B2B content portfolio decisions

B2B content audit: decide what to keep, improve and publish next

A B2B content audit reviews existing pages against the buying questions they should answer, their accuracy and their observed contribution. It helps decide what to keep, improve, combine or retire and where a genuinely new page is needed. Build an inventory, join comparable page and query evidence, review usefulness, then document each decision with its risks, owner and review date.

Written by Noah Levy · Updated October 28, 2026

On this page

Define the decision before collecting every metric

Choose the business problem: unclear service pages, outdated claims, overlapping guides or missing help for a buying decision. Define the audience, offer, languages and portfolio boundary. A content audit is a decision exercise; a spreadsheet containing every available metric is only an input.

Distinguish search acquisition, buyer evaluation, customer support and internal-only material. An important service explanation may influence a buyer without being the first landing page. A public guide can attract relevant demand without closing a sale on the same visit. Do not judge these pages by one common pageview target.

Ungated content planning

Record the decision for each existing page

Inventory the page, its evidence and its decision separately. Load the complete six-page portfolio, then build a report or CSV containing your own decisions. No thresholds automatically delete or merge pages.

1Inventory

Name the page and the buying question it serves.

2Evidence

Record the period, usefulness and any missing data.

3Decision

Write keep, update, combine, retire or hold, with a responsible owner.

Private, ungated workbook. Add up to 50 entries, load the complete fictional example, build a report and export CSV or print. Inputs stay in this page and disappear on reload. No automatic audit, score or verified outcome is produced.

Entry 1
Content inventory and decision record

No completed entries. Unknown is not zero.

No account, upload or data collection. Inputs disappear when this page reloads. Build and export the report before leaving. Without JavaScript, use the blank fields and the completed example in the article manually.

Inventory canonical pages and the job each one owns

Record URL, language, format, buyer question, buying stage, primary service, owner and last substantive review. Group language counterparts and genuine duplicates without assuming they should share a canonical. Include important pages missing from navigation; exclude private material from the public inventory.

Mark the primary URL for each buying question. A service page answers whether you can do the work; a buyer’s guide helps evaluate providers; a pricing guide explains units and scope. Similar vocabulary does not make those jobs interchangeable. Give each new-page proposal an explicit reason an existing URL cannot serve it.

Join evidence without turning missing data into zero

Bring together page-filtered Search Console queries, organic landing sessions, relevant referrals and source-reconciled booking evidence where available. Record dates, geography, device, attribution coverage and the last meaningful page change. Compare the same cohorts, with enough time after publication for the question being tested.

Read impressions and clicks together. Falling CTR can reflect a new mix of queries; declining clicks may follow lower demand, lost exposure, a position change or a different search presentation. Check query-level changes before changing the page. A long sales cycle can also separate the initial landing from a later booking.

Unknown attribution, hidden query rows and a page too new to assess remain unknown. An indexed page with no observed clicks is different from one that has never been crawled. Neither state automatically proves the content is worthless.

Review usefulness, accuracy and proof as an editor

Check whether the page gives a direct answer, helps a real decision, states relevant limits and supplies verifiable evidence. Review current prices only against approved facts; check every language’s commercial meaning, examples, tool instructions, sources and handoff. Record unsupported claims for correction rather than strengthening them.

Inspect useful competing answers for the same buyer question. Look for missing decision criteria, examples, responsibilities or evidence. Add only what improves the decision. More headings, repeated keywords and a larger word count do not establish an advantage.

  • Can a buyer understand who the page is for and when the advice does not apply?
  • Are examples clearly fictional or supported by approved public evidence?
  • Does the next link help the buying question and stay in the selected language?
  • Are the title, answer, visible FAQ and structured data consistent?

Choose keep, update, combine or retire with safeguards

Treat the four actions as editorial decisions with different risks. First fix demonstrated inaccuracies and broken journeys. Preserve useful pages while testing the smallest relevant change. Combining or retiring pages needs checks of unique intent, links, conversions and language relationships before implementation.

Content decision matrix
ActionWhen evidence supports itSafeguard
KeepAccurate, useful and owns a distinct buyer questionRecord next review; preserve working elements
UpdateSame intent; a demonstrated answer, proof or usability gapKeep URL; record exact change and baseline
CombineReal duplication of intent with no valuable unique roleReview demand, links and outcomes; map redirects and references
RetireNo remaining useful public job and a reviewed replacement or removal planCheck backlinks, internal links, language alternatives and appropriate response

Work through a fictional three-page inventory

These rows are invented to show the decision method. They are not client performance, market benchmarks or recommendations to delete equivalent pages on your site.

Fictional content inventory
PageObservationDecisionNext check
Service pageRelevant buyers visit; scope description is outdatedUpdate approved scope at the same URLValidate facts and later buyer outcomes
New localized guideNo mature query or booking evidence yetKeep; verify discovery and language linksReview after a documented observation window
Two older guidesPossible overlap; unique questions and backlinks not checkedHold consolidation pending evidenceMap intent, links and outcomes first

Approve a focused improvement and publishing plan

Prioritize by buyer value, strength of evidence, risk, effort and dependency. Give each action an owner and a verification date. If a new page fills a distinct gap, define its primary question, boundaries, language equivalents, useful original help and internal links before drafting. Do not publish another page solely because a keyword tool shows volume.

After implementation, verify the live page, assets, navigation and source tracking. Compare mature page/query cohorts and genuine booking quality, not an immediate total-session increase. Keep a change log so later reviews can distinguish a content edit from seasonality or an attribution change.

Build an inventory that can support a decision

Start with one record per canonical public URL and a separate language-family identifier. Keep the observed URL and redirect destination when they differ; do not silently join unrelated pages because their titles resemble one another. Add language, buyer question, stage, linked service, owner and last substantive change.

Join Search Console page/query records to the canonical using an explicitly documented mapping. Record dates, country, device and disclosed-query coverage. Keep GA4 landing sessions separate from pageviews and from all sessions in which a page appeared. Join booking evidence only where a legitimate identifier or source record supports it. Missing joins remain unknown.

The workbook separates these fields so a future editor can reproduce your decision. Search performance, landing activity, original-created bookings, held calls and qualification describe different things; do not put them in one undifferentiated “conversions” column.

A completed fictional portfolio: six reasons to make different decisions

All counts, dates and paths below are invented training data, not Beespoke or client results. The comparison window is two complete 28-day periods, in one unchanged reporting setup. Booking evidence is separate from search and landing records.

Fictional portfolio and action register
Page and roleEvidence and unknownsDecision with reasonApproval and retest
/services/calibration/ · evaluate providerRelevant query clicks 8→9; organic landings 12→14; one original-created website booking verified, qualification unknownKeep accurate service content; improve the specific unanswered scope questionCommercial owner approves scope; review matched query and genuine outcomes
/guides/calibration-cost/ · budgetClicks 7→3; old source describes a retired process; query mix changedUpdate verified facts and missing cost drivers; retain URLSubject owner verifies primary source; annotate revision and compare mature cohort
/fr/guides/calibration-cost/ · French equivalentPublished recently; index and performance evidence not yet availableHold performance judgment; retain and check discoveryEditor verifies full language journey; date next observation
/downloads/site-preparation/ · prepare visitOne landing; buyers use it after discovery; SEO volume smallKeep useful supporting asset; do not delete for low trafficSales owner reviews usefulness, accuracy and next step
/guides/calibration-basics-a/ · basic explanationSame apparent question as B; external links and unique passages unreviewedPossible combine, not authorized removal; compare both firstSEO/editor checks intent, links, equivalents and redirect plan
/guides/retired-procedure/ · discontinued taskOwner confirms procedure retired; no useful replacement of the same job; links still need auditRetirement candidate after review, not automatic redirect to homepageOwner approves; resolve incoming links and language dependencies; validate final status

Use conditional decisions rather than traffic thresholds

Keep when the page is accurate and serves a distinct useful job. Update when the job remains valid but facts, answers, proof or usability have a demonstrated gap. Hold when age, attribution, index state or ownership is not assessed. Low clicks alone do not distinguish these cases.

Combine only when the same buyer job is genuinely duplicated and the retained destination can satisfy both audiences without losing useful material. Review incoming links, unique queries, booking contribution, language relationships and bookmarks. Prepare the destination and explicit redirect mapping before an approved retirement. Never redirect an unrelated retired page to a convenient service merely to avoid a 404.

Retire only after the responsible owner confirms that the job is obsolete, wrong or outside scope and useful alternatives have been assessed. Preserve the reasoning and rollback dependency. A content-audit worksheet does not authorize deleting URLs or changing commercial claims.

Make the handoff repeatable and keep the inventory current

Assign an editorial owner for facts, an implementer for changes and a validator for live acceptance. A task needs the URL, buyer question, current evidence, smallest proposed change, approval hold, release and retest. Keep source-review dates separate from publication dates.

After substantive publication, check the final language pages, metadata, FAQ parity, assets, links and contact path. Then use sufficiently mature matched periods for search and source-aware booking evidence. A technical pass completes implementation; it does not prove that the new article caused an additional sale.

Use the local CSV as a portable decision register. Copy references rather than private prospect details. Load the six-row example to inspect the full decision trail; export your own notes before leaving because this page does not save them.

End with decisions, not a publishing quota

Improve the existing portfolio where evidence supports it. New pages should fill distinct buyer gaps and preserve valuable current journeys.

  1. Name one primary buying question per page.
  2. Write the evidence, unknowns and risk of each action.
  3. Approve redirects or removals only after the dependencies are reviewed.
  4. Track the release and compare mature outcomes.

Frequently asked questions

What is the difference between a content inventory and an audit?

An inventory lists what exists. An audit adds evidence, a review of buyer usefulness and accuracy, and an explicit decision with an owner and verification date.

Should we delete every page with no organic traffic?

No. Check its age, discovery and index state, buyer role, unique information, links and available conversion evidence. Missing or immature data does not establish that removal is beneficial.

When should we build a new guide?

When a distinct, relevant buying question cannot be served well by an existing URL. Define the scope and internal journey first; use keyword estimates as supporting evidence, not a publishing quota.

How do we know an update worked?

Record the baseline and exact change, verify implementation, then compare mature equivalent query/page cohorts and source-aware business outcomes. Treat external changes and incomplete attribution as limitations.

Sources and methodology

The linked publisher guides inform the scope comparison. The decision matrix and worksheet are Beespoke planning methods, not a validated ranking model. All worked rows are fictional; no client data or automatic deletion recommendations are supplied. Sources reviewed October 27–28, 2026.

  1. PathFactory: B2B content audit
  2. Lform: B2B content audit checklist
  3. Beespoke editorial policy

Build the right demand system for your market

Bring your offer, buyer, current acquisition data and sales process. We will identify whether paid search, organic search, AI visibility, outbound or a focused combination deserves the next test.

Book a conversation