Candidate Data: The Fields, the Sources, and How Fast It Decays (2026)
Candidate data is the record a recruiter holds on a person, covering employment, tenure, skills, education and contact details.
Published
Written by
Ainesh S.
Reviewed by
Nithish A.
Read time
15
minutes

Candidate data is the record a recruiter holds on a person, covering employment, tenure, skills, education and contact details. The employer and title fields change each time the person changes jobs or is promoted, while the record stays as it was. In short: a candidate database is where records are stored, and candidate data is what each record says about the person.
Crustdata measured how often people in six US occupations changed employer or title, using employment histories in the Crustdata people database. Averaged across four start dates, 17.6% of people holding one of these jobs on a start date had changed employer or title within 12 months (2,619 person-dates from 781 people). A record copied on that date and never re-checked would no longer match them 12 months later.
Over 24 months, 29.3% of people holding one of these jobs changed employer or title. That figure averages the two start dates whose 24-month window had closed (1,312 person-dates from 704 people).
Knowing where each field comes from and how often it changes lets a recruiting team re-check the riskiest records first and keep the rest on a slower schedule.
At a glance
Fields: employment, tenure, skills, education and contact details.
Rates measured by Crustdata: employer and title changes among people holding one of six US occupations on a start date.
Changed within 3, 6 and 12 months: 5.3%, 9.9% and 17.6% (four start dates pooled).
Changed within 24 months: 29.3% (two start dates pooled, 704 people).
Highest risk: people in their first 2 years in a role, 30.8% of whom changed employer or title within 12 months.
Public check: 20.6% of US wage and salary workers had been with their employer for a year or less in January 2026 (US Bureau of Labor Statistics).
Main legal duty: keep candidate data accurate and only for as long as it is needed (European Commission).
What is candidate data?
Candidate data is the information a recruiter, sourcing team or recruiting software product holds about a potential hire. Recruiters use it to find people who fit a role and to decide whom to contact first.
Each record is a set of fields, and each field comes from a source and changes on its own trigger. One fictional record, for a made-up engineer called Jordan Lee, shows how the fields behave:
Field | Example value (fictional) | How it is usually collected | When it changes |
|---|---|---|---|
Current employer | A payments company in Austin | What the person posts, an application, the employer's own ATS | When the person changes jobs |
Job title | Senior software engineer | Same as the employer field | On a promotion or internal move |
Tenure | 1 year 8 months in the role | Worked out from the role's start date | Every day, if it is stored as a number |
Skills | Go, Kafka, payments APIs | Listed by the person, inferred from titles, or tested | When a skill is added or tested |
Education | BSc Computer Science, 2017 | Resume, application, public profile | Rarely, once a person is working |
Contact details | Work email, phone, Austin, Texas | Application, enrichment, the person | Work email with the employer, location with a home move |
Is candidate data the same as a candidate database?
Candidate data is the content of each record, and a candidate database is the system that stores the records. An applicant tracking system (ATS) or a recruiting CRM is a typical candidate database. The same candidate data can sit in several databases at once, and each copy goes out of date on its own unless something refreshes it.
What fields make up candidate data?
Candidate data has five groups of fields, and they change at very different speeds. The fields an enrichment API returns are covered in what data a candidate enrichment API should return. The five groups are:
Employment
Tenure
Skills
Education
Contact details
What does the employment field hold?
The employment field holds the person's current employer and job title, plus their past roles. It changes whenever the person moves to a new employer or takes a new title at the same one.
Example: Jordan's record lists a payments company. If Jordan moves in month 9, a search for "engineers at payments companies" keeps returning Jordan until the record is refreshed.
What does the tenure field hold?
Tenure is how long a person has been in a role, worked out from the start date. A stored number such as "2 years at company" is wrong the next day. For that reason, tenure is best computed from the start date each time it is used.
Example: Jordan started 1 year 8 months ago, so a value of "1 year" saved 8 months ago is now 8 months short.
What does the skills field hold?
The skills field holds what the person can do. A skill is listed by the person, inferred from titles and work history, or shown in a test. Skills carry no dates in most records, so nobody can say from the record alone when a skill was added.
Example: "Kafka" on Jordan's record could date from a role 5 years ago or from last month.
What does the education field hold?
The education field holds schools, degrees and graduation years. It changes rarely once a person is working.
Example: Jordan's 2017 degree will read the same in 2030.
What do the contact fields hold?
The contact fields hold email addresses, phone numbers and location. A work email address belongs to the employer, so it tends to stop working when the person leaves. Location changes when people move home, and the US Census Bureau puts that at 8.7% of Americans in 2022.
Example: If Jordan leaves the payments company, a message sent to Jordan's old work email 1 week later is likely to bounce.
How much should each field be trusted?
Employer and title fields carry the most known risk, because only they have a change rate measured by Crustdata. The suggested re-check for each field depends on how often it changes, using Crustdata's measured job-change rates and the US Census Bureau figure for location:
Field | Rate measured by Crustdata? | Change rate | Suggested re-check |
|---|---|---|---|
Current employer | Yes | 13.9% of people within 12 months, 24.2% within 24 months (six US occupations) | Monthly for records in active use |
Job title | Yes | 3.7% of people within 12 months, at the same employer (six US occupations) | With the employer field |
Tenure | Derived | Changes daily if stored | Recompute from the start date |
Skills | No | Not dated in most records | Before outreach on a skills match |
Education | No | Rarely changes after a person starts work | Rarely |
Work email | No | Tied to the employer field | Whenever the employer changes |
Personal phone | No | Changes on the person's own schedule | Before first contact |
Location | No, Census figure | 8.7% of Americans moved home in 2022 | Yearly, or before a location search |
Where does candidate data come from?
Candidate data comes from five kinds of source, and each one updates on a different trigger:
What candidates post about themselves
Resumes and job applications
An employer's own ATS or CRM (customer relationship management system)
Company websites
Job boards
Each source has its own answer, and a comparison of all five comes after them.
What do candidates post about themselves?
Candidates post their own details on public professional profiles, personal websites and code repositories. These sources often hold the most detail, and each person edits them when they choose.
Example: One public profile can list Jordan's 4 past roles, 12 skills and a 2017 degree in one place.
What do resumes and applications show?
A resume or application is accurate on the day it is sent and then stays fixed. It gives a full work history for that one moment and starts to age the same day.
Example: A resume Jordan sent in 2024 still lists the 2024 employer.
What does an employer's own ATS or CRM show?
An ATS or CRM shows what a recruiting team recorded. It changes only when a recruiter edits it or a scheduled refresh updates it. Records added years ago usually carry the employer and title from that time.
Example: A 2023 import of Jordan's record still shows Jordan's 2023 job.
What do company websites show?
Company websites show team pages, leadership lists and press releases. They change when the company updates the page, which may happen weeks after a hire or departure.
Example: A team page updated once a quarter can list Jordan for up to 3 months after Jordan leaves.
What do job boards show?
Job boards show which roles a company is hiring for. A posting is company data, but it signals that someone in that role may be leaving or that a team is growing.
Example: A posting for Jordan's exact role at Jordan's employer, put up 30 days ago, can mean Jordan's seat is opening.
How do the sources of candidate data compare?
The five sources differ in what they tell a recruiter and in what makes them change:
Source type | What it tells a recruiter | When it changes | Main limit |
|---|---|---|---|
What candidates post about themselves | Employer, title, skills, education | When the person edits it | Edited on the person's own schedule |
Resumes and applications | Full history at one moment | Never, after it is sent | Ages from day one |
An employer's own ATS or CRM | What the team recorded | When someone edits or refreshes it | Ages unless refreshed |
Company websites | Who works there, in some roles | When the company updates the page | Partial and slow |
Job boards | Open roles at a company | When postings go up or down | Describes companies only |
How fast does candidate data go out of date?
Of people in six US occupations on a start date, 17.6% changed employer or title within a year. Crustdata's measurement of employment histories in its people database shows the share growing with each interval:
Months after the start date | Changed employer | Changed title at the same employer | Changed employer or title |
|---|---|---|---|
1 | 1.7% | 0.3% | 2.0% |
3 | 4.3% | 1.0% | 5.3% |
6 | 7.8% | 2.1% | 9.9% |
12 | 13.9% | 3.7% | 17.6% |
24 | 24.2% | 5.2% | 29.3% |
Each figure is the share of people holding one of the six jobs on a start date who changed employer or title within the period. The columns are rounded separately, so the 24-month parts add to 29.4% while the total is 29.3%.
The rows in that table draw on two pools of start dates:
Rows | Start dates | Person-dates | People |
|---|---|---|---|
1 to 12 months | 1 January 2023, 1 July 2023, 1 January 2024, 1 July 2024 | 2,619 | 781 |
24 months | 1 January 2023, 1 July 2023 | 1,312 | 704, all among the 781 |
The share of people changing employer or title nearly doubles between 6 and 12 months, and employer changes make up most of it at every interval.
The job-change rates raise three further questions:
Which candidates change employer or title fastest?
Do these rates apply outside the US?
How does this compare with government job-change data?
Which candidates change employer or title fastest?
People who are new in a role change employer or title fastest. Splitting Crustdata's 12-month figure by whole months in the role on the start date gives:
Time in the role on the start date | Changed employer or title within 12 months |
|---|---|
Under 2 years | 30.8% |
2 to 5 years | 21.4% |
5 years or more | 3.5% |
A person under 2 years into a role was about nine times as likely to change employer or title within a year as someone past 5 years.
Do these rates apply outside the US?
Crustdata measured job-change rates on a sample of US people only. The closest UK figure uses a different measure. The Office for National Statistics found that an average of 9.1% of UK workers changed jobs each year from 2000 to 2018, including moves inside the same firm.
Teams with mainly non-US records should test a sample of their own before relying on the US figures.
How does this compare with government job-change data?
Government data shows job change on a similar scale to Crustdata's measurement. The US Bureau of Labor Statistics found that 20.6% of US wage and salary workers had a year or less with their current employer in January 2026.
Median tenure was 4.1 years for all wage and salary workers and 3.0 years for workers aged 25 to 34 (BLS Table 1).
Turnover data gives the monthly view. The BLS Job Openings and Labor Turnover Survey put the 2025 average monthly quits rate at 2.0% and the total separations rate at 3.3%.
Crustdata's 12-month employer figure for the six US occupations, 13.9%, sits below the BLS share of 20.6%, which also counts people starting or returning to work.
How did Crustdata measure this?
Crustdata measured how often people changed employer or title, using employment histories in the Crustdata people database. The method compared each person's job on a start date with their job at fixed points afterwards, in five parts:
Sample: 2,400 US profiles from the Crustdata people database, drawn at random starting points. The six occupations were software engineer, recruiter, account executive, registered nurse, accountant and marketing manager. The measurement used the People Search API and no live data.
Who counted: a person counted on a start date if they held a job in one of the six occupations that day, so one person can count on more than one start date.
Employer change: the job held on the start date had ended by the end of the window, and the person held no role at that company then.
Title change: a role with a different title at the same company started inside the window.
Start dates: the 12-month figures use the four start dates whose 12-month window closed at least 15 months before the measurement. The 24-month figures use the two of those whose 24-month window also closed by then.
Dates in employment histories are month-level, so a 3-month window can be off by up to a month.
How much of an old ATS is out of date?
If an ATS matches the six US occupations Crustdata measured, about 17.6% of year-old records are wrong. The figures in this section estimate an ATS from Crustdata's measured job-change rates and do not measure any real one. For a database of 10,000 candidate records, our arithmetic on the unrounded rates gives these estimated counts:
Time since the records were last checked | Estimated records with the wrong employer or title |
|---|---|
3 months | 531 |
6 months | 989 |
12 months | 1,760 |
24 months | 2,934 |
Old ATS data comes up often in conversations between recruiting teams and Crustdata. The co-founder of an ATS company described to our team what customers bring when they switch systems:
"If I'm migrating their system from a different ATS, they might have data that is like two, three years old or maybe five years old and we just need to refresh that profile for them."
How do you clean up an old ATS?
A practical way to clean an old ATS is to re-check records in order of risk. Subtract each role's start date from the date the record was last checked to get tenure at that time. Re-check people who were under 2 years into a role first, then 2 to 5 years, then the rest. Records with no start date go to the front of the queue.
What are the uses of candidate data?
Candidate data supports four common recruiting jobs:
Sourcing: finding people who match a role by employer, title, skills and location.
Matching: scoring candidates already in a database against a new opening.
Re-engagement: contacting past applicants and runners-up when a similar role opens.
Market mapping: counting who works where, to plan a search or advise a hiring manager.
However, every one of these uses depends on the employer and title fields, which change whenever a person changes jobs.
What are the risks and limitations of candidate data?
Candidate data carries two main risks in day-to-day use:
Out-of-date records lead to the wrong outreach
Inferred fields can describe the person wrongly
Why do out-of-date records lead to the wrong outreach?
Out-of-date records send messages to the wrong place or with the wrong context. A note that congratulates someone on a job they left, or a work email that bounces, costs the recruiter credibility with that person.
Why can inferred fields describe the person wrongly?
Inferred fields can describe the person wrongly because they are guesses from other fields. A title such as "senior software engineer" says nothing certain about which languages the person uses, so a skill inferred from it can be wrong.
Is it legal to store and use candidate data?
Holding candidate data comes with a duty to keep it accurate and to keep it no longer than needed. Under the UK GDPR a recruiter must also have a valid lawful basis to handle it. This is a summary, not legal advice.
The European Commission lists seven data protection principles, and the rules that matter most for candidate data fall into three areas:
Accuracy and retention under the GDPR
Notice under the UK GDPR
California's rules for job applicants
What does the GDPR require for candidate data?
The General Data Protection Regulation (GDPR) requires candidate data to be lawful, accurate and kept no longer than needed.
The European Commission states that "personal data must be processed in a lawful and transparent manner" and that "personal data must be stored for the shortest time possible." It adds that an organisation "should establish time limits to erase or review the data stored."
The Commission's accuracy example is about recruiting. It says that "the recruitment office must ensure that the data is kept up-to-date." The same page says that inaccurate personal data must be rectified or deleted without delay. An out-of-date employer or title on a candidate record is therefore an accuracy question as well as a retention one.
What does the UK GDPR require when candidate data comes from another source?
The UK GDPR requires a recruiter to tell people about data collected from another source. The ICO (UK Information Commissioner's Office) sets the deadline as "within a reasonable period of obtaining the personal data and no later than one month." If the first message to the person comes sooner, the notice goes with it.
What does California require for job applicants?
California's privacy law covers job applicants who live in the state. The California Privacy Protection Agency states that the law "includes California residents that are employees or job applicants," because the exemption for employment-related data expired on 31 December 2022.
How do you keep candidate data up to date?
Candidate data stays up to date when records are re-checked as often as their fields change. A scheduled refresh is shown step by step in how to build an auto-updating candidate database. Three questions cover how to do that:
How often should candidate data be refreshed?
Which practices keep candidate data accurate?
What does a last-updated filter show about a candidate dataset?
How often should candidate data be refreshed?
A monthly refresh suits most candidate records in the six US occupations Crustdata measured. In those occupations, the share of people who changed employer or title grows with the gap between checks:
Re-check every | People who changed employer or title before the next check |
|---|---|
1 month | 2.0% |
3 months | 5.3% |
12 months | 17.6% |
Recruiting teams that talk to Crustdata most often expect monthly updates, and some want changes the week they happen.
Which practices keep candidate data accurate?
Four practices keep a candidate database accurate:
Store a date on every record: a last-checked date on each record shows how long ago it was confirmed. Teams that talk to Crustdata often ask for one.
Recompute tenure: work out tenure from the start date each time it is used.
Match the schedule to the use: check shortlists in an active search weekly and the rest of the database monthly or quarterly.
Watch for changes: set alerts on the people a team cares about, so a job move triggers a re-check of that one record. Change-alert services do this, including the Crustdata Person Entity Watcher, which sends a notification when a watched person has a new job or a title move.
What does a last-updated filter show about a candidate dataset?
A last-updated filter shows how recently each record was checked, which one buyer used to test a bought dataset. A head of growth at an AI recruiting software company filtered a candidate dataset bought from another vendor to profiles updated in the last month, and described the result to our team:
"I think when I put it to last updated a month ago, was like I lost like 90% of my data right there."
Candidate data FAQ
How do you define outdated candidate data?
Outdated candidate data is any field that no longer matches the person, most often the employer or title. A record checked recently can still be outdated if the person moved after the check. For that reason, the date of the last check and the date of the person's last job change are two separate tests.
Does a job change make personal contact details out of date?
A job change does not end a personal email address or phone number, because neither belongs to the employer. Those fields go out of date on their own schedule.
How long can you keep candidate data?
Candidate data has no single legal retention period. The ICO says "the UK GDPR does not set specific time limits for different types of data."
Do you have to delete candidate data when someone asks?
A deletion request must be answered, but deletion is not automatic in every case. Under the UK GDPR, the ICO notes that "the right is not absolute and only applies in certain circumstances." It expects a response "without undue delay and at the latest within one month."
Can you contact candidates found in public data under the GDPR?
A recruiter who contacts candidates found in public data must tell them what data is held. The ICO's guidance on the right to be informed sets out that duty. Whether the contact itself is lawful depends on the lawful basis the recruiter relies on, which a privacy lawyer should confirm.
Can old ATS records be refreshed in bulk?
Old ATS records can be refreshed in bulk by matching each one to an up-to-date profile. A match on name alone is weak, so the match is safer when it also uses a past employer, an email address or a profile URL from the record. Records that match nothing are worth a manual check before anyone decides to keep or delete them.
