Marketing Data Quality for media technology companies: Audit Checklist

Media technology companies generate marketing data across campaigns, content, partner placements, product trials, account systems and analytics tools. A field can be technically populated yet commercially misleading: an “engaged account” may be a repeated content view, a conversion may be a duplicate request, and a campaign label may change halfway through a reporting period.

This audit checklist looks at meaning and use as well as formatting. It helps an operating team decide which data can support a report, which needs repair and which should be removed from a decision. It is not a data-certification standard, advertising benchmark or privacy approval.

1. State the audit objective

Write the decision the audit must support: release a dashboard, reconcile campaign reporting, change a routing rule, evaluate a vendor, or retire a field. Define systems, markets, period, audience and excluded data. An audit without a decision can accumulate observations forever.

Record the audit owner, reviewer, evidence window and closure date. If a field is out of scope, write that down so a future report does not treat the omission as a pass.

2. Build a field inventory

List field name, business meaning, source, owner, transformation, allowed values, refresh expectation, downstream use, retention and correction path. Include derived metrics and spreadsheet columns, not only database schema.

| Field question | Evidence | Failure example | |—|—|—| | What does the field mean? | Definition and sample | Same label used for two states | | Where did it originate? | Source and timestamp | Manual overwrite with no note | | Who uses it? | Report, route or decision | No downstream owner | | How is it changed? | Mapping or formula version | Hidden transformation | | How is an error corrected? | Ticket, replay or change log | Correction stops in one system |

3. Check definition consistency

Compare the field definition across marketing, sales, product, analytics and partner reporting. Mark terms that sound similar but represent different states: response, engaged account, accepted lead, trial, subscriber, opportunity or customer.

The Google Analytics GA4 Event reference can help describe an observed event. It does not decide what an engaged media buyer or qualified account means. Keep event name, business rule and downstream action separate.

4. Test lineage and transformations

Choose a small sample and trace it from source to report. Record filters, deduplication, joins, attribution windows, currency or time-zone conversions, manual edits and late-arriving data. Repeat the trace after a field or campaign-label change.

If a result cannot be reproduced from the current mapping, mark lineage as failed even when the number appears plausible. Preserve the prior mapping and show the bridge to a new result.

5. Examine completeness and validity

Do not treat every blank as the same defect. A field may be intentionally absent, not yet available, invalid, or missing because a route failed. Define allowed blank states, format rules, range checks, referential checks and expected delay.

For a campaign source, test unusual characters, renamed campaigns, partner traffic, direct visits and a duplicate parameter. For an account field, test a changed domain, parent-child relationship and existing customer.

6. Review timeliness and freshness

A fresh-looking timestamp can still contain stale content. Record when the source was created, when it was transformed, when it reached the report and when it was last reviewed. A data-quality issue is sometimes a timing issue rather than a bad value.

Set a freshness rule tied to a decision. A daily operations route may need a shorter tolerance than a quarterly planning view. Do not invent a universal threshold.

7. Inspect duplicates and identity rules

Media technology records can arrive through several domains, devices, partners and content forms. Define what counts as a duplicate, possible duplicate, continuation or new business context. Keep the original source and the merge decision.

An aggressive merge can remove a valid buying signal; no merge can inflate a report. Test both false-match and missed-match cases and name who can reverse the result.

8. Reconcile platform transport

The Google Ads conversion import guidance is an implementation reference for moving conversion information between systems. It does not prove the quality of the internal state or the incrementality of media activity.

Compare source event, imported value, internal state, receiving action and correction history. If a campaign report uses an imported event, show the mapping and its limitation instead of assigning confidence to the transport alone.

9. Check privacy purpose and access

Marketing data may contain contact details, account context, content behaviour or partner-provided attributes. The NIST Privacy Framework is a voluntary tool for identifying and managing privacy risk. Use it to ask why a field is needed, who may access it, how it is communicated, and how a correction or deletion request travels.

Do not expand an audit into unnecessary personal-data collection. Use synthetic records for testing where names or identifiers do not affect the decision.

10. Grade issue severity

Use a local severity rule: blocking changes a public, financial or operational decision; material can distort a segment or route; contained affects a narrow view with a workaround; informational needs documentation. Explain the rationale and affected scope.

Severity is not a synonym for business value. A small field can be blocking if it controls a permission route, while a large missing field may be contained if no active decision uses it.

11. Link quality to a decision

The NIST Information Quality Standards offer a useful lens for utility, objectivity, integrity and correction. Translate the lens into “what decision would change if this issue were fixed?” If no decision changes, document the field rather than assigning urgency by intuition.

The GOV.UK Measuring Success guidance is a process reference for connecting measures to decisions. Use it to define the owner, action and review date for a data-quality metric, not to claim a media-technology benchmark.

12. Close issues with evidence

An issue is not closed because a ticket says “fixed.” Attach a new sample, mapping version, replay result, reviewer and affected report. State whether historical data was corrected or only future data will be clean.

If a correction changes a published report, record the communication and date. If the team accepts a limitation, make the limitation visible to the report audience.

13. Copy-ready audit record

“text Audit objective / decision / scope / period: Field, meaning, source, owner and downstream use: Lineage and transformation version: Completeness, validity, freshness and duplicate test: Privacy purpose, access and correction path: Issue severity and affected decision: Remediation, owner, dependency and stop rule: Closure sample / replay / reviewer / date: Historical correction or stated limitation: Next audit trigger and review cadence: “

Good marketing-data quality is not a claim that every field is perfect. It is the ability to explain what a field means, how it reached a decision, what is uncertain and how the team will correct it. That is the standard a media technology audit can actually test.