# Audit a content collection before adding another article

Complete original constructed teaching cases. No real people, studies, source datasets, customers, accounts or external publications.

Inventory the collection by reader task, exact edition, owner and dependencies, then choose what to keep, revise, consolidate or retire. A list of URLs is the starting record; maintenance decisions need the actual content and evidence. Count language counterparts together when assessing topic coverage, and preserve useful historical records without presenting them as current instructions.

## Define the collection and the unit of review

Choose a bounded collection you can legitimately inspect: your reporting guides, their English and German editions, downloadable worksheets and the posts that introduce them. Record where each item actually lives. An exported CMS list can omit attachments, old campaign pages or native articles; search results can miss material too. Write those discovery limits before calling the inventory complete.

GOV.UK’s maintenance guidance illustrates why tool coverage matters: its Govsearch excludes blogs, campaign sites and service domains. That is a government-specific tool boundary, not a service for auditing a creator’s blog. Apply the general checking question to your own sources: which channels, states and files does this inventory include, and which must be checked elsewhere?

## Read the job, edition and dependencies together

For each item, retain a stable local ID, real destination, language, intended reader task, content state, responsible owner, last substantive review and known supporting evidence. Add important incoming and outgoing references and attached artifact versions. Keep unpublished, scheduled, published and historical material separate. A recent modification time can reflect a template change rather than a factual review.

Read neighboring pieces before labeling a gap. Two titles may solve one task; one broad title may hide several unfinished questions. Record the actual decision a reader can complete and what the article leaves to a linked guide. An assigned topic code helps organize your judgment but does not prove semantic distinctness. Translations and small keyword changes do not supply extra reader jobs.

## Make a portfolio decision with an explicit loss check

The original five-item exercise has two language editions for J1, an overlapping English variant for J1, a past-event record J2 and a chart guide J3 with an unresolved source period. That is five artifacts and three declared jobs, not five established independent topics. Reading the bodies is still needed to defend the assignments. None of these rows describes an actual site or reader.

Keep useful current work; revise a material gap; consolidate genuine overlap only after preserving necessary differences; retain a clearly historical record when that is its legitimate purpose. Retirement needs a reason, affected references and an appropriate destination or explanation where the publishing system supports it. Do not invent a redirect, native attachment replacement or successful deletion. Unknown evidence can justify holding a claim while retaining the review record.

| Constructed item | Declared job | Maintenance proposal |
| --- | --- | --- |
| A1 · English | J1 period | Keep |
| A2 · German | J1 period | Keep counterpart |
| A3 · English | J1 period | Review overlap before consolidation |
| A4 · historical | J2 event | Label past context |
| A5 · source open | J3 chart | Hold unsupported period |

## Resolve three collection decisions

I1 — Constructed overlap: A1 explains the selected reporting period in English; A2 is its complete German counterpart; A3 repeats A1 with a different title. Proposed decision: keep A1/A2 as one bilingual job and compare A3 for any unique useful detail before consolidating it. Retain a destination map and actual publishing-state check. Outcome: a maintenance proposal, not an implemented redirect, an extra translation topic or a measured search improvement.

I2 — Constructed historical item: A4 announced an event that has ended. The date and original announcement remain useful in a project history, but the invitation no longer offers a current action. Proposed decision: label the event as past and remove misleading current invitation links where authorized. Keep legitimate historical context and necessary records. Outcome: a scoped change plan; neither low traffic nor age alone proves that every historical item should disappear.

I3 — Constructed unresolved evidence: A5 shows a chart whose caption names October while the available source has no verified period. Proposed decision: hold the unsupported period claim, identify the source version and review dependent image, alternative text and worksheet. Do not substitute the current dashboard total or change only the article date. Outcome: a visible evidence gap and prioritized repair, not a completed source recovery or better reader result.

## Prioritize and verify the actual maintenance

Prioritize by the consequence of an error, unresolved access or privacy, task importance, evidence strength and repair effort. Traffic can help locate problems but is not the sole measure of value. A rarely used correction guide can matter at the moment it is needed. Keep the reason for each priority and the proposed owner; do not turn a scoring spreadsheet into objective proof of usefulness.

Before changing a live item, freeze the exact affected editions and files. After authorized maintenance, reopen the real destination, links and obtainable attachments in the intended access context. Check both languages when the same fact appears in both. Keep an earlier useful record and a rollback or correction route where appropriate. An updated inventory row is not evidence that the actual page changed.

## Keep a usable audit record without hoarding data

The private worksheet records discovery scope, item/edition, reader-task evidence, overlap, source and file dependencies, proposed action, expected loss, responsible owner and actual post-change verification. Use local IDs in examples. Avoid copying subscriber lists, private messages or source datasets into an editorial inventory when a minimal private reference is sufficient.

The government content-design bodies were read on 1 November 2026. They describe public-sector publishing; our creator inventory, decisions and five-item example are an original adaptation. No content was retired, merged or redirected by this packet. Its local checker verifies declared fixture counts and identities, not coverage, permissions, semantic uniqueness, reader benefit or a live publishing result.

1. Record discovery scope and tool gaps.
2. Read actual editions and neighboring reader jobs.
3. Map counterparts, overlap and dependencies.
4. Choose maintenance with loss review and owner.
5. Verify authorized changes at the real destination.

## Primary sources

- [GOV.UK: Inventory scope and existing-content maintenance](https://guidance.publishing.service.gov.uk/writing-to-gov-uk-standards/plan-manage-content/manage-existing-govuk-content/)
- [GOV.UK: Evidence and task-based user needs](https://guidance.publishing.service.gov.uk/writing-to-gov-uk-standards/plan-manage-content/identify-user-needs/)
- [GOV.UK: Content design and duplication](https://guidance.publishing.service.gov.uk/writing-to-gov-uk-standards/plan-manage-content/understand-content-design/)