56  Financial Services Analytics: Credit Risk and Compliance

56.1 Why Financial Services Analytics Matters

A bank dashboard misread by half a basis point on a 50,000 crore portfolio is a 25 crore mistake; the same misread audited by a regulator is a fine plus a remediation programme.

Financial services — banks, NBFCs, insurers, asset managers — is the industry where analytics is both the product and the regulator’s audit trail. Every credit score is a model that will be examined; every transaction-monitoring alert is a process that will be inspected; every capital number reported to the central bank is a calculation that must reconcile to the trial balance, the risk system, and the model documentation. The BI analyst on a financial-services desk works under three audiences simultaneously — business managers who want decisions, risk officers who want defensibility, and regulators who want reproducibility.

For a BI analyst, financial services clusters into three jobs that distinguish the industry from generic finance (Chapter 50). Credit-risk analytics answers who do we lend to, at what price, and what happens when they stop paying? — origination scoring, portfolio monitoring, vintage and roll-rate analysis, IFRS 9 / CECL provisioning. Compliance and regulatory reporting answers can we prove to the regulator that we follow the rules? — KYC, AML transaction monitoring, sanctions screening, fair-lending and conduct-risk dashboards. Capital and stress-testing analytics answers do we hold enough capital, and how much would we lose in a downturn? — risk-weighted assets, capital ratios, regulatory stress tests, ICAAP. Lyn C. Thomas (2009) frames consumer credit analytics as the discipline of building scorecards that survive both the business cycle and the model-validation team. Bernd Engelmann & Robert Rauhmeier (2011) collect the analytical infrastructure of Basel — PD, LGD, EAD estimation, validation, stress testing — under one cover; the BI dashboards that surface those parameters are how risk officers and regulators read the bank.

TipThe financial-services dashboard contract

Three rules separate financial-services dashboards from every other kind:

1. Reproducibility is the headline metric. Every number on the dashboard must trace back to a model run, a data lineage, and an as-of timestamp the regulator can rerun.
2. Conservatism beats optimism. When two valid methods disagree, the more conservative number goes on the dashboard with the second clearly footnoted.
3. Audit trails carry version, not just value. Risk and compliance dashboards are versioned artefacts — last year’s view must still be reproducible exactly.

56.2 Credit-Risk Analytics: Origination, Portfolio, Provision

Credit risk runs along the lending lifecycle: origination (whom to approve and at what price), portfolio monitoring (how the book is performing and migrating), and provisioning (how much loss we expect under accounting rules). Each phase has its own visualisation grammar.

TipThree Credit-Risk Phases and Their Visualisations

Phase	Question	Core analytics	Visualisation lens
Origination	Whom to approve, at what price?	Application scorecard, KS/Gini, swap-set analysis.	Score-band approval rate chart, KS plot, Gini lift curve.
Portfolio monitoring	How is the book aging and migrating?	Vintage curves, roll-rate matrix, PD migration.	Vintage chart, transition heatmap, delinquency stacked bar.
Provisioning	What loss must we book under IFRS 9 / CECL?	PD, LGD, EAD, ECL by stage.	Stage-1/2/3 stacked bar, ECL waterfall, scenario-weight chart.

TipThe application scorecard and KS / Gini

A consumer-credit scorecard maps applicant attributes (income, employment, prior delinquencies) to a score that predicts probability of default over a defined horizon. Two visualisations validate the score’s discriminatory power.

The KS (Kolmogorov-Smirnov) plot shows two cumulative distributions on the same axis — cumulative percentage of good customers and cumulative percentage of bad customers, both as functions of score band. The maximum vertical distance between the curves is the KS statistic. KS above 30 is acceptable for retail unsecured lending; above 40 is strong (Lyn C. Thomas, 2009).

The Gini lift curve plots cumulative percentage of bads (y-axis) against cumulative percentage of total population (x-axis), sorted from worst score to best. The area between this curve and the diagonal, doubled, is the Gini coefficient. Gini and KS together are the dashboard score validation tile every model-risk function expects to see.

flowchart LR
  A[Applicant<br/>features] --> B[Scorecard<br/>model]
  B --> C[Score<br/>1-1000]
  C --> D[Score band<br/>cut-offs]
  D --> E{Approval<br/>policy}
  E -->|Above cutoff| F[Approve]
  E -->|Below cutoff| G[Decline]
  C --> H[KS / Gini<br/>validation]
  style F fill:#E6F4EA,stroke:#137333
  style G fill:#FCE8E6,stroke:#D93025
  style H fill:#FFF7E6,stroke:#F4B400

TipVintage curves and the roll-rate matrix

The vintage curve, introduced in Chapter 50 for trade-receivables, is the workhorse view in retail lending. Each origination cohort (auto loans booked in Q1 FY24, personal loans in Q2 FY24, etc.) is plotted as cumulative default rate against months on book. Stacked on a single chart, vintages reveal whether newer originations are deteriorating — a signal that scorecard cut-offs need tightening, or that a risky channel needs throttling.

The roll-rate matrix is a heatmap with rows = current delinquency bucket (Current, 1-30 dpd, 31-60, 61-90, 90+), columns = next-period bucket, and cell colour = transition probability. The diagonal-and-above-diagonal mass shows curing; below-diagonal shows worsening. Roll rates feed directly into IFRS 9 staging and PD estimation.

TipIFRS 9 staging and the ECL waterfall

Under IFRS 9 (and CECL in the US), every loan sits in one of three stages: Stage 1 (12-month ECL — performing), Stage 2 (lifetime ECL — significant increase in credit risk), Stage 3 (lifetime ECL — credit-impaired). The mix matters because Stage 2 and Stage 3 carry far heavier provisions per rupee of exposure.

The standard dashboard view is a stacked bar of exposure by stage, trended monthly, with an ECL waterfall walking from prior-period closing ECL through new originations, stage transitions, write-offs, recoveries, and parameter updates to current ECL. The waterfall is what the audit committee uses to interrogate the provision number — without it, the headline ECL is a black box.

WarningThe hindsight problem in scorecard monitoring

A scorecard built three years ago is being scored on customers it never saw — the loan book, the macro economy, and the application channel mix have all shifted. Population stability index (PSI) and characteristic stability index (CSI) dashboards monitor whether the through-the-door population still looks like the development sample. PSI above 0.25 on any major attribute is the trigger for model recalibration; without the dashboard, the deterioration is invisible until losses have already crystallised (Bernd Engelmann & Robert Rauhmeier, 2011).

56.3 Compliance and Regulatory Reporting

Compliance is the part of financial services where the dashboard is, in effect, evidence in a future regulatory audit. Three workstreams dominate.

TipKYC and customer due diligence

Know Your Customer (KYC) dashboards track who is on the books and whether their documentation is current. Standard tiles include:

• Customer count by KYC status (Full, Simplified, Pending, Expired).
• Days-past-due on KYC review per customer tier.
• Source-of-funds and source-of-wealth completeness for high-risk segments.

The visualisation idiom is the same operations heatmap pattern from earlier chapters: customer segments on rows, KYC review months on columns, cell colour = compliance status. Regulators ask for these views by name during inspections; a customer with expired KYC documentation that the bank kept transacting with is a finding even before any AML question is raised.

TipAML transaction monitoring and the alert funnel

Anti-Money-Laundering (AML) transaction monitoring runs rule-based and model-based detectors over every transaction and produces alerts. The dashboard is a funnel — raw alerts → enriched alerts → analyst review → escalated to investigation → STR (Suspicious Transaction Report) filed → confirmed productive. Banks generate millions of raw alerts; only a small fraction become productive STRs.

flowchart LR
  A[Raw alerts<br/>millions] --> B[Rule-tuning<br/>and de-duplication]
  B --> C[Enriched alerts<br/>with KYC context]
  C --> D[Analyst review]
  D --> E{STR<br/>filed?}
  E -->|Yes| F[Reported to FIU]
  E -->|No| G[Closed]
  F --> H[Productive<br/>confirmation]
  style F fill:#FFF7E6,stroke:#F4B400
  style H fill:#E6F4EA,stroke:#137333

The STR conversion rate (productive STRs divided by analyst-reviewed alerts) is the headline efficiency metric. The dashboard pairs it with time-to-decision and backlog age — regulators inspect the backlog as carefully as the throughput.

TipSanctions screening and false-positive analytics

Sanctions-screening dashboards report against OFAC, UN, EU, and local sanctions lists. The screen rate (matches per million transactions) is meaningless without the false-positive rate — name-matching algorithms produce huge numbers of soft matches that operations teams must clear in seconds. The dashboard view is a small-multiples chart per channel (wire, card, correspondent) showing screen rate and false-positive rate together. Regulators look for evidence of tuning — that the bank is systematically reducing false positives without missing true positives.

TipFair-lending and conduct-risk dashboards

Fair-lending analytics — required under ECOA in the US, similar regimes elsewhere — compares approval rates, pricing, and default rates across protected classes (gender, ethnicity, age) after adjusting for credit quality. The visualisation is a small-multiples bar chart with the protected-class dimension on the x-axis and the metric on the y-axis, plus an adverse-impact ratio annotation. Conduct-risk dashboards extend the idea to mis-selling — concentration of complaints, unsuitable product placements, and high-pressure sales patterns.

WarningCompliance dashboards are evidence, not signal

Operations dashboards exist to drive decisions; compliance dashboards exist to be defensible. The accuracy bar is higher, the change-management is heavier, and any visualisation that obscures the underlying numbers (chartjunk, smoothing, custom colour scales) is a liability. Default to plain bars, dated values, and visible methodology.

56.4 Capital and Stress-Testing Analytics

Capital analytics measures whether the bank holds enough cushion to absorb the losses it might face. Stress testing pushes the same machinery against severe-but-plausible economic scenarios that regulators specify (or that the bank itself constructs as part of ICAAP).

TipRisk-weighted assets and the capital stack

Risk-Weighted Assets (RWA) are the regulator’s view of the bank’s exposure adjusted for riskiness. Capital ratios — CET1, Tier 1, Total Capital — are computed as eligible capital divided by RWA. The standard dashboard view is a stacked bar of capital by tier, trended monthly, with horizontal reference lines at minimum (4.5%, 6%, 8%) and management buffer levels.

A second view is the RWA waterfall: starts at prior-period RWA, walks through credit-risk RWA changes (volume, mix, model recalibration), market-risk RWA changes, operational-risk RWA changes, and regulatory adjustments, landing at current RWA. The waterfall is the chart the CFO presents to the regulator’s quarterly review.

TipStress-test scenarios and the loss-distribution chart

Regulatory stress tests (DFAST in the US, BoE in the UK, RBI Pillar 2 in India) project losses under specified macroeconomic scenarios. The dashboard runs the same model under three (or more) scenarios — baseline, adverse, severely adverse — and plots projected ECL, RWA, capital ratio, and net income over a 9- to 13-quarter horizon.

The visualisation is a small-multiples line chart, four panels (one per metric) by scenario colour, with the regulatory minimum capital ratio drawn as a hard reference line. The chart answers how close to breaching the minimum do we get under severely adverse? in a single glance.

flowchart LR
  A[Macro scenarios<br/>baseline, adverse,<br/>severely adverse] --> B[PD, LGD, EAD<br/>scenario-conditioned]
  B --> C[ECL projection<br/>per quarter]
  C --> D[Capital ratio<br/>projection]
  D --> E{Breaches<br/>minimum?}
  E -->|Yes| F[Capital plan<br/>action required]
  E -->|No| G[Pass with<br/>buffer]
  style F fill:#FCE8E6,stroke:#D93025
  style G fill:#E6F4EA,stroke:#137333

56.5 Common Pitfalls

CautionWhat goes wrong

1. Scorecard performance reported only on full-population basis. Hides reject inference and channel-mix shifts. Always slice by recent vintage and origination channel.
2. Vintage curves stacked but not annotated. Without scorecard-version and policy-change annotations, the audience cannot interpret why one cohort moved.
3. ECL number without scenario-weight transparency. IFRS 9 ECL is a probability-weighted average across scenarios; show the weights and individual-scenario ECLs alongside the headline.
4. AML alert dashboards reported only as alert volume. Regulators read backlog age and STR conversion rate; volume alone tells them nothing.
5. Sanctions screen rates without false-positive context. A 12 percent screen rate is uninformative without the false-positive denominator.
6. Capital ratio shown as a single tile. Always trend it; always show the minimum and management buffer alongside.
7. Stress-test outputs without governance trail. The dashboard must record the scenario specification, the model versions used, and the sign-off chain — without it the test fails the reproducibility contract.
8. Fair-lending analytics without statistical significance. Adverse-impact ratios computed on small populations produce noise; show confidence intervals or sample sizes.
9. Compliance dashboards distributed by email screenshot. The audit trail vanishes the moment the dashboard becomes a PNG. Use Power BI Service or Tableau Server with view-history logging.
10. Forgetting model risk. Every dashboard sitting on a model output should link to the model documentation, the validation report, and the model risk rating; without those links the dashboard is undocumented analytics.

56.6 Illustrative Cases

NoteThree case sketches

Yuvijen Bank scorecard recalibration. Risk analytics team builds a Power BI PSI dashboard that monitors through-the-door population stability across the consumer-loan portfolio. PSI on self-employed proportion breaches 0.25 within four months of a new digital channel launch — the channel is bringing in a meaningfully different applicant mix. Scorecard recalibration in the next quarter improves Gini from 47 to 53 and reduces the 6-month default rate on the affected segment by 1.8 points.

Yuvijen NBFC AML alert backlog. Compliance team replaces a monthly alert-volume report with a Tableau dashboard that shows alert backlog age, STR conversion rate, and analyst-load distribution. The view reveals that 38 percent of alerts are sitting more than 30 days unassigned — a regulator finding waiting to happen. After redistributing alerts and tightening rule thresholds, backlog age drops below 7 days within two months and STR conversion rate rises from 0.4 to 1.1 percent. The compliance officer presents the dashboard at the next regulatory inspection without preparation; it answers the inspector’s questions live.

Yuvijen Bank stress-test capital trajectory. Capital management team builds a Power BI dashboard for the annual ICAAP submission. The four-panel small-multiples (ECL, RWA, CET1, net income) under baseline / adverse / severely-adverse scenarios shows CET1 dipping to 8.4 percent at quarter 6 of severely adverse — within the regulatory minimum but below the 9.5 percent management buffer. The board approves a 1,800 crore capital pre-funding action; the dashboard is the artefact submitted alongside the ICAAP narrative.

56.7 Hands-On Exercise: Build a Bank Risk and Compliance Dashboard

NoteThree-page bank dashboard

Aim. Build a three-page financial-services dashboard in Power BI that ties credit-risk monitoring, AML compliance, and capital ratios together, with the reproducibility, audit, and access discipline the function requires. Tableau equivalents are noted.

Scenario. You are the BI lead in risk and compliance at Yuvijen Bank. The Chief Risk Officer has asked for a dashboard that lets her see, in the Monday risk-committee meeting, which credit cohorts are deteriorating, which AML alerts are at risk of breaching SLA, and where the capital ratio is trending against the management buffer.

Deliverable. A three-page Power BI report — Credit Risk, Compliance, Capital — with role-based RLS, an as-of timestamp on every page, model-documentation drill-throughs, and a CRO summary that consolidates the highest-impact items.

56.7.1 Step 1 — Load and model the data

Use Get Data in Power BI to load six CSVs:

• loans.csv — LoanID, CustomerHash, Product, BookedDate, OriginationScore, Vintage, Stage, ExposureCurrent, DPDBucket.
• loan_events.csv — LoanID, EventDate, EventType (Disbursal, Payment, Default, Cure, WriteOff), Amount.
• aml_alerts.csv — AlertID, CustomerHash, AlertDate, RuleCode, RiskScore, AssignedTo, Status, ClosedDate, STRFiled.
• kyc_status.csv — CustomerHash, KYCTier, LastReviewDate, ExpiryDate, Status.
• capital_ratios.csv — Date, EntityCode, RWA_Credit, RWA_Market, RWA_Operational, CET1, Tier1, TotalCapital.
• stress_scenarios.csv — ScenarioCode, Quarter, MacroVariables (long format), ProjectedECL, ProjectedCET1, ProjectedNetIncome.

Type the columns. Build a DimDate calendar; mark it. Build a DimProduct table with Product, Segment, RegulatoryAssetClass. Build a RoleSecurity table mapping user UPN to allowed entity codes and product segments.

56.7.2 Step 2 — Page 1: Credit risk

Build five visuals.

Vintage chart. Line chart with one line per origination quarter, x-axis = months on book, y-axis = cumulative default rate. Annotations mark scorecard-version and policy-change boundaries.

Roll-rate matrix. Matrix with DPDBucket_Start on rows, DPDBucket_End on columns, cell value = transition probability over the last quarter. Conditional formatting from green (no movement) through amber to red (worsening transitions).

Stage-1/2/3 stacked bar trend. 100 percent stacked column trended monthly, with bands for IFRS 9 Stage 1, 2, 3.

ECL waterfall. Power BI Waterfall visual walking Prior-period ECL → New originations → Stage transitions → Write-offs → Parameter updates → Current ECL.

PSI tile. Card with current PSI on key origination attributes (income, employment, channel), with conditional colour: green under 0.10, amber 0.10-0.25, red above 0.25.

DAX measures:

StageWeightedECL =
SUMX(loans,
    SWITCH(loans[Stage],
        1, loans[ExposureCurrent] * loans[PD12m] * loans[LGD],
        2, loans[ExposureCurrent] * loans[PDLifetime] * loans[LGD],
        3, loans[ExposureCurrent] * loans[LGDImpaired],
        0
    )
)

PSI =
SUMX(
    VALUES(applicants[FeatureBand]),
    VAR p_now = DIVIDE(CALCULATE(COUNTROWS(applicants), applicants[Period] = "Current"),
                       CALCULATE(COUNTROWS(applicants), applicants[Period] = "Current", ALL(applicants[FeatureBand])))
    VAR p_dev = DIVIDE(CALCULATE(COUNTROWS(applicants), applicants[Period] = "Development"),
                       CALCULATE(COUNTROWS(applicants), applicants[Period] = "Development", ALL(applicants[FeatureBand])))
    RETURN (p_now - p_dev) * LN(DIVIDE(p_now, p_dev))
)

Tableau alternative: vintage chart with cohort dimension on Colour; matrix native; stacked column native; waterfall via Gantt marks; card via single-number sheet.

56.7.3 Step 3 — Page 2: Compliance

Build four visuals.

KYC compliance heatmap. Matrix with CustomerTier on rows and KYCStatus on columns, cell value = customer count. Conditional formatting flags Expired and Pending in red.

AML alert funnel. Funnel showing Raw alerts → Enriched → Analyst-reviewed → Escalated → STR filed → Productive, with stage-to-stage conversion percentages.

Backlog age trend. Line chart of median alert age in days over the last 13 weeks, with reference line at the 7-day SLA.

Sanctions screening table. Table per channel with screen rate, false-positive rate, true-positive count, and trend sparkline.

Tableau alternative: heatmap native; funnel as sorted bar; line with reference line; table with sparkline via reference-line trick.

56.7.4 Step 4 — Page 3: Capital

Build four visuals.

Capital-ratio trend. Line chart of CET1, Tier 1, and Total Capital trended monthly, with horizontal reference lines at regulatory minima and management buffer.

RWA waterfall. Power BI Waterfall visual walking Prior-period RWA → Credit RWA changes → Market RWA changes → Op RWA changes → Adjustments → Current RWA.

Stress-test small-multiples. Four panels (ECL, RWA, CET1, Net Income) with three lines per panel (baseline, adverse, severely adverse), reference line at minimum CET1.

Capital actions log. Table of dated capital-management actions (issuance, dividend, RWA optimisation) with amounts and approver. The audit-trail tile.

Tableau alternative: line chart native; waterfall via Gantt marks; small-multiples via Trellis; table native.

56.7.5 Step 5 — As-of timestamps and lineage

Every page must carry a header strip with: as-of date, source-system snapshot timestamp, model version, scenario code (where applicable), report-generation timestamp. The strip is built as a single text-box visual driven by a DAX measure that pulls the metadata. Without this strip, the dashboard does not satisfy the reproducibility contract.

56.7.6 Step 6 — Role-Based Security and audit logs

Implement Power BI RLS:

• Front-line credit officer. Sees their own segment’s credit-risk page only; AML and capital pages hidden.
• Risk officer. Sees all credit-risk and capital pages; AML aggregate but not customer detail.
• Compliance officer. Sees AML and KYC pages with customer-hash detail; not other pages.
• CRO. Sees everything plus the CRO summary page.
• Internal audit. Time-bound role with read-only access to all pages plus the audit log.

Enable workspace audit logging mandatorily — for regulated dashboards this is a control evidence, not a configuration choice.

56.7.7 Step 7 — CRO summary and model-documentation drill-throughs

Build a fourth page consolidating: top three credit cohorts on watch, top three AML alert backlogs, capital ratio versus minimum and buffer. Each tile drill-throughs to the relevant detail page and to a separate metadata page that links to the model documentation, validation report, and model-risk rating in the bank’s model inventory. This linkage is what makes the dashboard a defensible artefact rather than a stranded chart.

TipConnect to the Visualisation Layer

Financial-services analytics relies on the visualisation grammar built earlier in the book, with the regulatory layer added on top. Vintage curves and waterfalls (Chapter 50) carry credit risk and capital. Funnel charts (Chapter 49) carry AML alert flow. Heatmaps (Chapter 12) carry KYC compliance and roll rates. Small-multiples (Chapter 12, Chapter 51) carry stress-test scenarios. The reproducibility, RLS, and audit-trail discipline of Chapter 36 is more central in financial services than anywhere else — every dashboard is a piece of regulatory evidence. The storytelling discipline of Chapter 48 is what turns a credit-risk dashboard into a board recommendation that survives the model-validation team and the regulator both.

TipFiles and Screen Recordings

Power BI three-page bank risk and compliance dashboard with RLS and audit logs (yuvijen-bank-risk.pbix), Tableau equivalent (yuvijen-bank-risk.twbx), anonymised workshop dataset (yuvijen-bank-data.xlsx), CRO summary export (yuvijen-bank-cro-summary.pdf), and a screen recording of the dashboard tour (yuvijen-bank-walkthrough.mp4) will be embedded here.

Summary

Concept	Description
Financial-Services Contract
Reproducibility First	Every number traces back to a model run, data lineage, and as-of timestamp
Conservatism over Optimism	When two valid methods disagree, the more conservative number leads
Versioned Audit Trail	Risk and compliance dashboards are versioned artefacts; last year's view must be reproducible
Three Industry Jobs
Credit-Risk Analytics	Whom to lend to, at what price, what happens when they stop paying?
Compliance and Reporting	Can we prove to the regulator that we follow the rules?
Capital and Stress Testing	Do we hold enough capital, and how much would we lose in a downturn?
Credit-Risk Phases
Origination	Application scorecard, KS, Gini, swap-set analysis at the front gate
Portfolio Monitoring	Vintage curves, roll rates, PD migration as the book ages
Provisioning	PD, LGD, EAD, ECL by IFRS 9 stage
Origination Tools
Application Scorecard	Maps applicant attributes to a default-probability score
KS Plot	Maximum vertical distance between cumulative-good and cumulative-bad curves
Gini Lift Curve	Cumulative bads vs cumulative population sorted worst-to-best score
Portfolio and Provision
Vintage Curves	Cohort default rates plotted against months on book to spot deterioration
Roll-Rate Matrix	Heatmap of delinquency-bucket transition probabilities over a period
IFRS 9 Stage 1, 2, 3	Performing, significant-increase-in-credit-risk, credit-impaired stages
ECL Waterfall	Walks prior-period ECL through originations, transitions, writeoffs to current ECL
PSI and CSI Monitoring	Population and characteristic stability indices monitor through-the-door drift
Compliance Visuals
KYC Compliance Heatmap	Customer tier by KYC status with red flags for expired and pending
AML Alert Funnel	Raw alerts to enriched to reviewed to escalated to STR filed to productive
STR Conversion Rate	Productive STRs over analyst-reviewed alerts is the headline efficiency metric
Backlog Age Trend	Median alert age in days against the SLA reference line
Sanctions Screening	Screen rate and false-positive rate per channel must travel together
Fair-Lending Adverse-Impact	Approval, pricing, default rate by protected class with adverse-impact ratio
Compliance as Evidence	Compliance dashboards exist to be defensible, not just to drive decisions
Capital and Stress
Capital Ratio Trend	CET1, Tier 1, Total Capital trended with regulatory minimum and management buffer
RWA Waterfall	Walks prior RWA through credit, market, op, and adjustments to current RWA
Stress-Test Small-Multiples	ECL, RWA, CET1, Net Income across baseline, adverse, severely adverse scenarios
Scenario-Conditioned PD-LGD	PD and LGD parameters re-estimated under each macro scenario
Common Pitfalls
Pitfall: Full-Population Score Reporting	Hides reject inference and channel-mix shifts in the headline number
Pitfall: Unannotated Vintages	Without scorecard-version annotations the audience cannot interpret cohort movement
Pitfall: Opaque ECL Number	IFRS 9 ECL is probability-weighted; show the weights and per-scenario ECLs
Pitfall: Volume-Only AML	Regulators read backlog age and STR conversion rate, not just alert volume
Pitfall: Screen Rates Without FPR	A 12 percent screen rate is uninformative without the false-positive denominator
Pitfall: Single-Tile Capital Ratio	Capital ratio always trended with minimum and management buffer alongside
Pitfall: Stress Without Governance	Dashboard must record scenario specification, model versions, and sign-off chain
Pitfall: Email Screenshot Distribution	Audit trail vanishes the moment the dashboard becomes a PNG
Pitfall: Missing Model Documentation	Every model-driven dashboard must link to documentation, validation, model-risk rating
Hands-On Dashboard
Page 1 — Credit Risk	Vintage chart, roll-rate matrix, stage stack, ECL waterfall, PSI tile
Page 2 — Compliance	KYC heatmap, AML alert funnel, backlog age trend, sanctions screening table
Page 3 — Capital	Capital-ratio trend, RWA waterfall, stress-test small-multiples, capital-actions log
As-Of Timestamp Strip	As-of date, snapshot timestamp, model version, scenario code, report-generation timestamp
Role-Based RLS	Front-line, Risk, Compliance, CRO, Internal Audit roles with separate scopes
Model-Documentation Drill-Through	Tiles drill-through to model documentation, validation report, model-risk rating

Engelmann, B., & Rauhmeier, R. (Eds.). (2011). The basel II risk parameters: Estimation, validation, stress testing (2nd ed.). Springer.

Thomas, L. C. (2009). Consumer credit models: Pricing, profit and portfolios. Oxford University Press.