A suite of 6 products has 2,400 pages split across 11 team-owned sites, and customers cannot tell which product owns a task; design the information architecture for a 9-month consolidation.

I would organize the suite around shared customer tasks while preserving product context as structured metadata. - A content inventory records owner, audience, task, product, lifecycle state, traffic, and duplicate cluster for all 2,400 pages. - The target map uses task-based landing pages, product filters, and one canonical page when workflows are identical across products. - An information-architecture decision record defines URL, breadcrumb, navigation, and cross-product linking rules before migration starts. - I pilot the model on the 200 highest-traffic pages and require 80% task-findability before scheduling the remaining waves.

Define a content model for 1,800 pages covering 4 cloud products, 3 personas, and 6 locales, with Contentful entries feeding web and in-product help within 12 months.

I would model durable content semantics separately from either delivery channel. - Core types include concept, task, reference, troubleshooting, release note, and reusable notice with required audience, product, version, and locale fields. - Relationships connect prerequisites, next steps, API objects, and product variants without copying whole pages into Contentful entries. - A schema registry names field owners, validation rules, allowed values, and backward-compatible change procedures for web and in-product consumers. - I validate the model against 120 representative pages and reject it if more than 10% need unstructured escape fields.

A 3,200-page developer portal uses 700 inconsistent labels across 8 teams; design a taxonomy that must improve successful search sessions from 58% to 75% in two quarters.

I would replace local labels with a governed faceted taxonomy derived from user language and search evidence. - Query logs, support terms, and card sorts define controlled values for product, task, language, framework, API version, and experience level. - A synonym map redirects legacy and colloquial terms while one preferred label appears in navigation, metadata, and search results. - Taxonomy owners review proposed terms through a change request that records definition, scope, aliases, and affected pages. - I backfill the top 1,000 pages first and compare zero-result rate, refinement use, and successful sessions against the 58% baseline.

Twelve engineering teams own 2,600 pages, but a central team of 4 writers has a 5-day review SLA; design a federated ownership model without making the writers a bottleneck.

I would give product teams accountable ownership and reserve central review for shared or high-risk content. - CODEOWNERS maps directories to an engineering owner and a documentation steward, with no page allowed to have an unowned state. - A risk matrix routes authentication, billing, security, and cross-product pages to writers while routine release updates use automated checks and peer review. - Each team receives freshness, broken-link, and overdue-review reports for its own pages rather than a central cleanup queue. - The operating agreement sets a 5-day central SLA, a 2-day team response SLA, and quarterly ownership audits for all 2,600 pages.

Reorganize 1,500 pages for 5 products into a Diátaxis model in 6 months while 9 teams continue shipping weekly; what architecture and transition artifacts do you define?

I would use Diátaxis as an intent model, not as four disconnected folders. - A classification rubric distinguishes tutorials, how-to guides, reference, and explanation through user intent, entry state, and completion outcome. - A crosswalk maps every existing page to keep, split, merge, rewrite, or retire and names the target type and owner. - Navigation supports journeys across types, such as tutorial to task to reference, instead of forcing users to understand the framework. - A 150-page pilot tests classification agreement above 85% and task completion before six monthly migration waves begin.

Search across 4,800 pages in 8 products and 7 locales returns the wrong version in 18% of sessions; design the index, metadata, and relevance controls for a 99.9% search SLA.

I would make product, version, locale, and lifecycle state first-class search signals. - The index stores canonical ID, product, locale, version range, content type, freshness, permissions, and replacement URL for every document. - Query routing applies the current product and locale as boosts, excludes retired pages, and exposes version filters rather than silently guessing. - Relevance tests cover 200 high-value queries with expected top-three results and run before each search configuration release. - Dashboards track latency, zero-result rate, wrong-version clicks, and index freshness against the 99.9% SLA by locale.

Plan the migration of 3,500 Confluence pages from 14 spaces to one developer portal within 10 months and a $180,000 budget, while preserving 12,000 inbound links.

I would migrate by validated content cohorts rather than copy all 14 spaces into a new shell. - The inventory captures owner, traffic, last verification, attachments, restrictions, duplicate groups, and inbound links for all 3,500 pages. - A disposition register assigns keep, merge, rewrite, archive, or delete and requires product-owner approval for high-traffic removals. - Automated converters handle standard pages, while the budget reserves manual work for macros, diagrams, restricted content, and 12,000 redirect mappings. - Ten monthly waves include link-crawl, visual, permission, and owner sign-off gates before each source space becomes read-only.

A 900-page administration product serves operators, auditors, and developers across 5 teams; design navigation that supports 30 recurring tasks without duplicating role-specific pages.

I would build one task architecture with role-aware entry points and conditional detail only where permissions change the procedure. - A task-to-role matrix identifies the 30 workflows, shared steps, permission differences, and required evidence for each persona. - Global navigation follows product tasks, while persona landing pages curate relevant tasks without creating alternate canonical copies. - Shared procedures use reusable steps and explicit role prerequisites rather than nearly identical operator and auditor pages. - Tree tests with 15 participants per role must reach at least 80% first-click success before the 900-page structure ships.

Design lifecycle metadata for 2,200 pages across 7 products where 22% have no owner and regulatory pages require annual verification with a 30-day remediation SLA.

I would make ownership and verification state enforceable publishing data rather than optional labels. - Required fields include accountable team, subject expert, product, version, review interval, last verified date, risk class, and successor page. - Regulatory content receives a 365-day review interval, 60-day advance reminders, and automatic escalation when the 30-day remediation window opens. - Pages without owners enter a 45-day assignment queue and cannot receive feature updates until a product director accepts ownership. - A lifecycle dashboard reports unowned, overdue, expiring, and retired pages by product with weekly exports to Jira.

After an acquisition, merge a 1,100-page portal and a 1,700-page portal covering overlapping products in 4 locales within 8 months; define the canonicalization and migration plan.

I would establish one canonical topic map before moving either portal. - A pairwise audit groups pages as unique, equivalent, partially overlapping, or conflicting and records traffic, authority, owner, and locale coverage. - Canonical decisions favor the supported product behavior and strongest task outcome, not automatically the acquiring company's wording. - A merge specification names source sections, retained examples, terminology changes, redirect targets, and reviewers for every combined topic. - Four two-month waves require locale parity, redirect coverage above 99%, and no unresolved product conflict before cutover.

Architect a Docusaurus platform for 2,800 pages, 10 products, 6 locales, and 14 contributing teams using GitHub Actions, with pull-request feedback under 10 minutes.

I would separate shared platform code from product content while exposing one predictable contribution workflow. - Versioned content packages hold product docs, and a shared theme package owns navigation primitives, search metadata, accessibility, and locale behavior. - GitHub Actions runs changed-file linting, links, schema checks, and affected builds on pull requests, then the full matrix after merge. - Dependency and content caches key on lockfile, locale, product, and generator version so unrelated changes do not invalidate all 2,800 pages. - Platform SLOs set p95 preview feedback below 10 minutes, production availability at 99.9%, and named owners for each measure.

Design a Sphinx system for 1,600 Python-library pages across 12 packages and 3 supported versions, with API reference generated nightly and releases published in 20 minutes.

I would build each package independently and assemble versioned outputs through a shared Sphinx configuration. - Package projects own narrative RST or MyST sources and autodoc inputs, while one extension package controls theme, redirects, warnings, and version selectors. - Nightly reference generation compares public symbols with the prior artifact and opens a review issue for additions, removals, or signature changes. - Release workflows build only changed packages, then compose and validate the complete version manifest before publishing. - A 20-minute budget is split across environment setup, package builds, link checks, and deployment with timings retained for 90 days.

Choose a monorepo or multirepo model for 3,000 pages owned by 18 engineering teams and a 5-person docs platform group, with 40 product releases per week.

I would use product repositories for release-coupled content and a central repository for shared experience and governance assets. - API changes, examples, and release notes stay beside code so 40 weekly releases can update implementation and documentation atomically. - Shared tutorials, navigation, theme, Vale rules, and content schemas live centrally with versioned packages consumed by product repositories. - A federation manifest supplies canonical URLs, ownership, locale, and build inputs to one portal without copying source files. - An architecture decision record compares contributor access, atomic changes, dependency rollout, discovery, and recovery costs before adoption.

Design pull-request previews for a 2,100-page, 6-locale portal where 120 contributors need private links within 8 minutes and previews must expire after 14 days.

I would build isolated, access-controlled previews from the exact pull-request commit. - Changed products and locales are detected from the dependency graph, while shared-theme changes trigger the complete portal matrix. - Each preview gets an immutable URL, authentication through the company identity provider, and a visible commit, branch, and expiration banner. - GitHub checks post build status, changed-page links, visual-diff artifacts, and the private preview URL within the 8-minute target. - A scheduled job deletes preview storage and deployment records after 14 days and reports orphaned resources against a monthly budget.

Define the deploy architecture for a 4,500-page portal serving 9 products in 7 locales, with a 99.95% availability SLA and a 15-minute rollback objective.

I would publish immutable site artifacts behind atomic traffic switching. - CI creates a signed manifest containing commit, generator version, locale bundles, checksums, redirects, and search-index version. - Staging validates routes, links, accessibility, headers, and 100 critical journeys against the exact production artifact. - Deployment uploads to a new versioned origin and changes the CDN pointer only after health checks pass in every locale. - The previous two artifacts and indexes remain deployable so rollback completes within 15 minutes without rebuilding.

Specify GitHub Actions quality gates for 2,700 pages maintained by 16 teams, with a 12-minute pull-request budget and fewer than 1% false blocks per month.

I would divide checks by risk and run the cheapest deterministic gates first. - Every change receives schema, Markdown, Vale, internal-link, secret, and ownership checks scoped to affected files. - Code samples, external links, accessibility, and complete locale builds run only when their dependency paths or scheduled full validation require them. - Blocking rules need an owner, documented rationale, fixture tests, measured false-block rate, and a temporary waiver with expiry. - The dashboard tracks p50 and p95 duration, queue time, failures by rule, waivers, and false blocks against the 12-minute and 1% targets.

Design repository permissions for 1,900 pages across 13 teams, including security and billing docs that require two approvals and contractor access that expires after 90 days.

I would enforce least privilege through repository paths, identity groups, and merge-policy checks. - CODEOWNERS routes requests to product owners for normal paths but cannot itself enforce two separate approvals; for security and billing a required approval-policy check backed by a GitHub App or bot verifies one non-author domain-team approval and one documentation-team approval. - Branch protection blocks direct pushes and requires both the approval-policy check and a CI check that verifies release-artifact signatures and provenance before merge; branch protection does not verify signatures itself. - Contractors join scoped identity groups with automatic 90-day expiry and no access to restricted source or production deployment credentials. - A quarterly access review compares repository, CI, hosting, and secret-store permissions and records exceptions with owners and end dates.

A planned platform must build 5,000 pages across 8 locales in under 15 minutes on a fixed $6,000 monthly CI budget; design for build performance before migration.

I would establish a dependency-aware build model and a measured performance budget before importing all content. - A benchmark corpus represents page sizes, diagrams, generated references, navigation depth, and all 8 locales at 500, 2,500, and 5,000 pages. - Content, theme, API schema, and locale dependencies determine affected builds rather than rebuilding every output for each commit. - Reproducible caches cover dependencies, parsed content, generated API artifacts, images, and search segments with explicit invalidation keys. - Capacity tests must keep p95 full build below 15 minutes and projected compute below $6,000 before the migration receives approval.

Define observability for a 3,600-page docs platform serving 6 products and 5 locales, with 99.9% publishing and search freshness SLAs reported to 15 teams.

I would trace each source commit through build, deployment, index, and user-facing verification. - A deployment record links commit, workflow run, artifact manifest, route count, locale, CDN version, and search-index generation. - Metrics cover queue time, build duration, failure stage, deployment age, route availability, index lag, and preview success by product. - Synthetic checks validate 60 critical pages, locale selectors, version selectors, and search results every five minutes. - Team dashboards and monthly SLO reports show ownership, error-budget consumption, and unresolved platform defects without exposing user data.

Select between Docusaurus and Sphinx for a 2,300-page portal containing React guides, Python SDK reference, and 6 locales, with a $120,000 migration budget and 7-month deadline.

I would run a weighted proof of concept because neither generator dominates this mixed corpus by default. - The scorecard weights Python API extraction, React component support, localization, versioning, search integration, accessibility, contributor skills, and maintenance cost. - Each tool builds the same 120-page slice containing authored guides, generated SDK reference, diagrams, redirects, and two locales. - The proof measures build time, author effort, extension work, output quality, upgrade surface, and five-year ownership cost. - The decision record selects the highest evidence-backed score and keeps migration scope within $120,000 and seven months.

Technical Writer interview questions

100 real questions with model answers and explanations for Senior Technical Writer candidates.

See a Technical Writer resume example →

Practice with flashcards

Spaced repetition · Hunter Pass

Questions

designinformation-architecture

I would organize the suite around shared customer tasks while preserving product context as structured metadata.

A content inventory records owner, audience, task, product, lifecycle state, traffic, and duplicate cluster for all 2,400 pages.
The target map uses task-based landing pages, product filters, and one canonical page when workflows are identical across products.
An information-architecture decision record defines URL, breadcrumb, navigation, and cross-product linking rules before migration starts.
I pilot the model on the 200 highest-traffic pages and require 80% task-findability before scheduling the remaining waves.

Why interviewers ask this: The interviewer is evaluating whether the candidate can turn a fragmented product suite into a testable, metadata-driven information architecture.

discoverycontent-modelingcontentful

I would model durable content semantics separately from either delivery channel.

Core types include concept, task, reference, troubleshooting, release note, and reusable notice with required audience, product, version, and locale fields.
Relationships connect prerequisites, next steps, API objects, and product variants without copying whole pages into Contentful entries.
A schema registry names field owners, validation rules, allowed values, and backward-compatible change procedures for web and in-product consumers.
I validate the model against 120 representative pages and reject it if more than 10% need unstructured escape fields.

Why interviewers ask this: The interviewer is checking content-model design across products, personas, locales, and multiple publishing channels.

content-taxonomydesignsessions

I would replace local labels with a governed faceted taxonomy derived from user language and search evidence.

Query logs, support terms, and card sorts define controlled values for product, task, language, framework, API version, and experience level.
A synonym map redirects legacy and colloquial terms while one preferred label appears in navigation, metadata, and search results.
Taxonomy owners review proposed terms through a change request that records definition, scope, aliases, and affected pages.
I backfill the top 1,000 pages first and compare zero-result rate, refinement use, and successful sessions against the 58% baseline.

Why interviewers ask this: The interviewer is evaluating taxonomy as a governed retrieval system tied to a measurable search outcome.

trackingdesignownership

I would give product teams accountable ownership and reserve central review for shared or high-risk content.

CODEOWNERS maps directories to an engineering owner and a documentation steward, with no page allowed to have an unowned state.
A risk matrix routes authentication, billing, security, and cross-product pages to writers while routine release updates use automated checks and peer review.
Each team receives freshness, broken-link, and overdue-review reports for its own pages rather than a central cleanup queue.
The operating agreement sets a 5-day central SLA, a 2-day team response SLA, and quarterly ownership audits for all 2,600 pages.

Why interviewers ask this: The interviewer is checking whether governance distributes accountability while protecting scarce specialist review capacity.

diataxisarchitectureartifacts

I would use Diátaxis as an intent model, not as four disconnected folders.

A classification rubric distinguishes tutorials, how-to guides, reference, and explanation through user intent, entry state, and completion outcome.
A crosswalk maps every existing page to keep, split, merge, rewrite, or retire and names the target type and owner.
Navigation supports journeys across types, such as tutorial to task to reference, instead of forcing users to understand the framework.
A 150-page pilot tests classification agreement above 85% and task completion before six monthly migration waves begin.

Why interviewers ask this: The interviewer is evaluating practical Diátaxis adoption through classification rules, journeys, and migration evidence.

indexesdesignsessions

I would make product, version, locale, and lifecycle state first-class search signals.

The index stores canonical ID, product, locale, version range, content type, freshness, permissions, and replacement URL for every document.
Query routing applies the current product and locale as boosts, excludes retired pages, and exposes version filters rather than silently guessing.
Relevance tests cover 200 high-value queries with expected top-three results and run before each search configuration release.
Dashboards track latency, zero-result rate, wrong-version clicks, and index freshness against the 99.9% SLA by locale.

Why interviewers ask this: The interviewer is checking search architecture, relevance testing, and operational controls for a large versioned corpus.

migrations

I would migrate by validated content cohorts rather than copy all 14 spaces into a new shell.

The inventory captures owner, traffic, last verification, attachments, restrictions, duplicate groups, and inbound links for all 3,500 pages.
A disposition register assigns keep, merge, rewrite, archive, or delete and requires product-owner approval for high-traffic removals.
Automated converters handle standard pages, while the budget reserves manual work for macros, diagrams, restricted content, and 12,000 redirect mappings.
Ten monthly waves include link-crawl, visual, permission, and owner sign-off gates before each source space becomes read-only.

Why interviewers ask this: The interviewer is evaluating migration planning through disposition, automation boundaries, redirects, budget, and acceptance gates.

design

I would build one task architecture with role-aware entry points and conditional detail only where permissions change the procedure.

A task-to-role matrix identifies the 30 workflows, shared steps, permission differences, and required evidence for each persona.
Global navigation follows product tasks, while persona landing pages curate relevant tasks without creating alternate canonical copies.
Shared procedures use reusable steps and explicit role prerequisites rather than nearly identical operator and auditor pages.
Tree tests with 15 participants per role must reach at least 80% first-click success before the 900-page structure ships.

Why interviewers ask this: The interviewer is checking whether role-based findability can coexist with canonical, reusable content.

content-lifecycledesign

I would make ownership and verification state enforceable publishing data rather than optional labels.

Required fields include accountable team, subject expert, product, version, review interval, last verified date, risk class, and successor page.
Regulatory content receives a 365-day review interval, 60-day advance reminders, and automatic escalation when the 30-day remediation window opens.
Pages without owners enter a 45-day assignment queue and cannot receive feature updates until a product director accepts ownership.
A lifecycle dashboard reports unowned, overdue, expiring, and retired pages by product with weekly exports to Jira.

Why interviewers ask this: The interviewer is evaluating content lifecycle architecture with enforceable ownership and risk-based review controls.

migrations

I would establish one canonical topic map before moving either portal.

A pairwise audit groups pages as unique, equivalent, partially overlapping, or conflicting and records traffic, authority, owner, and locale coverage.
Canonical decisions favor the supported product behavior and strongest task outcome, not automatically the acquiring company's wording.
A merge specification names source sections, retained examples, terminology changes, redirect targets, and reviewers for every combined topic.
Four two-month waves require locale parity, redirect coverage above 99%, and no unresolved product conflict before cutover.

Why interviewers ask this: The interviewer is checking whether acquisition content is reconciled through evidence and explicit canonical decisions rather than bulk transfer.

ci-cddocusaurusgithub-actions

I would separate shared platform code from product content while exposing one predictable contribution workflow.

Versioned content packages hold product docs, and a shared theme package owns navigation primitives, search metadata, accessibility, and locale behavior.
GitHub Actions runs changed-file linting, links, schema checks, and affected builds on pull requests, then the full matrix after merge.
Dependency and content caches key on lockfile, locale, product, and generator version so unrelated changes do not invalidate all 2,800 pages.
Platform SLOs set p95 preview feedback below 10 minutes, production availability at 99.9%, and named owners for each measure.

Why interviewers ask this: The interviewer is evaluating a scalable Docusaurus architecture with clear boundaries, incremental CI, and platform SLOs.

system-designdesignapi

I would build each package independently and assemble versioned outputs through a shared Sphinx configuration.

Package projects own narrative RST or MyST sources and autodoc inputs, while one extension package controls theme, redirects, warnings, and version selectors.
Nightly reference generation compares public symbols with the prior artifact and opens a review issue for additions, removals, or signature changes.
Release workflows build only changed packages, then compose and validate the complete version manifest before publishing.
A 20-minute budget is split across environment setup, package builds, link checks, and deployment with timings retained for 90 days.

Why interviewers ask this: The interviewer is checking modular Sphinx architecture, generated-reference governance, versioning, and bounded release performance.

monorepo

I would use product repositories for release-coupled content and a central repository for shared experience and governance assets.

API changes, examples, and release notes stay beside code so 40 weekly releases can update implementation and documentation atomically.
Shared tutorials, navigation, theme, Vale rules, and content schemas live centrally with versioned packages consumed by product repositories.
A federation manifest supplies canonical URLs, ownership, locale, and build inputs to one portal without copying source files.
An architecture decision record compares contributor access, atomic changes, dependency rollout, discovery, and recovery costs before adoption.

Why interviewers ask this: The interviewer is evaluating repository topology as a trade-off between release coupling and shared platform control.

design

I would build isolated, access-controlled previews from the exact pull-request commit.

Changed products and locales are detected from the dependency graph, while shared-theme changes trigger the complete portal matrix.
Each preview gets an immutable URL, authentication through the company identity provider, and a visible commit, branch, and expiration banner.
GitHub checks post build status, changed-page links, visual-diff artifacts, and the private preview URL within the 8-minute target.
A scheduled job deletes preview storage and deployment records after 14 days and reports orphaned resources against a monthly budget.

Why interviewers ask this: The interviewer is checking preview isolation, permissions, incremental builds, traceability, and resource lifecycle.

deploymentrollbackarchitecture

I would publish immutable site artifacts behind atomic traffic switching.

CI creates a signed manifest containing commit, generator version, locale bundles, checksums, redirects, and search-index version.
Staging validates routes, links, accessibility, headers, and 100 critical journeys against the exact production artifact.
Deployment uploads to a new versioned origin and changes the CDN pointer only after health checks pass in every locale.
The previous two artifacts and indexes remain deployable so rollback completes within 15 minutes without rebuilding.

Why interviewers ask this: The interviewer is evaluating reliable static delivery through immutable artifacts, preproduction gates, atomic cutover, and fast rollback.

ci-cdgithub-actionsquality-gates

I would divide checks by risk and run the cheapest deterministic gates first.

Every change receives schema, Markdown, Vale, internal-link, secret, and ownership checks scoped to affected files.
Code samples, external links, accessibility, and complete locale builds run only when their dependency paths or scheduled full validation require them.
Blocking rules need an owner, documented rationale, fixture tests, measured false-block rate, and a temporary waiver with expiry.
The dashboard tracks p50 and p95 duration, queue time, failures by rule, waivers, and false blocks against the 12-minute and 1% targets.

Why interviewers ask this: The interviewer is checking whether quality gates are fast, risk-based, measurable, and governable rather than an indiscriminate checklist.

design

I would enforce least privilege through repository paths, identity groups, and merge-policy checks.

CODEOWNERS routes requests to product owners for normal paths but cannot itself enforce two separate approvals; for security and billing a required approval-policy check backed by a GitHub App or bot verifies one non-author domain-team approval and one documentation-team approval.
Branch protection blocks direct pushes and requires both the approval-policy check and a CI check that verifies release-artifact signatures and provenance before merge; branch protection does not verify signatures itself.
Contractors join scoped identity groups with automatic 90-day expiry and no access to restricted source or production deployment credentials.
A quarterly access review compares repository, CI, hosting, and secret-store permissions and records exceptions with owners and end dates.

Why interviewers ask this: The interviewer is evaluating permission architecture across content sensitivity, approvals, temporary access, and deployment authority.

migrationsdesignperformance

I would establish a dependency-aware build model and a measured performance budget before importing all content.

A benchmark corpus represents page sizes, diagrams, generated references, navigation depth, and all 8 locales at 500, 2,500, and 5,000 pages.
Content, theme, API schema, and locale dependencies determine affected builds rather than rebuilding every output for each commit.
Reproducible caches cover dependencies, parsed content, generated API artifacts, images, and search segments with explicit invalidation keys.
Capacity tests must keep p95 full build below 15 minutes and projected compute below $6,000 before the migration receives approval.

Why interviewers ask this: The interviewer is checking whether scale and cost are designed through representative benchmarks, dependency graphs, and explicit budgets.

observability

I would trace each source commit through build, deployment, index, and user-facing verification.

A deployment record links commit, workflow run, artifact manifest, route count, locale, CDN version, and search-index generation.
Metrics cover queue time, build duration, failure stage, deployment age, route availability, index lag, and preview success by product.
Synthetic checks validate 60 critical pages, locale selectors, version selectors, and search results every five minutes.
Team dashboards and monthly SLO reports show ownership, error-budget consumption, and unresolved platform defects without exposing user data.

Why interviewers ask this: The interviewer is evaluating end-to-end platform traceability and SLO reporting rather than build logs alone.

docusaurussphinxmigrations

I would run a weighted proof of concept because neither generator dominates this mixed corpus by default.

The scorecard weights Python API extraction, React component support, localization, versioning, search integration, accessibility, contributor skills, and maintenance cost.
Each tool builds the same 120-page slice containing authored guides, generated SDK reference, diagrams, redirects, and two locales.
The proof measures build time, author effort, extension work, output quality, upgrade surface, and five-year ownership cost.
The decision record selects the highest evidence-backed score and keeps migration scope within $120,000 and seven months.

Why interviewers ask this: The interviewer is checking tool selection through a representative prototype and total ownership evidence rather than preference.

Locked questions

21
Design an OpenAPI documentation program for 28 REST services, 1,400 reference pages, and 11 engineering teams, with reference updates published within 24 hours of a schema merge.
restopenapidocumentation
22
Architect AsyncAPI documentation for 16 event domains, 900 pages, and 7 producer teams, including Kafka examples and a 3-month schema-version transition.
documentationschemakafka
23
Create one developer-docs model for 5 SDKs, 1,200 pages, and quarterly releases, with Java, Python, JavaScript, Go, and .NET examples kept within one release of feature parity.
javascriptpython
24
A 2,000-page API portal is projected to be 55% generated from OpenAPI and SDK sources; define the boundary between generated and authored content before a 6-month rebuild.
openapi
25
Design API versioning documentation for 36 services, 1,800 pages, and 4 supported versions, with monthly releases and a 12-month compatibility promise.
versioningapi-versioningdesign
26
Define a deprecation system for 24 public APIs used by 6,000 developers, with 12 months of notice, 3 SDK release trains, and 1,100 affected pages.
content-deprecationsystem-designapi
27
Design code-sample testing for 2,400 snippets across 6 SDKs and 3 supported API versions, with pull-request results under 20 minutes and nightly full coverage.
coveragecode-sample-testingapi
28
Plan a migration guide from API v2 to v3 for 1,300 pages, 4 SDKs, and 2,500 customer applications, with v2 support ending in 9 months.
migration-guidesapimigrations
29
Redesign onboarding across a 750-page developer portal for 3 APIs and 4 SDKs, with a target to reduce median time to first successful call from 42 to 15 minutes in one quarter.
onboardingapi
30
Create release synchronization for 20 services, 5 SDKs, and 1,600 developer pages when engineering ships twice weekly and documentation has a 24-hour publication SLA.
documentation
31
Design a DITA information model for 4,000 pages covering 7 hardware products, 12 variants, and 6 locales in Oxygen, with an 18-month migration timeline.
ditaoxygen-xmlmigrations
32
Configure an Oxygen XML workflow for 2,500 DITA topics, 30 authors, and 8 product teams, with review turnaround under 5 business days and zero invalid topics at publish time.
configditaoxygen-xml
33
Architect a MadCap Flare project for 1,700 help topics, 5 desktop products, 3 release branches, and PDF plus web outputs within a $90,000 annual tooling budget.
madcap-flare
34
Create a component-reuse model for 3,100 topics across 9 products where 28% of text is duplicated and 6 teams need independent release schedules.
components
35
Design conditional publishing for 2,200 topics, 8 product editions, 4 operating systems, and 6 locales, with no unsupported instruction reaching customers.
system-designdesign
36
Design localization architecture for 4,600 topics, 8 locales, and monthly releases, with translation available 10 business days after English and a $420,000 annual budget.
designlocalization
37
Govern terminology across 3,800 pages, 10 products, and 7 locales after an audit finds 420 competing terms for 140 concepts; deliver a controlled glossary in 4 months.
terminology-managementforms
38
Define diagram source governance for 650 diagrams embedded across 2,900 pages and 6 products, with editable sources required within a 3-day update SLA.
39
Design media governance for 1,200 screenshots and 180 videos used across 3,400 pages, 5 products, and 6 locales, with a $75,000 annual production budget.
design
40
Set accessibility architecture for 5,000 pages, 900 diagrams, and 220 videos across 7 locales, targeting WCAG 2.2 AA within 12 months.
documentation-accessibilitywcaga11y
41
Create a style guide for 2,700 pages written by 85 contributors across 9 teams and 6 locales, with a 6-month rollout and no more than 20 mandatory rules.
style-guides
42
Define documentation completion criteria for 14 product teams shipping 35 features per quarter into a 3,200-page suite, with release approval required 2 days before launch.
documentation
43
Design review SLAs for 2,100 pages and 12 teams when writers receive 160 pull requests per month, including 48-hour standard and 4-hour launch-blocking targets.
code-reviewdesignpull-requests
44
Build a documentation quality score for 4,200 pages across 8 products, combining analytics and audits without letting high traffic hide inaccurate content; deliver the first baseline in 90 days.
documentation
45
Integrate documentation into the roadmaps of 10 product teams managing 2,800 pages and 60 planned releases over 2 quarters, with only 6 writers available.
roadmapdocumentation
46
Create a content deprecation policy for 3,600 pages across 9 products, with 540 legacy pages scheduled for retirement over 6 months and redirect coverage above 99%.
content-deprecationcoverage
47
A support organization handles 18,000 monthly cases linked to a 2,500-page knowledge base; design a 6-month documentation program targeting 15% verified support deflection across 5 products.
documentationdesign
48
Design a mentorship program for 6 writers supporting 3,000 pages and 11 teams, with 2 new hires expected to own a product area within 90 days.
mentoringdesign
49
A proposed documentation-platform redesign costs $360,000 for 12 teams maintaining 4,800 pages; Finance requires an 18-month payback, while manual release, localization, and support rework consume 6,400 staff hours annually. Build the investment case and phased architecture.
documentationlocalizationarchitecture
50
Establish governance for a 5,000-page documentation estate spanning 12 products, 20 teams, and 8 locales, with a quarterly council and a $900,000 annual program budget.
documentation
51
An API release renamed authentication headers and made 240 pages stale, while 18 customers opened 63 failed-login tickets in 2 hours; do you roll back the API, patch only the top pages, or stop the rollout?
authapirollback
52
A database migration guide tells 320 enterprise admins to run steps in the wrong order, 27 have started, and 6 environments are unavailable; do you retract the guide, issue an inline correction, or let Support handle each case?
databasemigrationsmigration-guides
53
Webhook documentation still shows a 10-second retry schedule after the product changed to exponential retries over 24 hours, causing 41 partners to create duplicate orders; do you revert the docs, revert the behavior, or document the new contract?
documentationwebhooksresilience
54
The Python SDK release changed create_user to create_users, but 96 code samples and 14 tutorials still use the old call and package downloads reached 38,000; do you yank the package, add an alias, or rewrite the docs only?
tutorialspython
55
Product wants to ship in 3 hours, Engineering changed 7 workflows across 180 pages this morning, and only 45 pages have been reviewed; do you delay release, publish partial docs, or waive review?
56
A beta OAuth guide told 74 teams to put a client secret in browser apps, and 11 of those secrets appeared in public repositories; do you edit quietly, pull the beta, or rotate every credential?
oauthsecrets
57
An API enum started with 4 values and gained 12 more, for 16 total; generated reference lists all 16 correctly, but 68 hand-written pages still allow only the original 4. Do you delete the prose, regenerate it, or keep both?
reference-docsapi
58
Rate-limit documentation says 1,000 requests per minute, production now enforces 300, and 9 large customers saw 429 rates above 22%; do you restore the old limit, announce the new one, or create temporary exceptions?
documentationerror-handling
59
A CLI release removed 3 flags used in 46 automation guides, breaking 1,700 nightly jobs before the release note was published; do you restore the flags, fast-track docs, or ask users to rewrite scripts?
60
Release 9.4 was rolled back after 52 minutes, but search and cached pages still direct 26,000 daily visitors to 130 version 9.4 procedures; do you purge all release content, keep history, or redirect to 9.3?
caching
61
The documentation portal has returned 503 for 47 minutes during a product launch, affecting 82,000 visits; do you fail over to a static mirror, wait for the vendor, or send users to cached search results?
documentationcachingprocurement
62
Docs build time rose from 9 to 46 minutes after 24,000 generated API pages were added, and 31% of pull requests are merged without preview; do you buy larger runners, split the site, or reduce generation?
code-reviewapipull-requests
63
After a search reindex, zero-result queries rose from 7% to 29% across 1.4 million searches and API reference clicks fell 38%; do you restore the old index, tune ranking live, or wait for fresh crawling?
indexesqueriesapi
64
A URL restructure created 18,600 broken inbound links, organic visits dropped 44%, and 2,300 support articles still use old paths; do you revert the structure, add redirects, or update links gradually?
65
A permissions change exposed 6 internal runbooks to 3,400 external users for 19 minutes, including private hostnames but no credentials; do you take the portal offline, hide the pages, or rotate infrastructure details?
runbooks
66
The OpenAPI generator failed silently and published reference for only 312 of 740 endpoints, while the hand-written guides still link to missing operations; do you revert, publish partial reference, or restore from the prior contract?
endpointsopenapigenerators
67
A monorepo migration moved 9,200 pages and 37 packages, but 16% of cross-package links fail and writer previews take 28 minutes; do you roll back, freeze edits, or repair in place?
monoreporollbackmigrations
68
A compromised documentation theme dependency injected a cryptominer into 4% of sessions for 3 hours, and the package has 11 transitive dependencies; do you patch the theme, revert the site, or disable all scripts?
documentationdependenciessessions
69
Preview environments fail to provision for 43% of 86 documentation pull requests because the hosting quota is exhausted; do you remove previews, increase quota, or share one staging site?
code-reviewdocumentationpull-requests
70
One CDN region serves documentation 3 releases behind to 17% of Asia-Pacific traffic, producing 290 wrong-procedure sessions; do you disable that region, purge globally, or wait for cache expiry?
documentationcachingsessions
71
An audit finds 1,480 duplicate pages across 6 product areas, and contradictory setup pages account for 31% of onboarding tickets; do you merge by traffic, delete older pages, or assign teams to clean their own areas?
onboarding
72
A link-graph review finds 620 orphaned pages, 190 still receive direct traffic, and 74 contain current compliance instructions; do you delete all orphans, add them to navigation, or archive them?
73
Forty percent of 3,600 pages have no active owner, 280 are tied to services being changed this quarter, and Engineering refuses blanket ownership; do you freeze those pages, assign writers, or escalate to directors?
escalationownership
74
Six months after a style guide launch, only 3 of 14 teams use it, Vale reports 12,400 violations, and writers spend 90 hours monthly debating style; do you enforce every rule, retrain teams, or simplify the guide?
style-guidesvale
75
The documentation review backlog reaches 430 pull requests, median age is 19 days, and 38 product releases are waiting; do you lower the review bar, add contractors, or triage by risk?
backlogdocumentationcode-review
76
A team met its target of 200 pages per quarter by splitting 48 guides into 236 thin pages, while task completion fell from 78% to 61%; do you keep the output target, reverse the split, or replace the metric?
monitoring
77
Three acquisitions introduced 11 competing terms for the same account object across 2,700 pages, and search reformulation reaches 34%; do you mandate one term immediately, preserve product brands, or build aliases?
78
Sales submits 96 urgent documentation requests in one month, but analytics show only 8 affect active deals and 22 duplicate existing pages; do you honor urgency, reject the queue, or impose an intake model?
documentationdata-structures
79
A cleanup project archived 900 low-traffic pages and cut the freshness denominator enough to raise the score from 72% to 94%, but Support still uses 140 archived pages; do you celebrate, restore pages, or redefine freshness?
80
Two subject-matter experts hold 67 reviews, median response is 16 days, and releases bypass documentation after day 5; do you add more writers, remove expert approval, or redesign review ownership?
documentationownership
81
A DITA conditional-publishing error exposes enterprise-only setup steps in 14 locales, and 6,800 community users see commands that fail; do you pull all locales, patch conditions, or publish English only?
dita
82
A single-source project reduced 1,900 topics to 620, but 23% of reused warnings are wrong for at least one of 8 products; do you undo reuse, add conditions, or allow product forks?
83
A DITA key-space refactor breaks 4,300 cross-references in 9 localized manuals two weeks before manufacturing; do you revert the keys, repair links manually, or postpone the manuals?
reference-docsrefactoringdita
84
A locale fallback bug serves 7-month-old English instructions on missing strings across 12 locales, affecting 18% of update pages; do you disable fallback, block every locale, or show mixed-language current content?
85
A redesign invalidates 760 screenshots across 20 locales, but replacing them all delays launch by 3 weeks; do you ship stale images, remove screenshots, or delay the product?
86
An architecture diagram used in 6 locales still shows a retired queue, and 14 customer designs copied it before an engineer noticed; do you delete diagrams, issue a correction, or restore the old architecture?
designdata-structures
87
An accessibility audit across 8 locales flags 1,240 images for alternative-text review and finds 310 keyboard-inaccessible code tabs; do you block releases, auto-generate alt text, or remediate by traffic?
documentation-accessibilityalt-textkeyboard-accessibility
88
A conditional PDF build omits mandatory safety warnings in 5 of 16 locales because one variable is translated differently; do you stop all shipments, ship unaffected locales, or insert an English warning?
89
A translation-memory import propagates an incorrect destructive-command warning to 3,600 segments in 10 locales, and 84 pages are already live; do you roll back the memory, patch live pages, or pause localization?
localizationmemoryrollback
90
Web, PDF, and in-product help are generated from one source, but 22% of 1,100 topics render different prerequisites across 6 locales; do you split the sources, normalize conditions, or accept channel differences?
normalization
91
The product roadmap funds 4 growth features but no work for 520 stale enterprise pages that generate 38% of support escalations; do you accept the roadmap, demand a feature delay, or propose a bounded remediation?
roadmapescalation
92
An external audit is due in 12 days, 46 of 180 controlled procedures lack approval evidence, and Operations wants writers to backdate signatures; do you delay the audit, reconstruct evidence, or accept the request?
forms
93
A documentation vendor will end support in 90 days, migration covers 28,000 topics and 16 locales, and the replacement loses DITA keyrefs; do you renew for one year, migrate now, or run both tools?
documentationmigrationsprocurement
94
Engineering rejects a documentation release gate because it added 11 minutes to 260 weekly builds, but 7 of the last 10 incidents involved undocumented breaking changes; do you remove it, keep it blocking, or redesign it?
incidentsdocumentationversioning
95
Product, Support, and Documentation maintain separate launch calendars, and 17 of 32 releases last quarter reached customers before documentation or support readiness, generating 410 avoidable tickets; do you centralize ownership, add status meetings, or create one readiness model?
documentationownershiphealth-checks
96
A junior writer publishes an authentication example with an admin token after 2 reviewers miss it, and the page receives 9,000 views; do you remove the writer from security work, treat it as a team failure, or focus only on token rotation?
authtokens
97
A mid-level writer's reviews average 96 comments and 4 rounds across 12 pull requests, delaying releases by 8 days; do you reduce standards, reassign their work, or mentor for earlier alignment?
mentoringcode-reviewpull-requests
98
Four engineers contribute 70% of internal docs, but their 210 pages have a 17% reader-failure rate and they reject editorial review as too slow; do you take authorship away, mandate review, or coach them?
99
Legal requires a 900-word disclaimer on 46 onboarding pages, Product says completion will fall 12%, and the audit date is in 9 days; do you accept the text, hide it behind a link, or propose layered disclosure?
onboarding
100
A new writer closes 55 tickets in a month but reopens 18 after weak evidence, while another writer closes 24 with zero reopenings; do you reward volume, slow the first writer, or change team coaching and metrics?
monitoring

Questions

1A suite of 6 products has 2,400 pages split across 11 team-owned sites, and customers cannot tell which product owns a task; design the information architecture for a 9-month consolidation.

2Define a content model for 1,800 pages covering 4 cloud products, 3 personas, and 6 locales, with Contentful entries feeding web and in-product help within 12 months.

3A 3,200-page developer portal uses 700 inconsistent labels across 8 teams; design a taxonomy that must improve successful search sessions from 58% to 75% in two quarters.

4Twelve engineering teams own 2,600 pages, but a central team of 4 writers has a 5-day review SLA; design a federated ownership model without making the writers a bottleneck.

5Reorganize 1,500 pages for 5 products into a Diátaxis model in 6 months while 9 teams continue shipping weekly; what architecture and transition artifacts do you define?

6Search across 4,800 pages in 8 products and 7 locales returns the wrong version in 18% of sessions; design the index, metadata, and relevance controls for a 99.9% search SLA.

7Plan the migration of 3,500 Confluence pages from 14 spaces to one developer portal within 10 months and a $180,000 budget, while preserving 12,000 inbound links.

8A 900-page administration product serves operators, auditors, and developers across 5 teams; design navigation that supports 30 recurring tasks without duplicating role-specific pages.

9Design lifecycle metadata for 2,200 pages across 7 products where 22% have no owner and regulatory pages require annual verification with a 30-day remediation SLA.

10After an acquisition, merge a 1,100-page portal and a 1,700-page portal covering overlapping products in 4 locales within 8 months; define the canonicalization and migration plan.

11Architect a Docusaurus platform for 2,800 pages, 10 products, 6 locales, and 14 contributing teams using GitHub Actions, with pull-request feedback under 10 minutes.

12Design a Sphinx system for 1,600 Python-library pages across 12 packages and 3 supported versions, with API reference generated nightly and releases published in 20 minutes.

13Choose a monorepo or multirepo model for 3,000 pages owned by 18 engineering teams and a 5-person docs platform group, with 40 product releases per week.

14Design pull-request previews for a 2,100-page, 6-locale portal where 120 contributors need private links within 8 minutes and previews must expire after 14 days.

15Define the deploy architecture for a 4,500-page portal serving 9 products in 7 locales, with a 99.95% availability SLA and a 15-minute rollback objective.

16Specify GitHub Actions quality gates for 2,700 pages maintained by 16 teams, with a 12-minute pull-request budget and fewer than 1% false blocks per month.

17Design repository permissions for 1,900 pages across 13 teams, including security and billing docs that require two approvals and contractor access that expires after 90 days.

18A planned platform must build 5,000 pages across 8 locales in under 15 minutes on a fixed $6,000 monthly CI budget; design for build performance before migration.

19Define observability for a 3,600-page docs platform serving 6 products and 5 locales, with 99.9% publishing and search freshness SLAs reported to 15 teams.

20Select between Docusaurus and Sphinx for a 2,300-page portal containing React guides, Python SDK reference, and 6 locales, with a $120,000 migration budget and 7-month deadline.

Locked questions