Glossary

A controlled experiment comparing two versions of something by assigning users at random to each, so that a difference in outcome can be attributed to the change.

See also: Statistical significance

An organisation’s rules on which AI tools staff may use, for what purposes, and with what data — including what must never be entered into an external service.

See also: Shadow AI, AI system inventory

Amounts owed to suppliers for goods and services received on credit, recorded as a current liability until paid.

See also: Three-way match

Amounts owed to the business by customers for goods and services delivered on credit, recorded as a current asset until collected.

See also: Expected credit loss (ECL)

An expense or income recognised in the period to which it relates, before the invoice is received or issued. It is reversed when the actual document is recorded.

See also: Accrual basis of accounting, Prepayment, Cut-off

Recognising transactions when they occur rather than when cash changes hands. IFRS financial statements, other than the cash flow statement, are prepared on this basis.

See also: Accrual

Under IAS 10, an event after the reporting period that provides evidence of conditions existing at the reporting date, and so requires the amounts in the financial statements to be adjusted.

See also: Non-adjusting event

A structured assessment, made before deployment, of how an AI system could affect individuals, groups and the organisation, and of the controls needed in response.

See also: AI risk classification, Human oversight

Assigning each AI system a risk level according to the consequence of it being wrong — who could be harmed, how seriously and how easily an error would be noticed — so that controls are proportionate.

See also: AI system inventory, EU AI Act

A register of every AI system an organisation uses or builds, recording what each does, the decisions it informs, its owner, its data and how its outputs are checked.

See also: AI risk classification, Shadow AI

A defined sequence of steps for solving a problem or performing a calculation.

See also: Machine learning

Systematic error in a model’s outputs that disadvantages particular groups, often arising from unrepresentative training data or from features that act as proxies for protected characteristics.

See also: Training data, Explainability

Under IFRS 9, the measurement of a financial asset or liability at its initial amount, adjusted for repayments and for interest calculated using the effective interest method, less any loss allowance.

See also: Expected credit loss (ECL)

Processing personal data so that individuals can no longer be identified by any means reasonably likely to be used. Truly anonymised data falls outside data protection law; pseudonymised data does not.

See also: Pseudonymisation, Personal data

The laws, controls and procedures designed to prevent the proceeds of crime being disguised as legitimate funds, including customer due diligence, monitoring and reporting of suspicion.

See also: Know your customer (KYC), Sanctions screening

Computer systems that perform tasks associated with human intelligence, such as recognising patterns, making predictions, understanding language or generating content.

See also: Machine learning, Generative AI

Under the IFRS Conceptual Framework, a present economic resource controlled by an entity as a result of past events, where the resource is a right with the potential to produce economic benefits.

See also: Liability, Recognition

A single, documented list of every input a financial model relies on, with its source, owner and date. It lets a reviewer trace any output back to the assumption that drives it.

See also: Financial model, Driver-based forecasting

An objective examination that gives a stated level of confidence that a process, control or report works as intended.

See also: Internal audit, Three lines model

A rule or algorithm for sharing credit for a conversion among the marketing touches in a customer’s journey: last-click, first-click, linear, position-based, time-decay or algorithmic. None is a controlled test of which conversions would have happened without the spending.

See also: Incrementality, Marketing mix modelling (MMM), Correlation and causation

The documented path from a figure in the financial statements back through the ledger to the source transaction and its approval.

See also: Journal entry, Internal control

A test in a three-statement model confirming that total assets equal total liabilities plus equity in every period. A balance check that passes because of a plug proves nothing.

See also: Three-statement model, Plug

Agreeing a balance sheet account to independent supporting evidence — a bank statement, a sub-ledger, a schedule — and explaining and resolving any difference.

See also: Reconciliation, Sub-ledger, Month-end close

The scenario a model presents as most likely, built on central assumptions. Upside and downside cases are measured against it.

See also: Scenario analysis

A measure of how much a security’s returns move with the market as a whole. A beta above 1 indicates greater sensitivity to market movements than the market itself.

See also: Capital asset pricing model (CAPM), Cost of equity

The step in a process with the least capacity relative to demand. It limits the output of the whole process, so improvement elsewhere does not raise throughput.

See also: Capacity, Throughput

The level of activity at which total revenue equals total costs, so that profit is zero. It is found by dividing fixed costs by the contribution margin per unit.

See also: Contribution margin

The return on ad spend at which contribution from attributed sales covers the advertising: one divided by the contribution margin rate. The rate is net of payment and delivery costs, returns, discounts and every other variable cost except the advertising.

See also: Return on ad spend (ROAS), Contribution margin, Break-even point

A plan that allocates resources for a period and sets the baseline against which managers are held to account. Its usefulness depends on its stability once agreed.

See also: Forecast, Target, Variance analysis

A documented plan for keeping critical activities running, or restoring them within set times, after a disruption.

See also: Risk register

Labels appended to a link telling analytics which source, medium, campaign and ad sent a visit. A spelling or case difference splits a campaign into several, so links follow a naming convention and are fixed at source, not in reports.

See also: Attribution model, Data quality

The maximum output a process, resource or organisation can produce in a period under normal conditions.

See also: Capacity utilisation, Bottleneck

Actual output as a percentage of capacity. Running close to full utilisation lengthens queues and lead times sharply.

See also: Capacity

A model that estimates the return investors require on equity as the risk-free rate plus beta multiplied by the equity risk premium.

See also: Beta, Cost of equity

Spending on assets expected to provide benefit over more than one period, such as property, plant and equipment. It is capitalised on the balance sheet and depreciated rather than expensed at once.

See also: Free cash flow to the firm (FCFF)

The number of days between paying for inputs and collecting cash from customers, calculated as days inventory outstanding plus days sales outstanding minus days payable outstanding.

See also: Days sales outstanding (DSO), Days payable outstanding (DPO), Days inventory outstanding (DIO)

Managing a group of related purchases as a single strategic unit, with a plan based on the organisation’s demand and the supply market for that category.

See also: Spend analysis, Supply positioning

The structured approach to moving people and processes from a current state to a new one, so that the change is adopted and sustained.

See also: Stakeholder

The structured list of every account in the general ledger, with codes that determine how transactions are classified and reported.

See also: General ledger

A loop in which a calculation depends, directly or indirectly, on its own result — commonly interest calculated on average debt that the interest itself affects. It is resolved by controlled iteration or by an opening-balance convention.

See also: Financial model

Holding evidence, before publication, that supports a factual, comparative, price or environmental claim as it is worded. That competitors make the same claim is not evidence, and evidence about one component does not support a claim about the whole product.

See also: Transcreation, Compliance

A machine learning task that assigns inputs to categories, such as flagging a transaction as fraudulent or not.

See also: Supervised learning, Confusion matrix

Clicks as a percentage of impressions. In organic search, a fall in CTR while position holds can mean the results page is answering the query itself, though more ads above the result or a changed snippet can do the same.

See also: Cost per click (CPC), Conversion rate (CVR), Search engine optimisation (SEO)

The schedule of tasks, owners, dependencies and deadlines for completing the month-end or year-end close.

See also: Month-end close

A valuation method that applies multiples observed for similar listed companies, such as EV/EBITDA, to the company being valued, after normalising for differences.

See also: Enterprise value, EBITDA, Exit multiple

Conformity with the laws, regulations, internal policies and contractual obligations that apply to an organisation.

See also: Compliance risk, Policy

The risk of legal or regulatory sanction, financial loss or reputational damage from failing to comply with applicable rules.

See also: Compliance, Risk register

A situation in which a personal interest could influence, or appear to influence, the performance of a professional duty. It is managed by disclosure and, where needed, by removing the person from the decision.

See also: Governance

A table comparing a classification model’s predictions with actual outcomes, showing true and false positives and negatives.

See also: Precision and recall, Classification

Combining the financial statements of a parent and its subsidiaries into those of a single economic entity, eliminating intra-group balances and transactions. The principles are set out in IFRS 10.

See also: Subsidiary, Intercompany elimination, Non-controlling interest

The maximum amount of text, measured in tokens, that a language model can take into account at once. Material beyond it is not considered, which can cause instructions or facts to be lost in long tasks.

See also: Token, Large language model (LLM)

Under IAS 37, a possible obligation whose existence depends on uncertain future events, or a present obligation not recognised because an outflow is not probable or cannot be measured reliably. It is disclosed rather than recognised.

See also: Provision

Under IFRS 15, an entity’s obligation to transfer goods or services to a customer for which it has already received, or is owed, consideration. Often called deferred revenue.

See also: Revenue recognition, Performance obligation

Managing a contract after award so that both parties meet their obligations and the intended value is delivered, including performance monitoring, variations and closure.

See also: Service level agreement (SLA), Supplier relationship management

Revenue less variable costs, expressed per unit or in total. It is the amount available to cover fixed costs and generate profit.

See also: Break-even point, Operating leverage

In risk management, an action, process or mechanism that reduces the likelihood or impact of a risk.

See also: Residual risk, Control environment

A general ledger account whose balance summarises the detail held in a sub-ledger, such as the receivables control account. It should agree to the sub-ledger total.

See also: Sub-ledger, General ledger

The attitudes, structures and standards across an organisation that set the tone for how controls are designed and followed.

See also: Control, Governance

Conversions as a percentage of the clicks, visits or users that could have converted, with the base named, since each gives a different figure. A funnel stage’s rate is the users completing it divided by those who reached it.

See also: Click-through rate (CTR), Cost per click (CPC), A/B test

Action taken to remove the cause of a detected problem so that it does not recur, as distinct from correcting the problem itself.

See also: Root cause analysis

Correlation is a statistical association between two variables; causation means one brings about the other. Correlation alone does not establish causation.

See also: Statistical significance

The return shareholders require for bearing the risk of holding a company’s equity, commonly estimated using the capital asset pricing model.

See also: Capital asset pricing model (CAPM), Weighted average cost of capital (WACC)

Advertising spend divided by the number of clicks it bought. For a lead goal, the most that can be paid per click is the target cost per lead multiplied by the conversion rate per click.

See also: Conversion rate (CVR), Click-through rate (CTR), Negative keyword

Advertising spend divided by impressions, multiplied by 1,000. Impressions include repeat showings to the same person, so CPM is not the cost of reaching a thousand people, which is spend divided by reach, multiplied by 1,000.

See also: Frequency, Cost per click (CPC)

The decline in response to an ad as the same people see it repeatedly, which fresh creative shown to the same audience restores. In audience saturation, by contrast, reach stops growing and only a new audience restores response.

See also: Frequency, Click-through rate (CTR)

A document issued to a customer that reduces the amount they owe, for example after a return or a pricing error.

See also: Accounts receivable

The cost of winning one new customer: marketing and sales cost in a period divided by the new customers won in it. The highest affordable CAC is the contribution a new customer earns within the payback period the business sets.

See also: Lifetime value (LTV), Payback period, Unit economics

Recording transactions in the correct period, so that revenue, costs, receivables and payables at period-end reflect only what occurred up to that date.

See also: Accrual, Month-end close

The time taken to complete one unit of work from start to finish in a process.

See also: Lead time, Throughput

A visual display of the key measures needed to monitor a process or objective, arranged so the most important information is read first.

See also: Data visualisation, Descriptive analytics

Detecting and correcting errors, duplicates, inconsistent formats and missing values in a dataset before analysis.

See also: Data quality

The roles, policies and standards that determine who owns, may use and is accountable for an organisation’s data, and how its quality and security are maintained.

See also: Data quality, Data lineage

The documented path of data from its source through each transformation to where it is used, so that a figure in a report can be traced back.

See also: Data governance, ETL

The data protection principle that personal data must be adequate, relevant and limited to what is necessary for the purpose.

See also: Personal data

The degree to which data is accurate, complete, consistent, timely and fit for its intended use.

See also: Data cleansing, Data governance

A spreadsheet feature that recalculates a model’s output across a range of values for one or two inputs, used to build sensitivity tables without rewriting formulae.

See also: Sensitivity analysis

Representing data graphically — charts, maps, tables — so that patterns and comparisons can be understood quickly and accurately.

See also: Dashboard

A central store of structured data integrated from several sources and organised for reporting and analysis.

See also: ETL, SQL

The average number of days inventory is held before it is sold, calculated as inventory divided by cost of sales, multiplied by the days in the period.

See also: Cash conversion cycle

The average number of days a business takes to pay its suppliers, calculated as trade payables divided by purchases or cost of sales, multiplied by the days in the period.

See also: Cash conversion cycle

The average number of days it takes to collect payment after a sale, calculated as trade receivables divided by revenue, multiplied by the days in the period.

See also: Cash conversion cycle, Working capital

Under IAS 12, tax payable or recoverable in future periods because of temporary differences between the carrying amounts of assets and liabilities and their tax bases.

See also: Asset, Liability

The ability of a sender’s messages to reach the inbox rather than being blocked or filtered as spam. Bounce, complaint and click rates indicate it; delivery rate counts mail filed as spam, and automatic image loading can inflate open rates.

See also: Suppression list, Marketing consent

The systematic allocation of the depreciable amount of an asset over its useful life, recognising the consumption of its economic benefits.

See also: Useful life, Impairment

Removing a previously recognised asset or liability from the balance sheet, for example when it is sold, settled or expires.

See also: Recognition

Analysis that summarises what has happened, through measures, aggregations and reports.

See also: Diagnostic analytics, Predictive analytics

Analysis that investigates why something happened, by drilling into data, comparing segments and testing explanations.

See also: Descriptive analytics

The rate used to convert future cash flows to their present value, reflecting the time value of money and the risk of the cash flows.

See also: Net present value (NPV), Weighted average cost of capital (WACC)

A valuation method that estimates value as the present value of expected future cash flows, discounted at a rate reflecting their risk, plus a terminal value.

See also: Terminal value, Free cash flow to the firm (FCFF), Discount rate

Forecasting built from the operational quantities that cause financial results — volumes, prices, headcount, conversion rates — rather than from prior-period totals, so that each change in the forecast traces to a named driver.

See also: Forecast, Rolling forecast

A structured investigation undertaken before a decision — appointing a supplier, onboarding a customer, making an acquisition — to identify risks and verify claims.

See also: Know your customer (KYC)

Earnings before interest, tax, depreciation and amortisation. A widely used proxy for operating cash generation that excludes financing, tax and non-cash charges; it is not a measure defined by IFRS.

See also: Enterprise value

The order size that minimises the combined cost of ordering and holding inventory, calculated as the square root of twice annual demand times the cost per order, divided by the annual holding cost per unit.

See also: Reorder point, Safety stock

Engagements, counting specified actions such as likes, comments, shares and saves, as a percentage of reach. A rate worked on followers or impressions is a different figure, so two rates compare only when the actions counted and the base match.

See also: Frequency, Click-through rate (CTR)

The value of a business’s operations to all providers of capital, broadly equity value plus net debt and other claims such as non-controlling interests.

See also: Equity value, Enterprise-to-equity bridge, Net debt

The reconciliation from enterprise value to equity value, deducting net debt and other claims senior to ordinary shareholders and adding non-operating assets.

See also: Enterprise value, Equity value

The value attributable to a company’s shareholders; for a listed company, its market capitalisation.

See also: Enterprise value

Referring an issue to a higher level of authority under defined criteria, so that it is decided by someone with the authority to act.

See also: Governance

Extract, transform, load: the process of taking data from source systems, converting it into a consistent structure and loading it into a target store such as a data warehouse.

See also: Data warehouse, Data lineage

Regulation (EU) 2024/1689, the European Union’s law on artificial intelligence. It classifies AI systems by risk, prohibits certain practices and places obligations on providers and deployers of high-risk systems.

See also: AI risk classification

A multiple, typically of EBITDA, applied to the final forecast year to estimate terminal value on the assumption that the business is sold at that point.

See also: Terminal value, Comparable company analysis

Under IFRS 9, the probability-weighted estimate of credit losses over the relevant period, recognised as a loss allowance against financial assets such as trade receivables.

See also: Accounts receivable, Amortised cost

The extent to which the reasons for an AI system’s output can be understood by the people who rely on it or are affected by it.

See also: Human oversight, Model card

Under IFRS 13, the price that would be received to sell an asset or paid to transfer a liability in an orderly transaction between market participants at the measurement date.

See also: Amortised cost

An input variable used by a machine learning model, such as transaction amount or customer tenure.

See also: Training data

A condition in a loan agreement requiring the borrower to maintain stated financial ratios, such as maximum leverage or minimum interest cover. A breach can allow the lender to demand repayment.

See also: Net debt, Liquidity

A structured calculation, usually in a spreadsheet, that represents a business or decision in numbers so that the effect of assumptions on outcomes can be examined.

See also: Three-statement model, Model audit

The structured report of an entity’s financial position and performance: a statement of financial position, a statement of profit or loss and other comprehensive income, a statement of changes in equity, a statement of cash flows, and notes.

See also: International Financial Reporting Standards (IFRS)

Further training of a pre-trained model on a smaller, task-specific dataset to adapt its behaviour.

See also: Large language model (LLM), Training data

A root cause technique that asks “why” repeatedly, each answer forming the basis of the next question, until the underlying cause is reached.

See also: Root cause analysis

A budget recalculated for the actual level of activity achieved, so that variances separate the effect of volume from the effect of price and efficiency.

See also: Budget, Variance analysis

The best current estimate of what will actually happen. Unlike a budget, it should change whenever the evidence does.

See also: Budget, Target, Forecast bias

A consistent tendency for forecasts to err in one direction. It is detected by comparing successive forecasts with actual outcomes over time.

See also: Forecast

A control requiring that a transaction or journal be prepared by one person and reviewed or approved by another before it takes effect.

See also: Segregation of duties, Journal approval

Cash generated by a business that is available to shareholders after operating costs, reinvestment and payments to and from lenders.

See also: Free cash flow to the firm (FCFF)

Cash generated by operations after tax and reinvestment in working capital and fixed assets, available to all providers of capital. It is the cash flow discounted at the WACC in an enterprise DCF.

See also: Discounted cash flow (DCF), Weighted average cost of capital (WACC)

In advertising, the average number of times each person reached was shown an ad in a period: impressions divided by reach. Rising frequency with falling click-through fits both creative fatigue and audience saturation, so on its own it shows neither.

See also: Cost per thousand impressions (CPM), Creative fatigue, Engagement rate

Under IAS 21, the currency of the primary economic environment in which an entity operates. Transactions in other currencies are foreign currency transactions.

See also: Presentation currency

The complete record of an entity’s accounts, from which the trial balance and financial statements are prepared.

See also: Chart of accounts, Trial balance, Sub-ledger

AI systems that produce new content — text, images, code, audio — in response to a prompt, based on patterns learned from their training data.

See also: Large language model (LLM), Hallucination

The assumption that an entity will continue in operation for the foreseeable future. Management must assess it, and disclose material uncertainties that cast significant doubt on it.

See also: Financial statements

Under IFRS 3, the excess of the consideration paid in a business combination over the fair value of the identifiable net assets acquired. It is not amortised but tested for impairment at least annually.

See also: Impairment, Consolidation

The system of rules, roles and processes by which an organisation is directed and controlled, including who decides what and how they are held to account.

See also: Three lines model, RACI matrix

Output from a generative AI system that is fluent and plausible but factually wrong or unsupported, such as an invented figure or citation.

See also: Generative AI, Human oversight

A fixed number typed into a formula or calculation cell instead of being linked to an input. Hardcodes are a common cause of model errors because they do not update when assumptions change.

See also: Model audit, Integrity check

A forecast of employee numbers by role and period, used to drive salary, benefit and related costs in a driver-based plan.

See also: Driver-based forecasting

Arrangements that allow people to monitor, question and override an AI system. It is real only where the reviewer has the time, information and authority to intervene.

See also: AI impact assessment, Explainability

Under IAS 36, the amount by which an asset’s carrying amount exceeds its recoverable amount — the higher of fair value less costs of disposal and value in use.

See also: Goodwill, Depreciation

The conversions that marketing caused, over and above those that would have happened without it, measured against a randomly held-out group or matched control regions. Platform- and retailer-reported conversions, by contrast, can include sales that would have happened anyway.

See also: Attribution model, A/B test, Marketing mix modelling (MMM)

The level of a risk before any controls are applied.

See also: Residual risk, Control

A calculation built into a model to confirm it is behaving correctly — a balance check, a cash reconciliation, or a flag when a total does not match its components.

See also: Balance check, Model audit

Removing balances and transactions between members of a group on consolidation, so that the group’s statements show only dealings with outside parties.

See also: Consolidation

An independent function that provides assurance to the board on whether governance, risk management and controls are effective.

See also: Assurance, Three lines model

A process designed to provide reasonable assurance that transactions are authorised, recorded accurately and completely, and that assets are safeguarded.

See also: Segregation of duties, Four-eyes principle

The discount rate at which the net present value of a series of cash flows is zero. It can mislead where cash flows change sign more than once or projects differ in scale.

See also: Net present value (NPV)

The accounting standards issued by the International Accounting Standards Board, together with the International Accounting Standards (IAS) it has adopted, used for financial reporting in many jurisdictions.

See also: Financial statements

Cost of sales divided by average inventory, showing how many times inventory is sold and replaced in a period.

See also: Safety stock

A formal request to suppliers to submit a bid against a defined specification, terms and evaluation criteria.

See also: Request for quotation (RFQ), Tender evaluation criteria

The international standard, published in 2023, specifying requirements for an artificial intelligence management system within an organisation.

See also: NIST AI Risk Management Framework

The review and authorisation of a manual journal entry by someone other than its preparer, with evidence of what was checked.

See also: Journal entry, Four-eyes principle

A record of a transaction in the ledger, with equal debits and credits, a date, a description and supporting evidence.

See also: General ledger, Journal approval

Continuous improvement through frequent small changes, typically led by the people who do the work.

See also: Lean

A measure chosen to show progress against an objective. A useful KPI has a defined calculation, an owner and a target.

See also: Leading indicator

A measure that signals rising exposure to a risk, with thresholds that trigger review or escalation before the risk crystallises.

See also: Risk appetite, Risk register

Verifying the identity of customers and understanding the nature of their business and the source of their funds, as part of anti-money laundering controls.

See also: Anti-money laundering (AML), Due diligence

A generative AI model trained on large volumes of text to predict and produce language, used for drafting, summarising, answering questions and similar tasks.

See also: Generative AI, Token, Context window

Awarding leads points for fit and behaviour, such as job role and pages visited, so that those above a threshold pass to sales. The threshold matches sales capacity, passing the best leads sales can promptly follow up, not the most.

See also: Conversion rate (CVR), Capacity

The total time from a request or order to its delivery, including waiting time as well as processing time.

See also: Cycle time, Reorder point

A measure that changes before the result it predicts — for example, sales pipeline ahead of revenue — and so gives early warning.

See also: Key performance indicator (KPI)

A management approach that improves flow and value to the customer by systematically removing waste from processes.

See also: Waste, Value stream map, Kaizen

Under IFRS 16, a lessee’s obligation to make lease payments, measured at the present value of the payments not yet made.

See also: Right-of-use asset

Under the IFRS Conceptual Framework, a present obligation of an entity to transfer an economic resource as a result of past events.

See also: Asset, Provision

The contribution a customer is expected to generate over the whole relationship, worked from margin per order, purchase frequency and retention rather than revenue. Also called customer lifetime value (CLV), it caps what is worth paying to acquire a customer.

See also: Customer acquisition cost (CAC), Contribution margin, Unit economics

A business’s ability to meet its obligations as they fall due, from cash and assets that can readily be converted to cash.

See also: Working capital, Financial covenant

A branch of AI in which systems learn patterns from data to make predictions or decisions, rather than following explicitly programmed rules.

See also: Supervised learning, Unsupervised learning

Internal financial reports prepared to support decisions, typically monthly. They are not governed by IFRS, but should be prepared consistently and reconciled to the ledger.

See also: Month-end close

A person’s recorded agreement, given freely by positive action, to receive marketing from an identified sender through a named channel. Pre-ticked boxes, bundled consent and a list seller’s generic consent do not count; when consent is required varies by market.

See also: Suppression list, Personal data

A statistical model that estimates each channel’s contribution to sales, offline channels included, from aggregate spend and sales over time, not individual journeys. It models carry-over, where advertising keeps working after it stops, and saturation, where extra spend returns less.

See also: Attribution model, Incrementality, Regression

The threshold above which an omission or misstatement could reasonably be expected to influence the decisions of users of financial information. It is judged by both size and nature.

See also: Management accounts

Purchasing that bypasses agreed contracts, approved suppliers or procurement procedures.

See also: Spend analysis

The middle value of a set of numbers arranged in order. Unlike the mean, it is not pulled by extreme values.

See also: Outlier

An independent review of a financial model’s logic, formulae and assumptions to find errors before the model is relied on.

See also: Financial model, Hardcode, Integrity check

A short document describing an AI model’s intended use, training data, performance, limitations and known risks.

See also: Explainability

The decline in a model’s accuracy over time as the data it meets in use departs from the data it was trained on.

See also: Training data

The process of recording, reconciling and reviewing a month’s transactions so that reliable management accounts can be produced.

See also: Close calendar, Balance sheet reconciliation, Cut-off

A word or phrase that stops a paid search campaign or ad group showing ads for queries that match it, such as job-seeking or wrong-language searches. Negatives are set at launch, then added from the search terms report.

See also: Search intent, Cost per click (CPC)

Borrowings less cash and cash equivalents. Definitions vary, so the items included should be stated.

See also: Enterprise-to-equity bridge, Financial covenant

The sum of the present values of a project’s expected cash flows, including the initial investment. A positive NPV indicates value created at the chosen discount rate.

See also: Discount rate, Internal rate of return (IRR)

A voluntary framework published by the US National Institute of Standards and Technology in 2023 to help organisations identify, assess and manage the risks of AI systems.

See also: ISO/IEC 42001, AI risk classification

Under IAS 10, an event after the reporting period that indicates conditions arising after the reporting date. It is disclosed where material but does not change the amounts reported.

See also: Adjusting event

The equity in a subsidiary not attributable, directly or indirectly, to the parent. It is presented within equity in the consolidated statements.

See also: Consolidation, Subsidiary

The degree to which operating profit responds to changes in revenue, determined by the proportion of fixed costs in the cost base.

See also: Contribution margin

A data point far from the others. It may be an error to correct or a genuine and important observation, and should be investigated rather than removed by default.

See also: Data cleansing, Median

When a model learns the noise in its training data as well as the signal, performing well on that data but poorly on new data.

See also: Test set

The time taken for an investment’s cumulative cash inflows to repay its initial cost. It ignores cash flows after payback and, in its simple form, the time value of money.

See also: Net present value (NPV)

Under IFRS 15, a promise in a contract to transfer a distinct good or service to a customer. Revenue is recognised as each obligation is satisfied.

See also: Revenue recognition, Transaction price

A way of estimating terminal value by treating the final year’s cash flow as growing at a constant rate for ever: next year’s cash flow divided by the discount rate less the growth rate.

See also: Terminal value, Exit multiple

Under UK and EU data protection law, any information relating to an identified or identifiable living individual.

See also: Data minimisation, Anonymisation

A spreadsheet tool that summarises a table of records by grouping and aggregating them along chosen dimensions.

See also: Descriptive analytics

A balancing figure inserted to force a model to agree — for example, to make the balance sheet balance. It hides the error it compensates for.

See also: Balance check

A statement of what an organisation requires and why, approved at an appropriate level. A procedure sets out how the policy is carried out.

See also: Standard operating procedure (SOP), Compliance

Precision is the share of cases a model flags that are truly positive; recall is the share of truly positive cases the model flags. Improving one usually costs the other.

See also: Confusion matrix, Classification

Analysis that estimates what is likely to happen, using statistical or machine learning models built on historical data.

See also: Prescriptive analytics, Machine learning

An amount paid in advance for goods or services to be received in a later period, held as an asset and released to expense as the benefit is consumed.

See also: Accrual

Analysis that recommends a course of action, typically by combining predictions with optimisation or decision rules.

See also: Predictive analytics

Under IAS 21, the currency in which financial statements are presented, which may differ from the functional currency.

See also: Functional currency

The breakdown of a change in revenue or margin into the effects of selling prices, quantities sold and the mix of products or customers.

See also: Variance analysis

Drawing the steps, decisions, handoffs and roles in a process, so that delays, duplication and control gaps become visible.

See also: Value stream map

The instruction and context given to a generative AI system. A well-specified prompt states the task, the relevant context, the constraints and the expected form of the output.

See also: Generative AI, Retrieval-augmented generation (RAG)

Under IAS 37, a liability of uncertain timing or amount, recognised when there is a present obligation from a past event, an outflow is probable and the amount can be estimated reliably.

See also: Contingent liability, Liability

Replacing identifying details with a code, so that data cannot be attributed to an individual without separately held information. Pseudonymised data remains personal data.

See also: Anonymisation, Personal data

A chart assigning, for each task or decision, who is Responsible, Accountable, Consulted and Informed.

See also: Governance

Including an item in the statement of financial position or of financial performance because it meets the definition of an element and the recognition criteria.

See also: Derecognition, Asset, Liability

Comparing two sets of records that should agree and investigating every difference until it is explained.

See also: Balance sheet reconciliation, Suspense account

A journal prepared on a fixed basis each period from a schedule — depreciation, prepayment releases, standard accruals — so that at close it needs review rather than construction.

See also: Journal entry, Month-end close

Deliberately attempting to make an AI system fail — to produce harmful, wrong or prohibited output — in order to find weaknesses before users do.

See also: Robustness

A statistical or machine learning method for predicting a continuous value, such as sales, from one or more input variables.

See also: Supervised learning, Predictive analytics

Under IAS 24, a person or entity closely connected to the reporting entity, such as key management personnel or entities under common control. Transactions with related parties must be disclosed.

See also: Financial statements

The inventory level at which a new order is placed: expected demand during the supply lead time plus safety stock.

See also: Safety stock, Lead time, Economic order quantity (EOQ)

An early approach to the market to learn about possible suppliers and solutions before a requirement is finalised.

See also: Invitation to tender (ITT)

A request to suppliers for prices against a clearly defined requirement, typically used where price is the main differentiator.

See also: Invitation to tender (ITT)

The level of a risk remaining after controls are applied. It is compared with risk appetite to decide whether further action is needed.

See also: Inherent risk, Risk appetite

A technique in which relevant documents are retrieved and supplied to a language model with the prompt, so that its answer is grounded in that material rather than in its training alone.

See also: Large language model (LLM), Hallucination

Attributed revenue divided by the advertising spend it is attributed to. It measures revenue, not profit: attributed sales cover their advertising only when ROAS reaches break-even ROAS, with both on one revenue basis, such as net of VAT and returns.

See also: Break-even ROAS, Contribution margin, Attribution model

Under IFRS 15, a five-step model: identify the contract, identify the performance obligations, determine the transaction price, allocate it to the obligations, and recognise revenue as each is satisfied.

See also: Performance obligation, Transaction price, Contract liability

Under IFRS 16, a lessee’s asset representing its right to use an underlying asset for the lease term. It is usually depreciated over the shorter of the lease term and the asset’s useful life.

See also: Lease liability

The amount and type of risk an organisation is willing to accept in pursuit of its objectives, approved by its board.

See also: Residual risk, Key risk indicator (KRI)

A grid that scores risks by likelihood and impact, often coloured as a heat map, to prioritise attention. Its scores are judgements and should be read as such.

See also: Risk register

A record of identified risks with their causes, owners, ratings, controls and actions, reviewed regularly.

See also: Risk matrix, Inherent risk, Residual risk

An AI system’s ability to keep performing correctly when inputs are noisy, unusual or deliberately manipulated.

See also: Red-teaming, Model drift

A forecast that is extended by one period each time a period closes, so that it always covers the same horizon — typically four to six quarters — rather than ending at the financial year-end.

See also: Forecast, Driver-based forecasting

A structured method for identifying the underlying cause of a problem, rather than its symptoms, so that corrective action prevents recurrence.

See also: Five whys, Corrective action

A projection of a full period’s result from recent performance, such as annualising the latest month. It assumes the recent period is representative.

See also: Forecast

Inventory held above expected demand to protect against variation in demand or supply lead time.

See also: Reorder point

Checking customers, suppliers and transactions against government sanctions lists before dealing with them.

See also: Anti-money laundering (AML)

Testing a model under coherent sets of assumptions — base, upside, downside — that change several inputs together to represent a possible state of the world.

See also: Sensitivity analysis, Base case

Improving a site’s visibility in unpaid search results through crawlable, indexable pages matched to the intent behind queries, original content and links earned rather than bought. Search engines’ published spam policies prohibit links bought to pass ranking credit.

See also: Search intent, Click-through rate (CTR)

The purpose behind a search query, usually classed as informational, navigational, commercial or transactional. Each query, in any language, dialect or script, is matched to the page type that serves its purpose, such as a guide for a question.

See also: Search engine optimisation (SEO), Negative keyword

Dividing responsibilities for authorising, recording and holding assets among different people, so that no one person can both commit and conceal an error or fraud.

See also: Internal control, Four-eyes principle

Changing one or two inputs at a time to see how much an output moves, in order to identify which assumptions matter most.

See also: Scenario analysis, Data table, Tornado chart

A defined, measurable standard of service a supplier or internal team commits to, such as response time or availability, with the consequences of missing it.

See also: Contract management

The use of AI tools by staff without the organisation’s knowledge or approval, often with data that should not leave it.

See also: Acceptable use policy, AI system inventory

A data-driven improvement method that reduces variation and defects, typically following the DMAIC cycle: define, measure, analyse, improve, control.

See also: Lean, Root cause analysis

The statement of what is required from a supplier. A technical specification describes the product or method; an output-based specification describes the result to be achieved.

See also: Invitation to tender (ITT)

Collecting, cleaning and classifying an organisation’s purchasing data to show what is bought, from whom and at what cost, as the basis for category strategies.

See also: Category management, Maverick spend

Structured Query Language, the standard language for querying and managing data in relational databases.

See also: Data warehouse

A person or group who affects, or is affected by, a decision, project or organisation.

See also: Change management, RACI matrix

A measure of how widely values are spread around their mean. Roughly two thirds of a normally distributed set lie within one standard deviation of the mean.

See also: Outlier

A written, step-by-step instruction for carrying out a routine activity consistently.

See also: Policy

An indication that an observed result would be unlikely if there were no real effect. It does not show that the effect is large or important.

See also: A/B test, Correlation and causation

A detailed ledger for a class of transactions — receivables, payables, fixed assets — whose total is posted to a control account in the general ledger.

See also: Control account, General ledger

Under IFRS 10, an entity controlled by another entity, the parent. Control rests on power over the investee, exposure to variable returns and the ability to use that power to affect them.

See also: Consolidation, Non-controlling interest

Machine learning from examples labelled with the correct answer, so that the model learns to predict the label for new cases.

See also: Classification, Regression, Training data

A structured approach to managing interactions with important suppliers to improve performance, reduce risk and develop value over time.

See also: Contract management, Supply positioning

Classifying purchases by their value to the organisation and the risk in their supply market — often shown as the Kraljic matrix — to choose a sourcing approach for each.

See also: Category management

A list of contacts never to be messaged or uploaded to ad platforms, such as those who unsubscribed, complained or hard-bounced, checked before every send and upload. Opted-out contacts are suppressed, not deleted, so a later import cannot restore them.

See also: Marketing consent, Deliverability

A temporary account holding entries that cannot yet be classified. A balance left in suspense at period-end is an unresolved error.

See also: Reconciliation

An ambition set to stretch performance, usually above what is expected. It should not be treated as a forecast.

See also: Budget, Forecast

The weighted criteria against which bids are scored, fixed and published before bids are opened so that the evaluation cannot be shaped to a preferred bidder.

See also: Invitation to tender (ITT), Total cost of ownership (TCO)

The value of a business’s cash flows beyond the explicit forecast period, estimated by the perpetuity growth method or an exit multiple. It often accounts for most of a DCF valuation.

See also: Discounted cash flow (DCF), Perpetuity growth method, Exit multiple

Data held back from training and used only to measure how well a finished model performs on data it has not seen.

See also: Training data, Overfitting

A governance model, published by the Institute of Internal Auditors, that distinguishes management’s ownership of risk (first line), specialist risk and compliance oversight (second line) and independent assurance by internal audit (third line).

See also: Governance, Internal audit

A financial model in which the income statement, balance sheet and cash flow statement are linked, so that each transaction flows through all three and the balance sheet balances by construction.

See also: Financial model, Balance check

A payables control that agrees the purchase order, the goods received record and the supplier invoice before an invoice is approved for payment.

See also: Accounts payable, Internal control

The rate at which a process produces finished output, set by its bottleneck.

See also: Bottleneck, Cycle time

The unit of text a language model processes — often a word or part of a word. Context limits and usage costs are measured in tokens.

See also: Context window

A bar chart ranking inputs by how much each moves an output across a stated range, with the most influential at the top.

See also: Sensitivity analysis

The full cost of acquiring, operating, maintaining and disposing of an asset or service over its life, not only its purchase price.

See also: Tender evaluation criteria

The data from which a machine learning model learns. Its quality, coverage and representativeness largely determine the model’s behaviour.

See also: Algorithmic bias, Test set

Under IFRS 15, the amount of consideration an entity expects to be entitled to in exchange for transferring promised goods or services, including estimates of variable consideration.

See also: Variable consideration, Revenue recognition

Adapting marketing copy into another language so that its intent, tone and effect survive, rewriting idiom, humour and cultural references rather than translating them word for word. Regulated wording, such as offer terms, prices and legal statements, is translated exactly.

See also: Claim substantiation

A list of every general ledger account and its balance at a date, whose debits and credits must be equal. It is the starting point for preparing financial statements.

See also: General ledger

The revenue and costs attributable to a single unit of the business — a customer, an order, a product — used to test whether growth adds value.

See also: Contribution margin

Machine learning that finds structure in unlabelled data, such as grouping similar customers into segments.

See also: Machine learning

The period over which an asset is expected to be available for use, or the number of units it is expected to produce. It determines the depreciation charge.

See also: Depreciation

A diagram of every step, delay and information flow needed to deliver a product or service, used to separate value-adding time from waste.

See also: Lean, Process mapping

Under IFRS 15, consideration that can change — discounts, rebates, bonuses, penalties. It is included in the transaction price only to the extent a significant reversal is highly unlikely.

See also: Transaction price

The comparison of actual results with budget or forecast, explaining each difference by its cause so that management can act on it.

See also: Budget, Flexed budget, Price-volume-mix analysis

In lean practice, any activity that consumes resources without adding value for the customer, such as waiting, rework, excess inventory and unnecessary movement.

See also: Lean

The average return required by all providers of capital, weighted by the proportions of debt and equity, with the cost of debt taken after tax. It is the discount rate for free cash flow to the firm.

See also: Cost of equity, Discount rate

Reporting suspected wrongdoing through a channel that protects the person reporting from retaliation.

See also: Compliance

Changing the sequence, roles and handoffs of a process — for example, to incorporate automation or AI assistance — rather than adding a new tool to an unchanged process.

See also: Process mapping, Change management

Current operating assets less current operating liabilities — principally inventory and receivables less payables. Increases absorb cash; decreases release it.

See also: Cash conversion cycle, Liquidity

Budgeting that requires every cost to be justified from zero for each new period, rather than adjusting the previous period’s figures.

See also: Budget