Course › Glossary

Glossary

52 terms  ·  every one written in plain language

Every term this course uses, explained without using other jargon to do it. Each entry links to the lesson that covers it properly.

Applicant tracking system (ATS) · Recruitment CRM · Resume parsing · OCR · Structured data · Hard filter · Weighted signal · Screening score · Banding · Over-filtering · Disagreement check · False negative · Keyword matching · Semantic matching · Match profile · Silver medallist · Intake meeting · Scorecard · Rating anchor · Must-have · Sourcing · Boolean search · X-Ray search · Talent map · Scraping · Sequence · Structured interview · Async interview · Debrief · Work sample · Stage · Status · Funnel · Conversion rate · Time to hire · Time to fill · Quality of hire · Source of hire · Workflow · AI agent · Prompt injection · Human-in-the-loop · Hallucination · Bias audit · Adverse impact · Proxy · EU AI Act · Local Law 144 · GDPR · DPDP Act · Legitimate interest · Retention period

The software

Applicant tracking system (ATS)

The database that holds your candidates, jobs, and where each person is in the process. Everything else in a hiring setup reads from it or writes to it. If two people asking the same question about a candidate get different answers, your ATS is not doing its job.

Covered in Module 1, Lesson 2

Recruitment CRM

Software for keeping in touch with people before a job exists. An ATS handles people applying now; a CRM handles people you might want later. CRM stands for customer relationship management.

Covered in Module 8, Lesson 3

Resume parsing

Turning a CV document into data fields — name, employers, job titles, dates, skills. It runs in seven steps and each one can go wrong. Every screening score is arithmetic done on its output.

Covered in Module 4, Lesson 1

OCR

Optical character recognition: reading letters out of a picture. Needed when a CV is a scan or a photo rather than a text document. It makes small errors, often in names and email addresses.

Covered in Module 4, Lesson 1

Structured data

A hidden block of information on a web page that tells search engines what the page is about. For job pages it is called JobPosting, and getting it right is what puts your jobs into job search results.

Covered in Module 8, Lesson 4

Screening

Hard filter

A rule that removes people from the pile. Should only be used for genuine, checkable, non-negotiable requirements — the right to work somewhere, a licence the law requires. Three at most.

Covered in Module 4, Lesson 3

Weighted signal

A rule that moves people up or down the list rather than removing them. Everything that is not a genuine must-have belongs here.

Covered in Module 4, Lesson 3

Screening score

A measure of how well a document matched the words you wrote. It is not a prediction that someone will be good at the job, and it cannot be compared between different jobs.

Covered in Module 4, Lesson 5

Banding

Grouping candidates into three or four groups — strong, worth reading, weak — instead of ranking them 1 to 50. More honest, because the difference between sixth and seventh is usually noise.

Covered in Module 5, Lesson 4

Over-filtering

Rejecting good candidates by setting rules too strictly. The most dangerous failure in AI screening, because a rejection produces no feedback — you never find out it happened.

Covered in Module 1, Lesson 3

Disagreement check

After every screening run, opening the three highest-scoring people who were rejected and the three lowest-scoring who got through, and deciding whether you agree. The step that separates screening from handing the job over.

Covered in Module 4, Lesson 4

False negative

A good candidate wrongly rejected. Costs you a hire and generates no evidence that it happened, which is why screening rules drift towards being too strict.

Covered in Module 1, Lesson 3

Matching and ranking

Keyword matching

Checking whether a word appears in a document. Fast and completely predictable. Cannot tell 'I use Python' from 'I have never used Python'.

Covered in Module 5, Lesson 1

Semantic matching

Matching on meaning rather than exact words, so 'built ETL pipelines' matches 'developed data ingestion workflows'. Its weakness is matching the topic rather than the requirement, and rewarding longer documents.

Covered in Module 5, Lesson 1

Match profile

The description the software compares candidates against. Build it from your scorecard, not from your job advert and not from your best employee's CV.

Covered in Module 5, Lesson 2

Silver medallist

Someone who nearly got hired for an earlier job. Already assessed, already interested, and will almost always reply. The cheapest source of candidates most companies ignore.

Covered in Module 5, Lesson 3

Defining the job

Intake meeting

The conversation with the hiring manager at the start of a search. Thirty minutes that decides the whole hire. Twelve questions, and the most useful one is about the last person who did not work out.

Covered in Module 2, Lesson 1

Scorecard

A written plan for what evidence you will collect about each skill, from which stage, and what counts as good enough. Not a list of skills.

Covered in Module 2, Lesson 4

Rating anchor

A written description of what each score means for a particular skill — what a 1, 2, 3 and 4 actually look like. Written before the first interview, it does more for consistency than any training.

Covered in Module 2, Lesson 4

Must-have

A requirement you would genuinely reject an outstanding candidate for missing. Most requirement lists contain one or two real must-haves and several habits.

Covered in Module 2, Lesson 5

Finding candidates

Sourcing

Finding candidates who have not applied to you. As opposed to inbound, which is people who came to you.

Covered in Module 3, Lesson 1

Talent map

A document showing where people who can do this job work, roughly how many exist, and what they cost. Built before searching, it turns an argument about difficulty into a market report.

Covered in Module 3, Lesson 5

Scraping

Automatically collecting data from websites. Most large professional networks ban it and enforce that by restricting accounts. The activity runs under your login, so the consequences land on you.

Covered in Module 3, Lesson 7

Outreach and interviews

Sequence

A planned series of messages to one candidate, usually five over about a month. Every message must add something the last one did not.

Covered in Module 6, Lesson 3

Structured interview

Everyone gets the same main questions, scores are written against agreed descriptions, and each interviewer submits scores before any discussion. Predicts job performance much better than an unstructured chat.

Covered in Module 7, Lesson 1

Async interview

A recorded interview the candidate completes on their own, watched later. Async just means not live. Useful for time zones and flexibility; costly in drop-out and exclusion.

Covered in Module 7, Lesson 3

Debrief

The meeting after interviews where scores are compared. Goes wrong when people score during the meeting instead of before it.

Covered in Module 7, Lesson 6

Work sample

A small version of the real task, done as part of the process. Predicts performance well and costs the candidate time, so keep it under an hour.

Covered in Module 7, Lesson 2

Process and measurement

Stage

Where a candidate is in the process — applied, screened, interview, offer. Stages only move forward.

Covered in Module 8, Lesson 1

Status

What is currently happening to a candidate — active, on hold, waiting on the candidate. A status is not progress. Mixing stages and statuses ruins every report you run.

Covered in Module 8, Lesson 1

Funnel

The shape of your process: how many people move from each stage to the next. Every hiring report is a funnel underneath.

Covered in Module 10, Lesson 1

Conversion rate

The share of people who move from one stage to the next. More useful than any count of activity.

Covered in Module 10, Lesson 1

Time to hire

Days from a candidate entering the process to accepting. Measures how fast your process moves a person.

Covered in Module 10, Lesson 2

Time to fill

Days from the job being approved to someone accepting. Measures how long the business waits, including the time before anyone could start searching.

Covered in Module 10, Lesson 2

Quality of hire

How well a hire works out. Cannot be measured cleanly — every available measure is a stand-in with a known flaw. Use two or three and name their limits.

Covered in Module 10, Lesson 2

Source of hire

Where a hire originally came from. Frequently wrong, because where someone applied is not where they first heard about you.

Covered in Module 10, Lesson 3

Automation

Workflow

A fixed sequence of automated steps that you set up in advance. The right level of automation for most hiring tasks, because you already know the steps.

Covered in Module 9, Lesson 1

AI agent

Software that chooses its own steps towards a goal rather than following a fixed sequence. More capable and much harder to check. Earns its complexity on open-ended research, rarely on hiring.

Covered in Module 9, Lesson 1

Prompt injection

Hidden instructions written into a document — such as a line in a CV saying to mark the applicant as highly qualified — aimed at software that reads it. Any automation reading candidate documents must treat their text as information, never as commands.

Covered in Module 9, Lesson 2

Human-in-the-loop

A person checking before something happens. Only real if the check asks a question that cannot be answered without reading. An approve button gets clicked.

Covered in Module 9, Lesson 3

Hallucination

When AI produces fluent, confident information that is not true — for example, adding a responsibility to a CV summary that the candidate never mentioned. The invented parts usually read better than the real ones.

Covered in Module 1, Lesson 3

Fairness and the law

Bias audit

Arithmetic on your own hiring data: what share of each group got through each stage, and how those shares compare. Required annually in some places, worth doing everywhere.

Covered in Module 11, Lesson 2

Adverse impact

When a rule affects one group much more than another, regardless of intent. A common screening rule of thumb treats a pass rate below four-fifths of the best group's as worth investigating.

Covered in Module 11, Lesson 2

Proxy

Something that stands in for a characteristic without naming it — postcode, university, gap length, CV template. Software does not need to know someone's background to act on it.

Covered in Module 1, Lesson 3

EU AI Act

European law classifying recruitment systems as high risk. The duties that cover recruitment now start on 2 December 2027, though the ban on inferring emotions at work has applied since February 2025.

Covered in Module 11, Lesson 1

Local Law 144

New York City's rule on automated hiring tools, in force since July 2023. Requires an independent bias audit every year, published results, and notice to candidates at least ten working days before use.

Covered in Module 11, Lesson 1

GDPR

European data protection law. Gives candidates rights to see, correct and delete their data, requires a reason to hold it and a date to delete it, and restricts decisions made purely by software.

Covered in Module 11, Lesson 4

DPDP Act

India's Digital Personal Data Protection Act. Became live when its detailed rules were published in November 2025, with duties phasing in to about mid-May 2027.

Covered in Module 11, Lesson 1

Legitimate interest

The usual legal reason for holding details of people you found rather than people who applied. It requires actually weighing your interest against theirs, and being able to show you did.

Covered in Module 3, Lesson 7

Retention period

How long you keep candidate data before deleting it. 'Indefinitely' is not a retention period, and a policy nobody applies automatically is worse than none.

Covered in Module 11, Lesson 4