Data quality / FIELD GUIDE

What is Data freshness?

Data freshness describes how recently information was observed or verified relative to how quickly it can change and how current it must be for the intended decision.

Key takeaways

  • Freshness describes how current information is for the decision being made.
  • Observed time, ingested time and last checked time are different timestamps.
  • Refresh volatile or consequential fields when they matter, not simply on one schedule for every record.

Overview

A recent database update does not prove a recent observation. Keep event time, collection time and last verification time separate when they differ. Refresh schedules should reflect field volatility and consequence: company registration details and active job openings do not need identical treatment. A stale value may remain useful if its date and limitations are visible.

How it works

  1. Record source observation and verification dates, not just ingestion time.

  2. Set freshness thresholds by field and use case.

  3. Refresh, expire or flag values when they no longer meet those thresholds.

Keep the relevant clocks separate

A dataset imported today may contain a job title observed last year. A page crawled this morning may repeat an announcement from several months ago. Store the event or observation date separately from ingestion and verification dates. Otherwise a recent pipeline run can make stale facts look current.

Define freshness according to the field and use. A broad industry classification may remain useful for a long period, while current employment or a time-sensitive buying signal may need a recent check. There is no single expiry date that accurately describes every attribute in a B2B record.

Timestamps that answer different questions
TimestampMeaningWhat it does not establish
Event timeWhen the underlying change happenedWhen your system learned about it
Observation timeWhen a source observed the valueThat the value remains true indefinitely
Ingestion timeWhen the value entered your systemThat the source information is new
Verification timeWhen a particular check ranThat every field in the record was checked

Set refresh rules around the decision

For prospecting, current role and contact status may deserve a check before delivery or outreach. For market sizing, a dated company-size estimate may be adequate if the model acknowledges uncertainty. Define the acceptable age and the action when a value exceeds it: refresh, review, downgrade confidence or stop the handoff.

Use change events where they are reliable. A detected employer change should invalidate the old employment relationship even if a periodic refresh is not due. Do not automatically discard historical observations; they may explain why an earlier decision was reasonable and support a later audit.

Spend refresh effort where it changes outcomes

An illustrative database contains 50,000 contacts, but only 500 are scheduled for a current campaign. Refreshing the fields required for those 500 can be more useful than rerunning every lookup for all records. The correct approach depends on other uses of the database and the consequences of stale values, not only the lookup cost.

Measure stale-value discoveries, correction rates and the age distribution of accepted records. A low average age can hide a tail of very old critical fields. When a provider claims a refresh cadence, ask whether it means every field is re-observed, records are revisited or only the database is updated somewhere during that period.

ILLUSTRATIVE EXAMPLE

What this looks like in practice

A provider reimports an old job posting today. The ingestion timestamp is current, but the hiring signal remains old; the workflow uses the posting date to decide whether it is still actionable.

Examples explain the concept; they are not reported customer results.

What to check

Measure age distributions and expired-field coverage. Verify that failed refreshes do not reset the “last verified” timestamp.

Common mistake

Updating a row’s timestamp after any background job and presenting every field in the row as newly verified.

Data freshness vs. Data accuracy

Freshness concerns timing. Accuracy concerns correctness. A newly collected value can be wrong, and an older value can remain correct, so both dimensions need attention.

Read the Data accuracy definition →

Questions answered

What is Data freshness?

Data freshness describes how recently information was observed or verified relative to how quickly it can change and how current it must be for the intended decision.

How often should contact data be refreshed?

Use the importance of the field, expected change rate and upcoming action to set the cadence. Rechecking before consequential delivery may be more useful than a universal schedule.

Does a current source page mean every fact is current?

No. Pages can retain historical facts or republish old events. Inspect the date associated with the specific claim.

Does a recently updated record mean all its fields are fresh?

No. One changed field can update the record timestamp while other attributes remain old. Use field-level observation dates where freshness matters. A recent import or sync timestamp should not replace the date of the evidence supporting employment, company size or another specific fact.

How quickly does B2B data decay?

The rate varies by field, source, market and definition of an error. Avoid applying one industry percentage to every dataset. Measure change and correction rates in your own population, then set refresh policies according to the decision and the consequences of using stale information.

References and further reading

Primary documentation and source material for this topic. Sources checked September 14, 2026; provider requirements can change.

  1. Data quality dimensionsIBM
  2. PROV overviewW3C

Continue reading on the blog

Explore all articles and guides →

Put the concept to work.

Explore the relevant AstroFabric workflow and see how the pieces connect.

Help keep this guide useful. Suggest a correction or browse the full glossary.