting...
This starts a temporary private draft. It is not public, listed for sale or shared automatically.
A loan can miss a payment.
The payment can remain overdue.
The maturity can be extended.
Interest collection can be postponed.
The borrower can still be under financial pressure.
And the loan can still carry a label that says:
Performing
That is not a contradiction if the label and the underlying credit state are treated as different things.
It becomes a problem only when they are collapsed into one field.
China Vanke now provides an unusually clear example.
On 22 September 2026, Reuters reported that Chinese financial regulators had informally asked some major banks not to classify overdue loans to China Vanke as non-performing.
The reported guidance also included extending repayment deadlines and, in some cases, postponing interest collection.
Reuters: China asks banks to keep Vanke loans off bad-debt books, sources say
Reuters described the instructions as window guidance.
That matters.
This was not presented as a published change to China’s formal loan-classification rules.
It was reported as informal supervisory direction.
The distinction is exactly where the DaDepo lesson begins.
A loan does not become economically healthy because the label stays “performing”.
One credit exposure can have several different states
A weak system may represent a loan using:
Status: Performing
or:
Status: Non-performing
That looks simple.
It is also dangerous.
A single loan can simultaneously have:
Contractual payment state: Overdue
Days past due: 45
Maturity: Extended
Interest collection: Deferred
Forbearance: Yes
Regulatory classification: Performing / Watch
Accounting impairment: Separate assessment
External rating: Deteriorated
Economic value: Discounted
Collection state: Active
Those states can diverge.
They should not overwrite one another.
The contractual state comes first
Every loan begins with contractual obligations.
A borrower may owe principal, interest, fees, scheduled amortisation, balloon repayment, covenant compliance and other contractual performance.
The first credit-state question is therefore:
Did the borrower perform according to the contract?
If the answer is no, the record should preserve:
- which amount was due;
- when it was due;
- how much was paid;
- how much remains unpaid;
- how many days have elapsed;
- whether the lender waived anything;
- whether the maturity changed; and
- whether interest continues to accrue.
That state exists independently of regulatory classification.
Payment state is not classification state
Consider a simplified example.
A loan has:
Principal due: RMB 1bn
Due date: 1 September
Amount paid: RMB 0
On 15 September:
Days past due: 14
The contractual fact is clear.
The payment is overdue.
A bank may then apply a regulatory category based on days past due, borrower capacity, collateral, restructuring, impairment, supervisory instructions, applicable rules and professional judgment.
That classification is another state.
The system should preserve both.
China’s formal framework already separates risk grades
China’s 2023 Commercial Bank Financial Asset Risk Classification Measures divide financial assets into five categories:
Normal
Special mention
Substandard
Doubtful
Loss
The last three categories are classified as non-performing assets.
The rules state that banks should classify assets using principles of:
- truthfulness;
- timeliness;
- prudence; and
- independence.
They also require classification to reflect the debtor’s ability to perform and changes in the asset’s risk profile.
The framework is therefore already more nuanced than:
Paid / Not paid
But the Vanke report shows that even this regulatory layer can differ from the observable payment layer.
Days past due remains an objective fact
The Chinese framework contains explicit overdue thresholds.
For non-retail assets, the rules provide that principal, interest or income that becomes overdue should generally move an asset at least into the special mention category, subject to limited short technical delays.
They also state that:
More than 90 days overdue
should generally be classified at least as substandard.
Substandard is already within the non-performing category.
This is important because:
Days past due
is observable.
It can be calculated from:
Contractual due date
+
Payment received date
Regulatory classification is an applied judgment or supervisory outcome.
The two should be linked.
They should not be merged.
A classification exception should not rewrite payment history
Suppose a lender receives supervisory guidance that says:
Do not classify this exposure as NPL yet
The system should not respond by changing:
Payment date
or:
Days past due
or:
Amount unpaid
Those facts did not change.
Instead, the system should add another state:
Regulatory classification: [category]
Classification basis: Supervisory guidance / bank decision
Effective date: [date]
Source: [source]
Reviewer: [party]
This preserves the underlying asset history.
Forbearance is another layer
If a bank extends a maturity or postpones interest collection, that may represent some form of accommodation.
The exact legal and regulatory label depends on the transaction and jurisdiction.
From an asset-data perspective, however, the lifecycle is clear:
Original terms
->
Payment pressure
->
Lender accommodation
->
Revised terms
That accommodation needs its own fields.
For example:
Original maturity
Revised maturity
Original payment schedule
Revised payment schedule
Interest collected / deferred
Waiver
Grace period
Amendment date
Effective date
Approval
Source document
A classification field cannot substitute for those terms.
Extension is not repayment
This distinction is fundamental.
Suppose:
RMB 500m due today
The maturity is extended by six months.
The loan may no longer be contractually overdue after the amendment takes effect.
But that does not mean:
RMB 500m repaid
The obligation still exists.
The timeline changed.
A structured system should record:
Original due date
Amended due date
Reason for amendment
Consideration
Approval
Effective date
Payment actually received
Without that history, later analysis may incorrectly conclude that the original obligation was fulfilled on time.
Interest deferral is not interest payment
The same applies to interest.
If interest collection is postponed:
Interest accrued
is not:
Interest paid
A loan record may need to distinguish:
Interest contractually due
Interest accrued
Interest capitalised
Interest deferred
Interest waived
Interest paid
Interest overdue
These values can have different legal, accounting and economic effects.
Vanke’s debt scale makes the distinction material
Reuters reported that Vanke had approximately RMB 351 billion of total debt at 30 June 2026.
About 72% was bank borrowing.
The company also reported a first-half 2026 net loss of approximately RMB 14.95 billion.
China Vanke: investor announcements
These numbers do not determine the classification of any specific loan.
They do show why classification treatment matters at scale.
When a borrower has hundreds of billions of renminbi of debt, the difference between:
Overdue
and:
Officially non-performing
can affect bank asset-quality ratios, provisions, capital, investor reporting, loan pricing, restructuring decisions, portfolio sales, risk limits and market interpretation.
That makes classification itself economically consequential.
A regulatory label can influence the market without changing the loan
Consider one bank exposure.
The underlying legal asset may remain:
Loan principal: RMB 2bn
Borrower: China Vanke
Interest: Contractual
Collateral: [if applicable]
The loan’s regulatory category can affect:
- how the bank reports asset quality;
- how management escalates the exposure;
- whether reserves increase;
- whether the asset enters a workout portfolio;
- whether it becomes eligible for NPL sale;
- whether internal capital is allocated differently; and
- how investors interpret the bank’s balance sheet.
The label matters.
But it does not become the asset itself.
Regulatory classification is a viewpoint
A useful way to think about classification is:
Asset
->
Observed facts
->
Rules
->
Institutional judgment
->
Classification
The asset is the loan.
The classification is a conclusion about the loan under a particular framework.
The record should therefore preserve:
Classification framework
Institution applying classification
Classification date
Classification category
Reason
Evidence
Override / exception
Reviewer
This becomes especially important across jurisdictions.
The same loan can have different classifications in different systems
A loan may be viewed under banking regulation, accounting standards, rating-agency methodology, investor credit policy, servicing policy, internal risk grades, securitisation rules, portfolio-sale definitions and regulatory reporting.
The result might look like:
Regulatory category: Special mention
Accounting: Credit-impaired
Internal risk grade: Watchlist
External rating: CCC
Servicing state: Workout
NPL sale eligibility: Yes
None of those fields necessarily contradicts the others.
They answer different questions.
Accounting impairment is not regulatory NPL classification
Accounting frameworks can require expected-credit-loss recognition before or after a regulatory NPL classification threshold is met, depending on the facts and standards.
The important data-model principle is:
Regulatory classification
should not overwrite:
Accounting impairment state
A useful record may contain:
Accounting standard
Stage / impairment category
Expected credit loss
Provision
Valuation date
Model
Reviewer
alongside:
Regulatory class
Regulatory framework
Classification date
The two can then be compared.
Provision is not loss
Another common error is to collapse:
Provision
into:
Realised loss
A bank may increase provisions because expected losses have risen.
The asset can still exist.
Cash may still be recovered.
Collateral may still have value.
Terms may still be restructured.
A later recovery can differ materially from the accounting provision.
Therefore:
Outstanding balance
Provision
Carrying value
Expected recovery
Realised recovery
should remain separate.
Regulatory forbearance is not economic cure
This is the broader Vanke lesson.
If a regulator or supervisor permits a softer classification treatment, that may help financial stability, bank capital management, restructuring, liquidity management or orderly resolution.
It does not automatically improve:
Borrower cash flow
or:
Borrower leverage
or:
Property sales
or:
Recovery value
A classification decision can change reporting state without changing economic fundamentals.
The Asset Passport needs both.
Informal guidance requires stronger provenance
Reuters’ reporting is particularly interesting because the guidance was described as informal.
That creates a provenance problem.
A formal rule may have:
- title;
- issuing authority;
- effective date;
- publication;
- article number; and
- legal text.
Informal supervisory guidance may instead be evidenced through regulator communication, bank internal memoranda, meeting notes, supervisory correspondence or reporting from informed sources.
The data model should distinguish:
Formal regulation
from:
Supervisory interpretation
from:
Informal guidance
from:
Market report
Each has a different evidentiary weight.
“Window guidance” should not become a permanent rule in the database
Suppose a system reads the Reuters article and records:
Chinese regulation: Vanke loans cannot be NPL
That would be an overstatement.
A better structure is:
Reported supervisory guidance
Borrower: China Vanke
Reported recipients: Selected major banks
Reported instruction: Avoid NPL classification for certain overdue loans
Publication source: Reuters
Publication date: 22 Sep 2026
Formal public rule identified: No
Scope: Unclear
Expiration: Unknown
Confidence: Source-reported
This is far more useful.
Classification needs an effective date
Loan classification is not permanent.
A loan can move:
Normal
->
Special mention
->
Substandard
->
Doubtful
->
Loss
It can also improve after repayment, restructuring, cure, collateral realisation, refinancing or improved borrower performance.
A current record therefore needs:
Classification
Effective from
Effective to
Previous classification
Reason for change
Source
Historical state should remain visible.
A loan can move economically before it moves regulatorily
This is common in credit.
The sequence may be:
Sales weaken
->
Cash flow deteriorates
->
Covenant headroom narrows
->
Rating falls
->
Payment delayed
->
Forbearance granted
->
Regulatory classification changes
The classification is often downstream from other signals.
If a system stores only the final classification, early warning disappears.
Early-warning indicators belong in the asset record
For a corporate borrower, relevant signals can include:
- revenue decline;
- operating cash flow;
- leverage;
- liquidity;
- debt maturities;
- refinancing;
- covenant headroom;
- asset sales;
- bond prices;
- bank-loan pricing;
- rating actions;
- payment extensions;
- unpaid interest;
- sponsor support; and
- creditor negotiations.
These signals do not prove default.
They describe credit state.
The borrower and the loan should remain separate
China Vanke is one borrower.
It has many obligations.
Not every instrument necessarily has the same maturity, security, lender, guarantor, payment state, amendment, classification, recovery expectation or legal ranking.
The system should not use:
Borrower status: Distressed
to overwrite every loan.
A useful graph is:
Borrower
->
Loan A
->
Loan B
->
Loan C
->
Bond D
->
Shareholder loan E
Each asset needs its own state.
Shareholder support is another state
Reuters reported that Vanke has relied on financing support from major shareholder Shenzhen Metro.
That support can affect liquidity.
But:
Shareholder support
is not:
Loan repayment
A structured record should distinguish:
- support provider;
- instrument;
- amount;
- date;
- maturity;
- security;
- ranking;
- use of proceeds; and
- whether proceeds were used to repay a specific obligation.
Forbearance creates a versioned contract
If repayment deadlines are extended, the loan agreement has effectively changed.
A structured system should preserve:
Version 1
Original maturity
then:
Amendment
then:
Version 2
Revised maturity
The old maturity should not disappear.
Otherwise later analysts cannot answer:
Was the borrower originally unable to meet the scheduled maturity?
NPL status affects portfolio eligibility
Many investment mandates define eligibility using terms such as:
Performing loan
or:
Non-performing loan
That can determine whether an asset may be held by a bank book, NPL vehicle, securitisation, fund, insurer, warehouse, distressed-debt investor or servicing platform.
If classification can diverge from contractual stress, eligibility rules need to specify:
Which definition?
“Performing” needs a definition
A fund may define performing as:
No payment more than 30 days overdue
A bank regulator may use another framework.
A servicing contract may use:
Current under amended terms
A rating agency may use another default definition.
The word alone is not enough.
A system should store:
Status label
Definition
Framework
Source
Effective date
Credit classification is not universal taxonomy
That is one of the biggest implications for DaDepo.
If Asset Passports are used across jurisdictions, the system cannot assume:
NPL = one universal state
Instead it needs:
Classification framework
Jurisdiction
Regulator
Category
Definition
Effective date
Source
Then data can be compared without pretending different rules are identical.
A calculated NPL indicator is not an official NPL classification
For example:
Days past due: 105
Rule threshold: >90 days generally at least substandard
Calculated indicator: NPL threshold met
Official bank classification: [category]
If those differ, the system should flag:
Classification divergence
not overwrite one with the other.
That can be a very valuable credit-monitoring control.
Divergence itself is information
When contractual state and regulatory classification diverge, the difference may reveal restructuring, regulatory forbearance, timing, data lag, classification error, special supervisory treatment, technical exception or another documented reason.
The system should preserve the reason.
An unexplained divergence deserves review.
Asset quality ratios can depend on classification policy
At bank level, non-performing-loan ratios are typically calculated from classified exposures.
If classification treatment changes, the reported ratio can move even without an equivalent change in borrower cash flows.
That means bank-level asset-quality analysis should distinguish:
Reported NPL ratio
from:
Underlying delinquency
where data is available.
The same principle applies to loan portfolios.
Credit state should be event-driven
A loan can produce events such as:
Payment due
Payment missed
Grace period begins
Grace period ends
Waiver granted
Maturity extended
Interest deferred
Classification reviewed
Classification changed
Provision increased
Restructuring signed
Payment cured
Default declared
Enforcement begins
Recovery received
Each event changes one or more states.
This is more robust than manually editing a single status field.
A credit-state Asset Passport
A useful DaDepo Asset Passport for a stressed loan could contain several linked sections.
Loan identity
- loan identifier;
- borrower;
- lender;
- facility;
- currency;
- original principal;
- outstanding principal;
- signing date;
- funding date;
- maturity;
- interest terms;
- amortisation;
- governing law; and
- current holder.
Contractual payment state
- next due date;
- amount due;
- principal due;
- interest due;
- amount paid;
- payment date;
- unpaid balance;
- days past due;
- grace period;
- default notice; and
- current contractual status.
Amendment and forbearance
- original maturity;
- current maturity;
- extension date;
- payment holiday;
- interest deferral;
- interest capitalisation;
- waiver;
- covenant reset;
- amendment agreement;
- approval;
- consideration; and
- effective date.
Regulatory classification
- jurisdiction;
- regulator;
- classification framework;
- classification category;
- classification date;
- prior category;
- NPL / non-NPL flag;
- basis;
- exception;
- supervisory guidance;
- source type;
- reviewer; and
- next review date.
Accounting impairment
- accounting standard;
- impairment stage;
- credit-impaired flag;
- expected credit loss;
- provision;
- carrying value;
- methodology;
- valuation date;
- reviewer; and
- source.
Internal credit state
- internal risk grade;
- watchlist;
- special-assets flag;
- workout status;
- probability of default;
- loss-given-default assumption;
- recovery estimate;
- collateral coverage; and
- latest credit review.
External credit state
- rating agency;
- borrower rating;
- instrument rating;
- rating date;
- outlook;
- rating action;
- bond / loan market price where available;
- CDS where available; and
- source.
Restructuring state
- restructuring proposed;
- negotiations;
- participating creditors;
- standstill;
- amended terms;
- debt exchange;
- maturity extension;
- principal haircut;
- interest reduction;
- new money;
- completion date; and
- current restructuring status.
Collection and recovery
- servicer;
- collector;
- demand sent;
- amount collected;
- recovery strategy;
- enforcement action;
- collateral action;
- recovery estimate;
- realised recovery;
- recovery date; and
- remaining claim.
Provenance
- loan agreement;
- amendment;
- bank statement;
- payment record;
- bank risk report;
- regulatory rule;
- supervisory communication;
- public filing;
- Reuters or other media report;
- rating release;
- accounting report;
- AI-extracted field;
- professional review;
- effective date; and
- last-updated timestamp.
This turns:
Performing / NPL
into a traceable credit history.
Credit state should be machine-readable
A loan could move through:
Current
->
Payment due
->
Payment missed
->
Past due
->
Special mention
->
NPL
But another path can be:
Payment pressure
->
Maturity extension
->
Current under amended terms
->
Forborne
->
Classification review
Or:
Overdue
->
Supervisory guidance
->
Regulatory NPL flag deferred
->
Economic stress remains
Or:
NPL
->
Restructuring
->
Payments resume
->
Cure period
->
Classification upgraded
The system should preserve the actual path.
“Current” must say under which terms
After restructuring, an asset may be:
Current under amended terms
That is not the same history as:
Never missed a payment
Both may currently show zero days past due.
The historical path remains economically important.
AI can reconstruct credit-state divergence
AI can help extract:
- contractual due dates;
- payment records;
- extension agreements;
- waiver terms;
- interest deferrals;
- risk classifications;
- impairment references;
- provisions;
- ratings;
- restructuring terms;
- regulator communications; and
- recovery records.
Across a portfolio, AI could flag:
- loan marked performing despite long overdue payment;
- maturity changed without linked amendment;
- interest marked paid when actually deferred;
- regulatory classification changed without source;
- NPL classification inconsistent with days past due;
- provision reduced while credit state deteriorated;
- old classification still shown after restructuring;
- upgraded loan without required cure evidence;
- borrower-level stress not reflected in instrument review; or
- media-reported informal guidance incorrectly stored as formal regulation.
That can materially improve credit data quality.
AI should not decide official NPL status by itself
AI should not independently conclude:
- that a regulator has granted an exception;
- that a bank has misclassified a loan;
- that a restructuring is regulatory forbearance;
- that a loan legally defaulted;
- that an amendment cured default;
- that a borrower is insolvent;
- that impairment is required;
- that a provision is adequate;
- that an exposure must be sold;
- that a regulator’s informal guidance applies to a specific facility; or
- that a loan should be upgraded or downgraded.
Those require governing rules, authoritative bank records and professional judgment.
What DaDepo can contribute
DaDepo does not need to become a banking regulator.
It does not need to decide whether Vanke loans are NPLs.
The useful role is information infrastructure.
An Asset Passport can connect:
Loan contract
->
Payment schedule
->
Actual payment
->
Days past due
->
Amendment / forbearance
->
Regulatory classification
->
Accounting impairment
->
Internal credit grade
->
Restructuring
->
Recovery
That can make it easier to answer:
- what the borrower was contractually required to pay;
- what it actually paid;
- what is overdue;
- whether terms were changed;
- whether interest was deferred;
- what regulatory category is currently applied;
- which framework produced that category;
- whether accounting impairment differs;
- whether internal credit views differ;
- what evidence supports each state; and
- how the asset changed over time.
DaDepo does not decide the classification.
It makes the classification explainable.
What DaDepo does—and does not do
Creating or reviewing a credit-state Asset Passport does not mean that DaDepo has:
- classified a loan for regulatory purposes;
- determined NPL status;
- determined default;
- approved regulatory forbearance;
- granted a waiver;
- amended a loan;
- calculated official provisions;
- determined accounting impairment;
- assigned a credit rating;
- determined borrower solvency;
- valued collateral;
- valued the loan;
- approved a restructuring;
- recommended enforcement;
- guaranteed recovery; or
- provided banking, regulatory, legal, accounting, tax, investment, restructuring or valuation advice.
Important: DaDepo provides technology and information tools. It does not provide banking supervision, regulatory classification, credit-rating, accounting, restructuring, legal, investment, tax or valuation advice or services unless a specific service is expressly identified and lawfully provided. Contractual payment status, regulatory classification, impairment, default, forbearance and restructuring depend on governing documents, authoritative records, applicable law, regulatory frameworks and qualified professional review.
A practical credit-classification checklist
Before a loan is described simply as performing or non-performing, ask:
- Loan: Which exact facility is being classified?
- Borrower: Which legal entity owes the payment?
- Current holder: Who owns the exposure today?
- Amount outstanding: What principal remains?
- Payment due: What amount was contractually due?
- Due date: When was it due?
- Payment received: What amount was actually paid?
- Days past due: What is the objective delinquency count?
- Grace period: Does a contractual grace period apply?
- Original maturity: What was the original date?
- Current maturity: Has it been extended?
- Amendment: Which document changed the terms?
- Interest: Is interest paid, accrued, deferred, capitalised or waived?
- Forbearance: Has the lender granted an accommodation?
- Restructuring: Has the exposure been formally restructured?
- Regulatory framework: Which classification rules apply?
- Regulatory category: What category is officially assigned?
- NPL flag: Is the exposure officially reported as non-performing?
- Classification date: When was that decision effective?
- Classification basis: What facts and rules support it?
- Exception: Does any exception or special treatment apply?
- Supervisory guidance: Is there formal or informal guidance affecting the treatment?
- Source type: Is that guidance published, private, reported or inferred?
- Accounting impairment: Is the asset credit-impaired under the relevant accounting framework?
- Provision: What expected-credit-loss provision is booked?
- Internal grade: What does the lender’s own risk system say?
- External rating: Has the borrower or instrument been downgraded?
- Collateral: What security supports the exposure?
- Recovery value: What is the current recovery estimate?
- Collection state: Is the asset in ordinary servicing, workout or enforcement?
- Cure: What would be required to improve the classification?
- History: Are prior missed payments, amendments and classifications preserved?
- Divergence: Do payment facts and official classification disagree?
- Explanation: Is that divergence explained by a source-backed reason?
- Current state: Can every material field be traced to a date and authoritative source?
If the answer to:
“What has the borrower actually paid, what contractual terms apply today, and why does the official classification say what it says?”
is unclear, the credit-state record is incomplete.
The broader lesson goes beyond Vanke
The same distinction matters across private markets.
Corporate loans
Borrowers can receive maturity extensions without curing underlying leverage pressure.
Commercial real estate
A lender may extend a loan while valuation and debt-service coverage continue deteriorating.
Private credit
Managers can amend covenants or payment terms while investor reporting still uses broad performing-loan categories.
Consumer NPLs
Days past due, charge-off, regulatory classification and collection status can differ.
Mortgage portfolios
Forbearance programmes can change contractual payment requirements without eliminating credit risk.
Receivables
An invoice can be disputed, overdue, restructured or insured while retaining a separate accounting or regulatory treatment.
Securitisations
Pool eligibility may depend on a contractual definition of default that differs from bank regulatory NPL rules.
Restructuring
A loan can be current under amended terms after having passed through a period of severe distress.
The infrastructure rule is the same:
Credit labels are interpretations of asset state, not substitutes for asset state.
The NPL flag is not the loan
That is the strongest lesson from the Vanke report.
A bank may show:
NPL: No
while the asset history also shows:
Payment pressure: Yes
Maturity extended: Yes
Interest collection deferred: Possibly
Borrower losses: Significant
Liquidity support: Ongoing
Those facts can coexist.
A robust private-credit system should never force one to erase the others.
The system should preserve:
What happened
separately from:
How the institution classified what happened
and separately again from:
What the asset is worth
That separation matters to lenders.
It matters to investors.
It matters to servicers.
It matters to regulators.
And it matters to anyone trying to understand the true state of a private credit asset.
A loan does not become performing because the label stays “performing”. The label is one state. The payment history is another.
That is why credit classification deserves its own Asset Passport lifecycle.
Further reading
- Reuters: China asks banks to keep Vanke loans off bad-debt books, sources say
- National Financial Regulatory Administration: Commercial Bank Financial Asset Risk Classification Measures
- State Council of the People’s Republic of China: China Banking and Insurance Regulatory Commission and PBOC issue Commercial Bank Financial Asset Risk Classification Measures
- China Vanke: Investor announcements
- Hong Kong Exchanges and Clearing: China Vanke filings