What is data hygiene?

Ongoing correction of duplicates, stale values, missing fields and broken references. Distinct from data governance, which assigns who may define and change data in the first place.

Why does data quality keep degrading?

Because the flow keeps producing defects. A cleanup fixes the stock; without entry controls the record set returns to its previous state at whatever rate the process generates errors.

What should you clean first?

Whatever a live report, routing rule or automation actually reads. Fields nothing consumes can be left dirty or retired; fields that feed decisions cannot.

How do you measure data hygiene?

As rates, not counts: duplicate rate, null rate on fields that rules read at the moment they fire, and staleness against a known-good source. Rates are comparable over time; counts grow with the database.

Data Hygiene

The continuous practice of keeping records accurate, complete and current — as distinct from data quality, which is the state that practice produces.

Stock and flow

A cleanup addresses the stock of bad records. The flow — form fills, imports, integrations, manual entry — keeps producing more. Without changing the flow, the corrected state decays at whatever rate the process generates defects, which is why the second cleanup is usually proposed within a year of the first.

The measurement that matters

Not how many bad records exist, but how fast they appear. Sample the records created last week rather than the whole database: that rate tells you whether the flow is fixed, and the total never will.

What to measure

Duplicate rate by object, on recently created records.
Null rate on every field a routing rule or automation reads, measured at the moment the rule fires rather than after enrichment completes.
Staleness — the share of records where a field disagrees with an external source you trust.
Divergence between systems, sampled on matched records. This is the health measure for every integration you run.

Fixing the flow, in order of yield

Match on submission, so a form fill updates the existing record instead of creating a second one.
Validate imports before they load — this is where large duplicate batches originate.
Retire fields nothing reads. Fewer fields means fewer to keep correct, and it reduces the work the system imposes on its users.
Enforce completeness before routing rather than after, which removes the rework that incomplete records generate downstream.

RELATED TERMS

Data Governance

The rules determining who may change what, how changes are reviewed, and how they are recorded. The control that stops a designed system from drifting back to whatever it was before.

Enrichment

Appending third-party data to records so routing, scoring and segmentation have fields to operate on.

Pipeline Hygiene

Whether pipeline records are accurate and current — as distinct from pipeline health, which is whether the pipeline is any good.

Single Source of Truth (SSOT)

The declared authority for a given piece of data — the system whose value wins when two systems disagree. Correctly applied it is decided per field, not per system.

COMMON QUESTIONS

What is data hygiene?: Ongoing correction of duplicates, stale values, missing fields and broken references. Distinct from data governance, which assigns who may define and change data in the first place.
Why does data quality keep degrading?: Because the flow keeps producing defects. A cleanup fixes the stock; without entry controls the record set returns to its previous state at whatever rate the process generates errors.
What should you clean first?: Whatever a live report, routing rule or automation actually reads. Fields nothing consumes can be left dirty or retired; fields that feed decisions cannot.
How do you measure data hygiene?: As rates, not counts: duplicate rate, null rate on fields that rules read at the moment they fire, and staleness against a known-good source. Rates are comparable over time; counts grow with the database.

WHERE THIS HAS BEEN APPLIED

Client work and research from RevOps HQ, our consulting practice.