INF2170H Chap.2 Content Inventories and Audits
Content Inventories and Audits
Build evidence for a decision
An audit is the purposeful review process; an inventory is the organised artifact created and refined during that work. This chapter begins with scope and unit of analysis, then shows how crawl data, manual observations and analyst defined dimensions can be combined without hiding their provenance.
Counts locate scale and outliers, while contextual review explains what action the pattern supports.
Carry findings into action
Cleaning, normalisation and classification change the dataset, so each transformation needs a rule.
The chapter links row level evidence to actions such as keep, revise, merge, archive or investigate, and finishes by verifying the implemented result rather than treating a spreadsheet status as completion.
Build the inventory around a real decision
A migration, redesign and accessibility review need different evidence.
Start by naming the decision, then define the corpus through hosts, paths, content types, dates and exclusions. Specify what one row represents. Pages, files, placements and reusable components can all be legitimate units, but mixing them without relationships produces misleading totals.
When one file appears on several pages, preserve both the authoritative object and its placements instead of choosing whichever count looks cleaner.
Choose fields because they support the decision. Identity fields establish which object is present. Structural fields locate it. Descriptive fields explain type or topic. Governance fields record ownership and review.
Judgment fields support an action under a documented criterion. A field derived by the analyst should be marked as such and retain the source value from which it was inferred. That distinction lets another reviewer reproduce or challenge the framing.
Read numbers through the content they describe
Counts provide scale and comparison, but they do not supply purpose or consequence.
A section with the most pages may be healthy; a single obsolete policy can carry greater risk. Inspect distributions rather than relying on averages, stratify records by meaningful type and review outliers against the underlying content. If an automated field is inconsistent, repair the extraction or narrow the claim before using it in a recommendation.
Translate findings into controlled actions.
Keep, revise, merge, redirect, archive and investigate should have clear meanings, owners and verification steps. An accurate page with no owner may be retained while governance is resolved. Two matching titles may require clearer labels rather than a merge. After implementation, rerun the crawl, inspect repaired metadata and test priority routes.
The evidence chain ends only when the action is verified against the problem that initiated the audit.
What this chapter covers
- 01
Audit and inventory
- 02
Scope definition
- 03
Unit of analysis
- 04
Qualitative evidence
- 05
Quantitative evidence
- 06
Action vocabulary
Worked example · free
Plan a migration inventory
- 2Define the corpus and row unit.
- 3Choose fields tied to migration.
- 2Separate automated and manual work.
Key terms
- Content Inventory
- Content Inventory is the organised artifact that records content objects and selected attributes.
- Content Audit
- Content Audit is the evidence gathering and interpretation process designed around a content decision.
- Unit Analysis
- Unit Analysis defines what one record represents for counting and comparison.
Content Inventories and Audits FAQ
Why is a crawl not a finished inventory?
A crawl provides scalable identity and technical fields, but it may include duplicates, irrelevant columns and missing context. Analysts must define the corpus, clean the rows, preserve raw values and add purposeful dimensions such as content type, audience or owner before drawing a recommendation.
How should qualitative and quantitative evidence be combined?
Use counts to locate scale, distributions and outliers, then inspect content to explain the pattern. A large section is not automatically poor and a small section may carry serious risk. Report what each form of evidence addresses instead of averaging different properties into one score.
What makes an audit recommendation traceable?
A reviewer should be able to move backward from the action to the finding and affected records, then forward to an owner and verification step. Use specific action verbs, preserve the transformation rules and rerun the relevant check after implementation.
Assessment move
Practise by taking a messy list and writing the decision, corpus boundary and row unit before adding fields. For each finding, name the quantitative pattern, the manual context and the action that follows. Keep a change log for cleaning and normalisation. Create a small original inventory with enough variation to expose duplicates, missing ownership and inconsistent values.
Before counting anything, define each column and decide how blanks, repeated values and unavailable pages will be recorded. Separate observations gathered automatically from judgments made during review. Then write three findings in the same pattern: the measured signal, the inspected examples, the decision and the verification step.
Challenge each finding with a counterexample so that a large count does not become a conclusion by itself. Practise changing the row unit from page to object or component and explain how the change affects totals and recommendations. Preserve raw values before normalising them, and record every merge or split in a change log.
End by checking whether the proposed action follows from the stated corpus rather than from unexamined assumptions about the whole site. This makes the audit reproducible and useful after the immediate migration decision.
