PRM flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

Recovery Time Objective

The maximum acceptable elapsed time to restore a business function or system after disruption. It sets failover design targets and must be shorter than the maximum tolerable downtime established by business impact analysis.

Recovery Point Objective

The maximum acceptable amount of data loss measured in time before a disruption. It determines backup and replication frequency, so an RPO of one hour requires at least hourly replication or snapshots.

Maximum Tolerable Downtime

The total time a business process can be unavailable before causing unacceptable harm to customers, revenue, or regulatory standing. RTO plus verification and reconciliation time must fit within it.

Impact Tolerance

The maximum disruption a firm can absorb for an important business service before breaching its obligations to clients and markets. Firms map services to tolerances and test severe scenarios against them.

Single Point of Failure

A component, person, or dependency whose failure alone halts a critical service because no substitute exists. Resilience work inventories these across people, process, technology, and third parties.

Warm Standby

A pre-provisioned recovery environment kept running with replicated data and periodic synchronization but limited capacity. It supports short RTOs that cold backups cannot meet because provisioning time is removed.

Fourth-Party Risk

Risk arising from a vendor's own subcontractors and upstream providers. Two suppliers sharing one cloud region or processor are effectively a single dependency despite separate contracts.

Incident Severity Classification

A predefined scale assigning incidents to tiers by customer impact, financial loss, and regulatory reporting triggers. The tier determines escalation paths, communications, and post-incident review requirements.

Bias-Variance Tradeoff

Expected prediction error decomposes into squared bias, variance, and irreducible noise. Increasing model complexity lowers bias and raises variance, so regularization and simpler models trade one for the other.

L1 Regularization

A penalty proportional to the absolute value of coefficients added to the loss function. It shrinks some coefficients to exactly zero, producing sparse models and performing automatic feature selection.

L2 Regularization

A penalty proportional to the squared magnitude of coefficients. It shrinks coefficients smoothly toward zero without eliminating them, reducing variance and improving stability when features are correlated.

Precision and Recall

Precision is true positives divided by predicted positives; recall is true positives divided by actual positives. Their harmonic mean is F1, useful when both false positives and false negatives matter.

ROC AUC

The probability a model ranks a random positive case above a random negative case, equivalent to the area under the ROC curve. It is threshold-independent but can look strong under severe class imbalance.

Data Leakage

Contamination of training features with information unavailable at prediction time, such as target-derived or future variables. It inflates validation performance and causes sharp degradation in production.

Concept Drift

A change over time in the relationship between input features and the target, so a previously accurate model becomes miscalibrated. It differs from input distribution shift and requires retraining or recalibration.

Cross-Validation

A resampling method that partitions data into folds, training on some and validating on others to estimate generalization error. Time-ordered data require expanding-window splits rather than random shuffling.

Gradient Boosting

An ensemble method that adds decision trees sequentially, each fitted to the residual errors of the current model, scaled by a learning rate. It reduces bias but can overfit without regularization.

Random Forest

An ensemble of decision trees trained on bootstrap samples with random feature subsets at each split. Averaging decorrelated trees reduces variance and typically resists overfitting better than a single tree.

Strong Customer Authentication

A requirement that electronic payments use at least two independent factors from knowledge, possession, and inherence, dynamically linked to the amount and payee. It underpins open banking and card payment security.

Tokenization

Replacing a primary account number with a surrogate token stored in a secure vault. It reduces PCI DSS scope because merchants and processors no longer hold usable card credentials.

Interchange Fee

The fee an acquirer pays an issuer on each card transaction, set by network rules and sometimes capped by regulation. It forms the largest component of the merchant discount rate.

Merchant Discount Rate

The total fee a merchant pays to accept a card payment, equal to interchange plus scheme fees plus the acquirer's margin. It is quoted as a percentage of transaction value.

Payment Authorization

The real-time approval step that reserves funds on a cardholder's account and typically expires after a set period. It is distinct from clearing and settlement, which actually move funds.

Instant Payment Irrevocability

Real-time credit transfer schemes settle within seconds and cannot be recalled once credited. Fraud controls therefore shift to pre-transaction scoring and beneficiary name verification rather than chargeback recovery.

Settlement Finality

The point at which a transfer becomes unconditional and cannot be reversed. On distributed ledgers with probabilistic consensus, finality is approximated by confirmation depth rather than a single moment.

Open Banking API

A regulated interface letting licensed third parties access account information or initiate payments with customer consent. Consent management, authentication, and data minimization are core obligations.

Value at Risk

The loss threshold not expected to be exceeded with a given confidence over a horizon. A one-day 99 percent VaR of ten million means losses exceed ten million on about one day in a hundred.

Expected Shortfall

The average loss conditional on exceeding the Value at Risk threshold. It is subadditive and coherent, capturing tail severity that VaR ignores, and is used in Basel market risk capital.

Modified Duration

The percentage change in a bond's price for a one percentage point change in yield, approximately. It equals Macaulay duration divided by one plus yield per period and is negative for long positions.

Expected Loss Formula

Expected credit loss equals probability of default multiplied by loss given default multiplied by exposure at default. It is covered by pricing and provisions, while unexpected loss is covered by capital.

Liquidity Coverage Ratio

High-quality liquid assets divided by net cash outflows over a thirty-day stress period. The ratio must meet a regulatory minimum so banks survive a short-term funding run.

Option Greeks

Delta measures value change per unit underlying move, gamma the change in delta, vega sensitivity to volatility, and theta sensitivity to time decay. Gamma and vega peak near the money and near expiry.