How the NU research moved through five independent challenges

• Added a fourteen year residual income transition and five year normalization stage.
• Improved reverse valuation, payout transition, discount rate sourcing, dilution disclosure, and scenarios.
• Preserved Challenger 1 FAIL as historical evidence.

• Rebuilt frozen company data and the package generator.
• Created formula driven operating, capital, valuation, dilution, and sensitivity schedules.
• Recalculated and audited 2,552 workbook formulas and caches.
• Ran deterministic, canonical, fixture, archive, and visual checks.

• Calibrated 2027 to H1 2026 and replaced the residual fee plug.
• Extended country capital schedules through 2040.
• Linked the exact 13.129576834% nominal USD cost of equity.
• Reworked dilution, terminal timing, peer calculations, scenarios, and sensitivities.

• Added balanced annual statements and broader funding mechanics.
• Added country capital, parent liquidity, issuance, and dividend suspension rules.
• Expanded sensitivity grids and class specific peer calculations.
• Rebuilt all mutable surfaces and closed byte integrity, but not analytical acceptance.

The fifth repair was launched after Challenger 5, but no repaired package was produced. The process stopped with exit code 130 and transitioned to publication preflight.

Numerical execution was partly reproducible, but the principal valuation artifact was not independently auditable.

1. Explicit multistage valuation and auditabilityTargets and reverse valuation were hard coded. The workbook omitted the full annual residual income and terminal schedule.
2. Operating, credit, and regulatory capital bridgesForecast ROE lacked customer, ARPAC, credit, funding, ECL, expense, tax, RWA, and capital support.
3. Terminal assumptions and transitionExplicit payout and terminal payout were economically inconsistent and lacked a transition.
4. Discount rate and country riskThe 14% sum worked, but its dated inputs, geographic weights, currency basis, and double counting controls were unsupported.
5. Dilution and per share book valuePoint in time equity was divided by a period average diluted EPS denominator.

The repair improved the economics, but targets, build controls, and several core bridges remained unreliable.

1. Valuation artifact auditability and reproducibilityTargets remained hard coded, formula caches were missing, and the builder and data schemas were incompatible.
2. Operating, credit, and regulatory capital bridgesA net margin plug replaced explicit credit, funding, ECL, tax, and country capital mechanics.
3. Sensitivity integrityThe central sensitivity cell did not equal the base valuation and the table was hard coded.
4. Point in time dilutionThe repair mixed dates, treasury shares, options, and antidilutive instruments.
5. Package consistency and accounting controlExpected return, reverse ROE, sensitivity, share count, and BVPS fields contradicted each other.
6. Peer cross check and scenario probability supportPeers were unnormalized Yahoo fields and 25/50/25 probabilities lacked evidence based calibration.

Formula presence and several mathematical identities passed. Economic calibration and cross surface control did not.

1. Operating and credit bridge calibration2027 produced implausible fee and earnings jumps. Fee income remained a balancing item.
2. Country capital, fungibility, and payout capacityCapital references pointed two years ahead, stopped in 2035, and did not transmit injections into owner value.
3. Terminal transition and dividend cross check2040 did not reach terminal economics and the DDM used an extra growth factor.
4. Nominal USD cost of equity and country risk controlThe exact sourced 13.129577% result was not linked into the central model.
5. June 30 dilution completenessRSUs, options, acquisition awards, and antidilutive ranges lacked a complete point in time bridge.
6. Sensitivity integrity and completenessPayout and country risk were missing. Other shocks did not rerun the complete model.
7. Filing normalized peer cross checkDisplayed multiples were secondary fields rather than calculations from filed shares, owner earnings, equity, FX, and ADR terms.
8. Decision tree calibration and frozen source consistencyTrigger thresholds differed across the frozen data and PDF, while scenarios had identical first year assumptions.
9. Single frozen source and package surface consistencyHard coded PDF capital values contradicted primary filings and the controlled data.

Scenario calibration, capital horizon, exact cost of equity, and trigger consistency improved. Deeper ownership and transmission problems remained.

1. Operating calibration and full balance sheet bridgeForecast assets did not reconcile with liabilities and equity. Several filed anchors were unused.
2. Country capital, payout capacity, issuance, and terminal fungibilityIssuance could fund desired dividends, creating a circular capital constraint. No 2041 recurring capital bridge existed.
3. Terminal convergence and continuing value timingFrozen terminal payouts were overwritten and continuing values retained incorrect timing.
4. Country risk sensitivity and nominal USD cost of equity controlOperator precedence broke the zero shock identity.
5. Diluted claim range and per share transmissionFuture financing claims were applied inconsistently across per share ranges.
6. Sensitivity integrity, full reruns, and central identitySeveral tables were partial overlays. Country buffer and remittance shocks did not transmit.
7. Filed peer trailing multiples and security basisOne preferred ADS price was applied to separately traded common and preferred classes.
8. Single frozen source, cross surface assertions, and closure truthfulnessThe manifest and closure record claimed controls that workbook inspection disproved.

Mechanical balance, source anchors, exact cost of equity, checksums, and archive integrity passed. Analytical acceptance still failed.

1. Reverse valuation reconstruction was wrongThe workbook discounted the wrong terminal component and reconstructed $53.92 instead of $15.17.
2. No-transaction identity failed to activateThe formula tested issue price instead of transaction quantity, so a cached zero proved nothing.
3. Future issuance and current shareholder ownershipFuture share counts were discounted like cash flows. Current owner value was not allocated by issuance vintage.
4. Mature ROE convergence and shocked terminal statesAn unsupported income normalization plug created terminal ROE, and terminal shocks did not converge through 2040.
5. ARPAC to IFRS bridgeThe scope bridge was a residual equal to IFRS revenue minus managerial revenue, not independently forecast components.
6. Balance sheet drivers and funding baseDrivers were dead or mislabeled and the $12 billion other funding anchor did not reconcile to filed liabilities.
7. Peer ADR and local market evidencePrices were not refreshed to one date and complete ADR, FX, implied local price, and tolerance checks were absent.
8. Analytically false closureByte integrity passed, but closure records said PASS while the analytical verifier and challenger still failed.