Every field has its favourite jargon and buzzwords – and People Analytics is no exception.
From “headcount envelopes” to “skills adjacency”, the language can feel like a puzzle if you’re not steeped in it every day. To make life easier (and meetings clearer), we’ve put together a plain-English guide.
Think of it as your decoder ring for the conference: no mystery acronyms, no baffling finance-speak, just simple explanations you can use right away.
A/B test | Compare two options by splitting users randomly. |
A/B/C option framing | Presenting three realistic choices (A, B, C), each with capacity, cost range and time‑to‑coverage, so Finance can compare and choose. |
Absence | Time not worked due to sickness, leave or other reasons. |
Accruals | Costs recorded when incurred, not when paid. |
Adoption | The extent to which teams are actually using the new process, model or plan. |
Adoption telemetry | Data showing whether teams are actually using the new process or tool. |
Adverse impact | Disproportionate negative effects on protected groups. |
AI assurance | Evidence that an AI system is safe, fair and robust. |
Algorithmic transparency | Explain how a model affects decisions. |
API | A standard way for systems to talk to each other. |
Artefacts | The simple documents you publish and maintain to run the process. |
Assumptions register | A list of the key assumptions behind the plan, each with its value/range, source, owner and date. |
Attrition | People leaving the organisation (voluntary or involuntary). |
Attrition bias | People drop out over time and skew results. |
Automation | Using technology to do work that people did before; sometimes an alternative to adding headcount. |
Backlog | The ordered list of work to do. |
Baseline | Your starting point for comparison. |
Batch vs real‑time | Move data in chunks vs continuously. |
Behavioural barrier | A habit or incentive that blocks adoption. |
Benchmarking | Comparing your numbers to a peer group. |
Blue‑green deployment | Run old and new versions side by side, then switch. |
Build / buy / blend / borrow / automate | Five ways to get work done: train your own; hire; mix internal and external; use contractors/partners temporarily; use technology instead of adding people. |
Build plan | Concrete steps and timelines to create skills or capacity internally. |
Cadence | The regular rhythm for reviews. Example: a monthly check‑in. |
Calibration | How well predicted probabilities match real outcomes. |
Canary release | Roll out to a small group first to check for issues. |
Capacity | The amount of work a team can deliver in a period (hours, cases, features, etc.). |
Capacity maths | Simple calculations that show how many hours or units of work your team can deliver. |
CDC (change data capture) | Streaming only what changed since last time. |
Census survey | Full workforce survey. |
CFO | The most senior finance leader in a company. Often signs off major spending. |
CFO note (monthly) | A short update to Finance: what changed (movement), how the method is holding up, adoption, confidence and the next decision. |
CFO‑ready artefacts | Documents laid out to answer Finance’s core questions fast. |
Champions | Enthusiastic early users who help others adopt. |
Change network | Named people in units who help land the change. |
CHRO | Chief HR Officer, the most senior HR leader. |
Class imbalance | When one outcome is rare (e.g., 2% attrition). |
Closing panel | The final group discussion with several speakers and audience Q&A. |
COE (centre of excellence) | A small expert team that sets standards and supports others. |
Cohort chart | Tracks groups over time from a common start point. |
Collective bargaining | Negotiation between employer and union on pay and conditions. |
Communications plan | Who needs to hear what, when and how. |
Compa‑ratio | Pay vs the midpoint of the pay band. |
Confidence | How sure you are about a number or forecast; often shown by a range rather than a single point. |
Confidence bands | Upper and lower bounds around a forecast to show uncertainty. |
Confidence interval | A range that likely contains the true value. |
Confounder | A hidden factor that distorts results. |
Contingent workforce | Contractors, temps, and agency staff. |
Cost range | A span instead of a single number (e.g., £450k-£520k) to reflect uncertainty. |
Coverage | Whether enough people are on shift to meet demand. |
Coverage & capacity test | Show how fast you can cover the work and how much work the team can deliver. |
Critical path | The steps that control the earliest finish date. |
Critical role | A role that strongly affects value or risk. |
Cross‑sectional | One‑off snapshot at a point in time. |
Cross‑validation | Test a model by training and testing on different splits. |
Data lakehouse | A data store that supports both raw files and analytics tables. |
Data lineage | Where data came from and how it was changed. |
Data mesh | Domain teams own their data as products with standards. |
Data quality dimensions | Accuracy, completeness, timeliness, consistency, uniqueness. |
Data retention | How long you keep data and when you delete it. |
Decision rights | Which roles are allowed to approve or reject a choice. |
Decision sheet | One page showing options A/B/C with capacity, cost range, time‑to‑coverage, risks and a recommendation. |
Decision test | Be explicit about the choice you want signed now (e.g., hire vs redeploy). |
Deduplication | Merging duplicate records into one. |
Definition of done | The checklist to call a story complete. |
Demand model | Estimate of future work to be done (e.g., cases per week next quarter). |
Difference‑in‑differences | Compare before/after across treated and control groups. |
Differential privacy | Adding noise so individuals can’t be identified. |
Domain data product | A well‑documented, trustworthy dataset owned by a domain team. |
DPIA | Data protection impact assessment; checks privacy risks before launch. |
DRI (directly responsible individual) | The named person who owns an outcome. |
Drift (data/concept) | When data or relationships change over time. |
Driver | An input that moves a result (e.g., hiring rate, attrition). |
Driver analysis | Which factors most move an outcome. |
Early attrition | Leaving within the first months (e.g., 90 days). |
Effect size | How big the difference or relationship is, not just whether it exists. |
Elasticity | The responsiveness of one variable to changes in another. |
Embeddings | Numeric vectors that represent meaning of text for search and clustering. |
Engagement | How committed and energetic people feel at work. |
eNPS | Employee Net Promoter Score; likelihood to recommend the employer. |
Entity extraction | Pulling out names, skills, locations from text. |
Envelope (finance/headcount) | The cap Finance sets on spend or roles for a period. |
Epic | A large piece of work broken into smaller stories. |
Erlang C | A formula to staff call centres based on wait targets. |
ETL/ELT | Move and transform data: Extract‑Transform‑Load or Extract‑Load‑Transform. |
Event stream | Real‑time feed of changes (e.g., new hires, status updates). |
Executive one‑pager | A single page with the headline, the number and the ask. |
Explainability vs interpretability | Explaining a specific prediction vs understanding the overall model. |
External labour supply | How many suitable people are available in the job market. |
External market checks | Using outside data to test realism (e.g., supply of engineers, wage trends). |
F1 score | The balance between precision and recall. |
Feature | An input variable used by a model. |
Feature store | A shared place to keep cleaned model inputs. |
Few‑shot/zero‑shot | Getting an AI to perform a task with few or no examples. |
Finance | The team that manages budgets, forecasts and company money. |
Finance‑grade lenses | Ways of evaluating a plan that meet Finance’s standards for rigour and auditability. |
Fine‑tuning | Training an AI on your examples to shift its behaviour. |
Forecasting | Projecting future numbers using current data and assumptions. |
Four‑fifths rule | A quick screen for adverse impact (rate should be at least 80% of the highest group). |
FP&A | Financial Planning & Analysis; the finance team that builds budgets and forecasts and partners with the business. |
FTE (Full‑Time Equivalent) | A way to add up different work patterns. 1.0 FTE = one full‑time person; 0.5 FTE = half‑time. |
Fully loaded cost | Salary plus benefits, taxes, tools and overhead. |
Funded decisions / funded choices | Options that have been approved and given budget to execute. |
Funnel chart | Shows drop‑off across stages in a process. |
Gen‑AI | Generative AI. Software that can read or generate text, code or images. |
Gen‑AI (generative AI) | AI that can generate or summarise text, code or images. |
Golden source | The single, agreed best source for a data element. |
Governance | Who decides what, and how those decisions are checked. |
Grounding | Forcing AI to use your trusted data as source material. |
Hallucination | When an AI confidently makes something up. |
Headcount envelope | The cap set by Finance for total people cost or number of roles. |
Heatmap | A coloured grid to show intensity. |
Hiring funnel | Stages from applicants to hires. |
Holdout set | A final, untouched sample to test real performance. |
HR-Finance operating handshake | A practical agreement: Finance sets the headcount envelope; HR/SWP bring options that fit the envelope and lift capability and productivity. |
Hub | A hiring/operating location you cluster roles in (e.g., Dublin vs Kraków). |
Hub‑and‑spoke | One expert hub supporting several local teams. |
Human‑in‑the‑loop | Keep people in key steps to review and override. |
ICE plots | Show feature effects for individual cases. |
Identity resolution | Matching records that belong to the same person. |
Impact (track) | Label for sessions that lead with business outcomes and value. |
Inclusion index | A composite score of belonging and fairness. |
Instruction tuning | Training an AI to follow instructions better. |
Internal mobility | Moves within the company, lateral or upward. |
Internal pipeline readiness | How many near‑ready candidates you already have inside the company and how fast they can be moved. |
Interrogate (numbers) | Examine and question the numbers to test reliability. |
JD parsing | Using software to read job descriptions and extract skills/requirements. |
JDs (job descriptions) | The written description of a role’s responsibilities and requirements. |
Job architecture | The structured map of roles, levels and bands. |
Job family | A set of related roles with similar skills and progression. |
K‑anonymity | Ensure each person looks like at least k‑1 others in a dataset. |
KPI | A key performance indicator; a metric that matters most. |
Label/target | The value a model tries to predict. |
Lagging indicator | A metric that moves after outcomes (e.g., retention after onboarding). |
Lawful basis | The legal reason to process data (e.g., consent, contract). |
Layering | Number of management levels from top to front line. |
Lead time | Time from request to delivery (e.g., 10 weeks to hire a data engineer). |
Leading indicator | A metric that moves before outcomes (e.g., pipeline health before hires). |
Leakage | Using information in training that wouldn’t be available in reality. |
Legitimate interest | A lawful basis that requires safeguards and balancing tests. |
Likert scale | 1 to 5 agreement scale. |
Longitudinal | Tracking the same people over time. |
MAE/MAPE/RMSE | Common error measures for forecasts and regressions. |
Market test | Prove the plan is realistic using labour‑market data and internal pipeline. |
Market tightness | When supply of a skill is scarce relative to demand. |
Market validation | Checking your plan against external data to confirm it’s realistic. |
Master data | Core entities like people, roles, locations with unique IDs. |
Method (for costs) | How you worked the number out: assumptions, rates, multipliers and what’s included/excluded. |
Mini‑case | A short, anonymised example that demonstrates the method. |
MLOps | The process to deploy and maintain models reliably. |
Model registry | A catalogue of models with versions and approvals. |
Model risk | The chance a model is wrong or misused. |
Money test | Show costs as a range, how you calculated them, and where the money will come from. |
Monitoring | Watching models for drift, errors and bias after launch. |
Movement | What has changed since last month (e.g., roles filled, cost variance, skills added). |
Multipliers | Factors applied to raw numbers to reflect reality (e.g., benefits on salary, utilisation). |
Narrative spine | The simple story arc of problem, options, choice, result. |
Net capacity | Capacity left after subtracting losses like attrition, holidays, training and ramp‑up. |
Next bet | The next decision or investment based on current progress. |
Non‑response bias | Missing views from those who didn’t respond. |
Nudge | A small design choice that makes the right action easier. |
Offer acceptance rate | Offers accepted divided by offers made. |
OKR | Objectives and key results; goals plus measurable results. |
One‑pager | A single‑page summary of the decision or plan. |
On‑ramp | An easy starting point that leads into a bigger topic or debate. |
Ontology/taxonomy mapping | Align different lists to a common standard. |
Operating cadence | The regular rhythm of meetings and reviews (e.g., monthly). |
Opex vs capex | Operating expenses vs capital investments. |
Orchestration | Scheduling and monitoring data jobs (e.g., nightly pipelines). |
Outcomes vs dashboards | Outcomes are business results (e.g., revenue saved). Dashboards are reports; they don’t prove value on their own. |
Overfitting | A model learns noise and fails on new data. |
Overtime | Hours beyond normal schedule, often at a higher rate. |
Ownership | The person accountable for a decision or deliverable. |
Ownership & cadence test | Name who is accountable and how often progress will be reviewed. |
Pack | A small bundle of handouts or digital files for delegates. |
Panel | A fixed group surveyed repeatedly. |
Partial dependence (PDP) | Shows average effect of a feature on predictions. |
Partner/contract | Use an external firm or contractor for a time‑boxed period. |
Pattern, not brand | Explaining a repeatable method without naming a specific company or product. |
Pattern‑based | Reusable approaches you can apply in other teams or companies. |
Pay equity | Paying fairly across comparable roles and levels. |
Pay progression | How pay moves over time within a band. |
PII | Personal data that can identify someone. |
Plenary | A session for the whole audience, not a breakout. |
Post‑mortem | Honest review after delivery to learn and improve. |
Power (sample size) | Chance of detecting a real effect given the sample. |
Precision | Of the predicted positives, how many were correct. |
Prediction interval | A range that likely contains the next observed value. |
Pre‑mortem | Imagine failure in advance to find and fix risks. |
Productivity | Output per unit of input (e.g., cases per agent per day). |
Proficiency level | Depth of skill (e.g., beginner, working, expert). |
Prompt | The instruction you give an AI model. |
Propensity score matching | Pair similar subjects to reduce bias in comparisons. |
Pseudonymisation | Replacing identifiers with codes while keeping links possible. |
Pulse survey | Short, frequent survey. |
p‑value | A measure used in statistical tests; lower means stronger evidence against chance. |
Quality‑of‑hire | How well new hires perform versus expectations (e.g., ramp time, performance, retention). |
Quarter | A three‑month business period (Q1, Q2, etc.). |
Quasi‑experiment | An experiment‑like test without full randomisation. |
Queueing | The maths of waiting lines and service times. |
RACI | Role map: Responsible, Accountable, Consulted, Informed. |
RAG (retrieval‑augmented generation) | Fetch documents first, then have AI answer using them. |
RAID log | Risks, Assumptions, Issues, Dependencies in one place. |
Ramp time / Ramp-up | Time for a new hire to hit expected output. |
RCT | Randomised controlled trial; the gold standard for experiments. |
Reallocation | Moving budget from one area to another rather than asking for new money. |
Recall | Of the actual positives, how many were found. |
Redacted | Sensitive details removed or masked. |
Redeployment | Moving current employees to different roles or teams instead of hiring new people. |
Red‑team | Try to break a system to find risks before launch. |
Reference data | Allowed lists like country codes or pay grades. |
Reforecast | Update the forecast mid‑period based on new facts. |
Regression to the mean | Extreme values tend to move closer to average next time. |
Reskill | Train someone to move into a different role. |
Response bias | Answers skewed by who chose to respond. |
Right to work | Legal permission to work in a country. |
ROC‑AUC | A score of classification performance across thresholds. |
Role archetype | A standardised role pattern used across teams. |
Roster optimisation | Arrange shifts to meet demand at least cost. |
R‑squared | How much of the variation a model explains. |
Run rate | Current monthly rate projected over a year. |
Safety filter | Blocks harmful or sensitive outputs. |
Sankey diagram | Shows flows between categories with bands. |
Scarcity index | A signal of how hard it is to find a skill in the market. |
Scenario | A plausible version of the future used to test a plan. |
Scenario planning | Preparing multiple plausible futures (A/B/C) with triggers for action. |
Scenario range | The low/medium/high outcomes considered. |
Schedule adherence | How well people stick to planned shifts. |
Scope (for costs) | What is in the estimate and what is not (e.g., includes salaries and tools, excludes relocation). |
Seat | A funded role or position. |
Seed the panel | Tee up targeted questions to spark the following discussion. |
Sensitivity analysis | How much a result changes if an input shifts. |
Sensitivity table | A small grid showing how outputs change with inputs. |
SHAP values | A way to explain model predictions by feature contributions. |
Shift bid | Workers choose preferred shifts, usually by seniority or points. |
Shrinkage | Time lost to meetings, breaks, training and absence. |
Sign‑off | Formal approval by the accountable person. |
Simpson’s paradox | A trend that reverses when you combine groups. |
Skill cluster | A group of related skills that travel together in roles. |
Skill inference (with Gen‑AI) | Using Gen‑AI to read job descriptions and extract the underlying skills required. |
Skill taxonomy | A structured list of skills that lets you group, compare and roll them up. |
Skills adjacency | Skills close enough that people can learn them quickly. |
Skills forecasting | Estimating which skills you will need, in what quantity, and when. |
Skills ontology/graph | A map of how skills relate to each other. |
Skills taxonomy | A structured list of skills with clear names and groups. |
SLA/SLO | Promises on availability or freshness of a service or dataset. |
Slide‑light | Very few slides; talk‑first. |
Slopegraph | Shows change between two points across categories. |
Small multiples | Many small charts with the same axes to compare groups. |
Source of funds | Where the money will come from (reallocated from another budget or new spending). |
Span of control | Number of direct reports per manager. |
Sparkline | A tiny line chart without axes for quick trend scans. |
Spine chart | Shows spread around a median across categories. |
Sprint | A short, fixed work cycle (e.g., two weeks). |
Story | A small, testable unit of work. |
Supply model | Forecast of people available, including hiring, exits and internal moves. |
SWP (Strategic Workforce Planning) | Planning the people and skills needed to deliver the business strategy. |
Talent density | Proportion of high performers in a team. |
Telemetry | Automated signals from systems that show usage or progress. |
Template | A pre‑formatted document you can reuse quickly. |
Throughput | The amount of work finished per period (cases per week, features per sprint). |
Time‑to‑coverage | How long until the work is adequately staffed. Example: ‘6 weeks to reach 90% coverage’. |
Time‑to‑fill | Days between job posting and offer acceptance. |
Time‑to‑start | Days between offer acceptance and start date. |
Tokens | Small chunks of text used for AI pricing and limits. |
Trigger | A condition that tells you to switch plan (e.g., vacancy rate passes 8%). |
TUPE | UK rules on staff transfers when services move supplier. |
Unit economics | Profit or cost per unit (e.g., per hire, per ticket). |
Uplift modelling | Predict who is more likely to change behaviour if treated. |
Utilisation | Share of time spent on productive work (e.g., 75%). |
Vacancy cost | Loss from an empty role (missed output or extra overtime). |
Vacancy rate | Open roles as a percentage of total roles. |
Variance | The gap between actual and plan. |
Variance to budget | Over or under the plan. |
Vector database | A store that finds nearest vectors (similar meaning text) fast. |
Vendor‑neutral | Not promoting a specific product or supplier; methods work with any toolset. |
Vignette | A brief, illustrative example or story. |
Wage trends / wage signals | Whether pay for a role is rising or falling in the market. |
Waterfall chart | Explains a total by showing adds and subtracts. |
Workload | Amount of work expected in a period. |
Works councils | Employee bodies with consultation rights in some countries. |
Zero‑based budgeting | Build the budget from zero, not last year’s spend. |