The L&D caption programme annual review: what to audit, when to schedule it, and how to present results to leadership

Why the annual review exists — and why it gets skipped

A caption programme is not a project with a completion date. It is a system — a policy, a vendor relationship, a quality control workflow, a glossary, a set of platform delivery configurations, and a documentation trail — that must remain aligned with a moving target: the organisation's content library, its compliance obligations, its technology stack, and its budget. Each of those four variables changes over 12 months in ways that can quietly break the alignment the programme achieved in year 1.

Content library changes. The library grows, usually faster than the caption workflow was designed for. New content is added — some captioned at the point of publish, some not. Existing content is revised — narration updated, new slides added, product names changed — and the associated caption file is rarely updated at the same time. Some content is retired or migrated to a new system. The content audit from the initial programme build is stale within 12 months. Without an annual count, no one knows the actual captioning rate for the current library. The common assumption — "we have a caption-before-publish gate, so everything new is captioned" — is almost always wrong, because exceptions accumulate silently. The engineering team uploads a product demo without knowing the gate applies to them. An HR partner adds a benefits orientation video via a different upload pathway. The manager who joined in March creates onboarding materials in Loom and shares them in Slack. None of these appear in the caption workflow until someone runs the inventory.

Regulatory and enforcement landscape changes. ADA Title II enforcement for state and local government entities and public universities began in April 2026 — organisations that were planning for future compliance now have active obligations. The European Accessibility Act has been enforceable for private sector companies operating in EU member states since June 2025. Several US states have passed or amended digital accessibility laws in the past 12 months. WCAG 2.2 published in October 2023 and the compliance landscape is gradually shifting from WCAG 2.1 as the reference standard. A governance policy written in year 1 that references specific regulatory frameworks may no longer accurately reflect the organisation's current obligations.

Vendor relationship drift. Vendors maintain their service quality most rigorously during the pilot and early production phase when the relationship is new and the account is at risk. Over time, accuracy can drift. Turnaround times can slip. The dedicated account manager from year 1 may have been replaced. The pricing schedule that was negotiated during the RFP may no longer reflect the vendor's current rates — either because market rates have changed or because the volume discount threshold was passed but never formalised into a contract amendment. A programme that is not actively managing the vendor relationship on an annual basis is a programme that is passively accepting whatever service level the vendor delivers.

Technology and platform changes. LMS platforms release updates that change how captions are delivered. A Cornerstone upgrade may reset the caption player configuration. A TalentLMS migration from an older plan to a newer one may change which caption formats are supported. An organisation that added Microsoft Stream for Teams recordings mid-year may not have connected Stream to the caption workflow at all. Platform changes that happen outside the caption programme's explicit view are the most common source of silent delivery failures — the programme is doing everything right, but the LMS is no longer serving the caption file to learners.

Budget misalignment. Year-1 budget assumptions are based on year-1 volume and year-1 vendor pricing. If production volume doubled in year 2, the subscription model that was cost-effective in year 1 may now be significantly more expensive per minute than a pay-per-minute contract. If production volume dropped, the reverse may be true. The correction labour assumption from year 1 may also be wrong: if the glossary has matured and accuracy has improved, correction labour should have declined — but if that savings is not being tracked, the budget will not reflect it and the efficiency gain will be invisible.

Why the review gets skipped. The annual review is not mandated by any external regulatory framework. There is no external audit body that requires a specific annual programme review format or schedule. The governance policy may reference it, but the organisation's own calendar is the only enforcement mechanism. The person who built the programme may have been promoted or moved to a different role. The institutional knowledge of what the programme is supposed to review annually lives in the programme builder's head, not in a documented process. The result is that the annual review gets scheduled in Q1, deferred to Q2 when the budget cycle competes, deferred again to Q3 when a product launch occupies the L&D team, and quietly dropped from the calendar when Q4 sprint work starts. The following year, the same conversation happens — "we should do an annual review" — with no prior-year baseline to compare against. This guide provides the structure that makes the annual review a repeatable, calendar-driven event rather than a discretionary effort.

When to schedule the annual review

The default timing for the annual review is Q1 of the organisation's fiscal year, approximately four to six weeks after fiscal year close. This timing works because the vendor's annual performance data is complete for the year just ended; the Q4 DCMP spot-check results are available; the budget cycle for the coming year is either open or about to open, making the next-year forecast actionable rather than theoretical; and the regulatory landscape for the prior calendar year has stabilised sufficiently for legal to provide a current-year analysis.

Why four to six weeks after fiscal close, not immediately. Running the annual review in the first two weeks of the new fiscal year means the vendor's annual performance report is likely not yet prepared, the finance team's actual-vs-budget reconciliation for the prior year may not be final, and the LMS administrator's content count may not reflect any year-end imports. Four to six weeks gives all data contributors time to prepare their inputs without the review becoming a planning exercise using incomplete data.

Calendar event versus working group. The annual review should be a standing calendar event, not a one-time meeting. Block it as a recurring annual event at the time the programme governance policy is signed — not after the programme is running and the schedule gets forgotten. The initial programme build post recommends scheduling the annual review as part of the programme launch documentation. If your programme was built without this, add it now: put a recurring Q1 calendar block for a two-hour review meeting and a pre-meeting data-collection window of two weeks before the meeting.

Triggers for early review. The annual schedule is the default. Several events should trigger a programme review outside the annual schedule:

• Vendor change or vendor merger. If the captioning vendor is acquired, merges with another company, or is replaced, a full vendor performance review should be conducted immediately — not deferred to the next annual review. Vendor transitions are the highest-risk moment for accuracy drift and SLA degradation. The vendor transition playbook covers the transition process; the annual review data serves as the baseline for evaluating the new vendor's performance.
• LMS migration. Any LMS migration should trigger a platform delivery audit immediately after go-live. Migrations consistently lose caption sidecar files without warning. The LMS audit methodology covers the post-migration verification process.
• Significant regulatory event. A new law, an enforcement action against a peer organisation, an OCR resolution agreement that changes the interpretation of existing obligations, or a court decision that expands accessibility obligations should trigger a regulatory refresh outside the annual schedule. The refresh should be conducted in collaboration with legal, not by the L&D team alone.
• Significant accuracy failure. If a DCMP spot-check returns accuracy below 95% for any content category, or if a learner or compliance audit identifies a material accuracy failure, treat it as an immediate programme incident — not a finding to queue for the next annual review. The caption QA methodology post covers the incident response process for accuracy failures.
• Organisational acquisition or restructure. If the organisation acquires another company or undergoes a restructure that changes the content scope, the content library audit component of the annual review should be run immediately to account for acquired content.
• Major product launch with new vocabulary. A product launch that introduces significant new technical terminology — new product names, new SDK symbols, new regulatory acronyms — should trigger an immediate glossary update, even outside the annual review cycle. If the glossary is not updated before the first training videos about the new product are captioned, those videos will have accuracy failures on exactly the terminology that matters most.

Who owns the schedule. The accessibility coordinator owns the review scheduling and logistics. The L&D director or VP of Learning sponsors the review and owns the leadership presentation output. The vendor manager, LMS administrator, and finance business partner are required data contributors. Legal or compliance should participate in the regulatory refresh section, at minimum as reviewers of the draft analysis before the meeting.

Component 1: Content library audit

The content library audit answers three questions: What new content was added in the past 12 months and what percentage is captioned? What existing content was revised and were the captions updated to match? What content was decommissioned and is it properly removed from accessible delivery? A fourth task — the accuracy spot-check — validates that the captioned content meets the 99% WCAG threshold, not just that caption files exist.

New content inventory

Pull the video count from the LMS using a date-added filter for the past 12 months. Cross-reference with the caption delivery log to identify uncaptioned items. The caption delivery log should be maintained by the vendor or the LMS administrator as a record of every caption job submitted and delivered during the year. If no delivery log exists, this is itself a finding — establish the log as an action item before the next annual review.

Calculate: new content total, new content captioned, new content not yet captioned, new content captioning rate. The target is 100% captioned before publish — but the actual rate in most organisations is lower because of exceptions that accumulated without being caught. The audit identifies the uncaptioned items; the action plan addresses remediation. Prioritise uncaptioned items by audience (mandatory compliance training first) and by regulatory obligation (content accessed by employees with documented accommodation requests first).

Identify the categories of uncaptioned content. Common patterns: content published during a product launch sprint before the caption gate was reinforced; content from a department that was not included in the original caption workflow training; content imported from a third-party vendor (purchased compliance training, vendor-produced onboarding materials) without a caption delivery requirement in the contract. Each category has a different remediation path. The third-party compliance training captioning post covers the purchased content remediation approach.

Content revision audit

Content revisions are the hidden compliance gap in mature caption programmes. An L&D team might record a new version of a product training module — updated narration to reflect a renamed feature, new slides added, a regulatory update incorporated into a compliance module — but not update the associated caption file. The old caption file remains attached to the new video, with accuracy errors wherever the script changed. The programme's captioning rate shows 100% coverage, but the coverage is inaccurate: the caption file transcribes audio that no longer matches the video.

How to identify: pull LMS records for all videos where the video file modified date is more recent than the associated caption file date. This requires the LMS to store both dates, which varies by platform. For platforms that do not expose this data directly, use the LMS audit log (most enterprise LMS platforms log content updates) and cross-reference against the caption delivery log date. Any video updated after its caption was delivered is a candidate for caption review.

How to handle revisions: a caption file is a transcription of the audio at a specific point in time. If the audio changes — even a single line of narration — the caption file must be updated for that segment. Partial re-captioning (only re-captioning the changed segment) is technically correct but in practice more expensive per-word than re-captioning the full video, because the vendor must perform a comparison between old and new audio to identify the changed segments. For short content revisions (one or two lines changed), partial re-captioning is worthwhile. For substantive revisions (more than 10% of content changed), re-captioning the full video is typically more cost-effective and produces a cleaner, more accurately timed caption file.

Decommissioned content

Content retired from the primary LMS may still be accessible in shared drives, legacy learning management systems, external partner portals, or archived SharePoint sites. The annual content audit should include a check of all access points for decommissioned content, not just the primary LMS. The audit should confirm that decommissioned content has either: been removed from all access points, or has been explicitly documented as out-of-scope with a legal rationale (for example, legacy content that pre-dates the compliance obligation).

For public entities and organisations with ADA Title II obligations, decommissioning a video does not retroactively satisfy any prior accessibility obligation that applied when the video was in active distribution. If a learner filed an accommodation request for a video while it was in distribution and the organisation failed to provide captioned access, that violation persists even after the video is decommissioned. The annual review should check whether any decommissioned content has pending accommodation requests or open complaint records associated with it.

Accuracy spot-check

The DCMP protocol accuracy check should be run annually against a stratified random sample of the captioned content library. Running it only at the point of vendor onboarding tells you how the vendor performs under scrutiny during the sales process; running it annually tells you how the vendor performs in steady-state production when there is no selection-pressure incentive to over-perform.

Sample size guidance: 3–5% of the total captioned library, or a minimum of 20 videos, whichever produces the larger number. For a library of 200 captioned videos, 3% produces 6 videos — the minimum floor of 20 is the binding constraint. For a library of 2,000 captioned videos, 3% produces 60 videos — the percentage threshold is the binding constraint. The goal is statistical coverage of the library, not exhaustive review.

Stratify the sample by: vendor (if multiple vendors or vendor tiers are in use), content category (compliance training, product training, technical training, onboarding, soft skills), and content vintage (year 1 vs. year 2 production). Vintage stratification is particularly important because it distinguishes between content produced when the glossary was first built (typically lower accuracy for proprietary vocabulary) and content produced after 12 months of glossary maturation. If year-2 accuracy is lower than year-1 accuracy, investigate: the vendor's service quality may have declined, the programme's QC rigour may have relaxed, or new content categories were added that the glossary does not yet cover.

The DCMP accuracy protocol for a 10-minute segment: identify a 10-minute segment from the captioned video. Transcribe the audio manually (word-for-word, not cleaned up). Count all captioning errors: omissions (words in the audio not present in the caption), substitutions (wrong word in the caption), insertions (words in the caption not in the audio), and typographical errors (misspellings or word-boundary errors). Note: for WCAG 99% accuracy threshold purposes, only substitution errors and omissions that change meaning count toward the error rate. Punctuation errors and typographical errors that do not affect meaning are measured separately under DCMP but do not count against the 99% WCAG threshold. Calculate: (total words in audio − qualifying errors) ÷ total words in audio. The result must be 0.99 or higher to meet the WCAG 2.1 AA threshold.

Document spot-check results in the programme accuracy log by video ID, content category, vendor, and date. This log is primary evidence for any regulatory audit or litigation and should be maintained as a permanent programme record. The caption QA methodology post has a sample accuracy log template.

Component 2: Vendor performance review

The vendor performance review covers three areas: accuracy performance against the 99% WCAG threshold, SLA adherence on turnaround time, and pricing and contract status. It uses the year's full production record, not just the annual spot-check sample. The goal is to identify whether the vendor's steady-state performance over 12 months of full production volume meets the programme's contractual and compliance requirements — not just whether they can perform under review-condition scrutiny.

Accuracy performance

Request the vendor's annual accuracy summary report before the annual review meeting. A reputable captioning vendor should be able to provide, for the full year's production: total jobs completed by content category, total minutes captioned by content category, accuracy rates by content category (based on their own internal QC), and a breakdown of correction requests and re-submissions by root cause.

Compare the vendor's self-reported accuracy against the independently conducted DCMP spot-check results from Component 1. If the vendor's self-reported accuracy is 99.2% and the independent DCMP spot-check finds 97.8% on a stratified sample, the discrepancy is itself a finding. Possible explanations: the vendor's internal QC protocol is less rigorous than the DCMP protocol; the vendor is selectively sampling easy content for their internal QC; the vendor's accuracy has improved in the categories they sample but has declined in categories they do not monitor closely. Any discrepancy greater than 0.5 percentage points between vendor-reported and independently measured accuracy should be raised in the vendor performance review meeting and documented in the annual review report.

Identify accuracy trends by content category. It is common for captioning vendors to maintain high accuracy on compliance training and soft-skills content (clear audio, standard vocabulary) while accuracy declines for engineering training or medical content (technical vocabulary, multiple speakers, domain-specific terminology). If the programme includes high-technical-vocabulary content, the glossary-biased captioning approach — where the caption engine is trained on the organisation's specific term list — is the primary defence against this accuracy gradient. Review whether the glossary coverage (Component 6) correlates with the accuracy category distribution.

SLA adherence

Pull the submission log and delivery timestamp for every job submitted during the year. Calculate: total jobs, jobs delivered within SLA window, jobs delivered outside SLA window, and average lateness for jobs that missed SLA. The calculation should use the contractual SLA window (e.g., 48 business hours), not the vendor's average turnaround time. Average turnaround time is not the same as SLA adherence — a vendor who delivers 80% of jobs in 24 hours and 20% of jobs in 96 hours (outside a 48-hour SLA) may report an excellent average turnaround while having a material SLA adherence failure.

Identify SLA breach patterns. Common patterns: SLA misses cluster around the vendor's peak production periods (Q1 compliance training season, Q4 year-end content pushes), or SLA misses are concentrated in specific content categories (very long videos, multi-speaker audio, or non-standard audio quality). If SLA misses cluster around the programme's own peak submission periods, address this in contract renegotiation: negotiate priority processing for time-sensitive content categories in exchange for advance submission scheduling for predictable peak periods.

Calculate SLA breach remediation owed. If the vendor contract includes SLA breach remediation (credit, refund, or service-level rebate), document every qualifying breach and calculate the remediation amount. Many organisations fail to claim SLA breach remediation because they do not track it systematically — the annual review is the point at which this is calculated and the claim is initiated with the vendor. The vendor SLA and contract review checklist has a breach remediation calculation template.

Pricing and contract review

Request the vendor's current pricing schedule and compare against: the pricing in the active contract, the market rate from the most recent competitive RFP (if available), and the pricing you would receive today if you ran a new vendor RFP for current production volume. Vendors typically provide introductory pricing that is below their standard rate, with the expectation that the rate will normalise at renewal. Pricing creep between year 1 and year 3 of a vendor relationship is common.

Volume threshold review: if the vendor contract includes volume discount thresholds, calculate total production volume for the year against the threshold levels. If the programme crossed a threshold during the year but did not invoke the corresponding discount (because the threshold was not monitored), negotiate a retroactive credit and amend the contract to include an automatic monitoring mechanism. If the programme is approaching a threshold and the back-catalogue plan will push volume above it in the coming year, negotiate the threshold discount into the contract before the volume reaches the threshold.

Contract renewal date: note the contract renewal date in the annual review documentation. If the contract renews within 90 days of the annual review, initiate renewal discussions immediately. Renewal negotiations that start 30 days before expiry are conducted under time pressure that favours the vendor. Renewal negotiations that start 90 days before expiry allow for a competitive RFP process if the initial renewal terms are not satisfactory. The captioning vendor RFP playbook covers the full 14-question scoring framework for use in renewal RFPs.

Component 3: Regulatory and standards refresh

The regulatory and standards refresh confirms that the programme's governance policy, compliance claims, and documentation accurately reflect the current legal and technical standard landscape. Regulatory obligations do not wait for annual reviews to change — but the annual review is the structured moment at which the programme's documentation catches up with any changes that occurred during the year.

WCAG version landscape

WCAG 2.1, published June 2018, remains the primary reference standard for ADA, Section 508 (2017 refresh), and EAA compliance obligations. WCAG 2.2, published October 2023, added nine new success criteria. None of the new WCAG 2.2 criteria directly modify the caption requirements (SC 1.2.2 and SC 1.2.4 in WCAG 2.1 are unchanged in WCAG 2.2). However, two WCAG 2.2 criteria — SC 2.4.11 (Focus Not Obscured, Level AA) and SC 2.5.7 (Dragging Movements, Level AA) — affect caption player user interface elements and should be reviewed if the programme has a custom caption player implementation.

WCAG 3.0 is in active development but has no enforcement date. The current compliance target for virtually all L&D programmes with a regulatory compliance obligation remains WCAG 2.1 Level AA. If your governance policy specifies "WCAG 2.1 AA" as the technical standard, no update is required for WCAG 3.0 — it is not yet an enforceable standard. If your policy specifies "the most current version of WCAG Level AA," review whether WCAG 2.2 introduces any requirements not previously in scope for your implementation.

US federal regulatory landscape

ADA Title II: The Department of Justice issued final rules in April 2024 requiring WCAG 2.1 AA compliance for web and mobile content of state and local government entities and public universities. Enforcement began in April 2026. If your organisation is a covered entity under ADA Title II, the annual review should confirm that all training content accessible by the public or by employees of a covered entity meets WCAG 2.1 AA captioning requirements for pre-recorded video.

ADA Title I: No final WCAG regulation applies to private employers under ADA Title I as of 2026. However, EEOC guidance, court decisions interpreting Title I reasonable accommodation obligations for digital content, and the 2026 enforcement environment have consistently referenced WCAG 2.1 AA as the de facto technical standard. Review whether any court decisions or EEOC guidance issued in the prior 12 months affect your organisation's Title I caption obligations. This review should be conducted with legal counsel, not independently by the L&D team.

Section 508: the 2017 Revised Section 508 Standards incorporate WCAG 2.1 Level AA for electronic and information technology. No significant changes to the section 508 standards occurred in 2025 or 2026. Federal contractors and federal agencies should confirm that any new ICT procurement during the year included the appropriate Section 508 conformance requirements in the vendor contract.

OCR investigations: the Department of Education Office for Civil Rights (OCR) is the primary enforcement body for accessibility complaints at educational institutions. OCR resolution agreements from prior years set precedent that affects the interpretation of compliance obligations at peer institutions. Review any OCR resolution agreements published in the prior 12 months that involve captioning or video accessibility, and assess whether they imply obligations that your programme has not yet addressed. The university lecture capture captioning post covers the OCR investigation landscape for higher education in detail.

EAA and international compliance

The European Accessibility Act has been enforceable for private sector entities with products and services in EU member states since June 2025. EAA references EN 301 549, which incorporates WCAG 2.1 Level AA for digital content including video. If your organisation produces training content accessed by employees or learners in EU member states, EAA compliance is an active obligation. The annual review should confirm: (1) the EU Accessibility Statement on your public-facing LMS portal or website is current and accurately describes the compliance status of training content, (2) any new content published in the year meets the WCAG 2.1 AA captioning requirement for EU-accessible content. The EAA accessibility statement template post covers the statement requirements.

UK accessibility: the Public Sector Bodies (Websites and Mobile Applications) Accessibility Regulations (PSBAR) apply to UK public sector bodies and reference WCAG 2.1 AA. The UK Equality Act 2010 duty to make reasonable adjustments applies to private employers for employee training content. Review whether any UK regulatory guidance issued in the prior year affects your organisation's UK obligations.

State law changes

Several US states have passed or amended digital accessibility laws and executive orders that create obligations beyond federal ADA and Section 508. California, New York, Illinois, and Washington have the broadest state-level accessibility obligations. Review whether any state law changes in the prior 12 months affect your organisation's state-level caption obligations — particularly if your organisation has employees, customers, or learners in states with broad accessibility laws. The state caption compliance law post covers the state-by-state landscape in detail.

Governance policy update

The output of the regulatory refresh component is an updated governance policy that accurately reflects the current compliance landscape. The policy should specify: the technical standard (WCAG 2.1 Level AA, SC 1.2.2), the accuracy threshold (99% per DCMP protocol), the scope of content covered, the compliance obligation basis (ADA Title I, ADA Title II, Section 508, EAA — whichever applies), the version date, and the next review date. Any changes from the prior year's policy should be tracked in the policy version history. The updated policy should be co-signed by the L&D Director and the Legal or Compliance officer who originally signed the policy. The governance policy template post has the full policy structure.

Component 4: Platform and LMS delivery audit

A caption compliance programme can have 100% caption coverage, 99.5% accuracy, and a current vendor SLA — and still fail learners at the point of delivery if the LMS or video platform is not serving the caption file correctly. Platform delivery failures are the most invisible compliance gap: they do not appear in the vendor's accuracy reports, the LMS content count, or the governance policy. They appear only when a learner with an accessibility need encounters the broken experience, or when an auditor runs a functional test.

Functional delivery check

For each video platform in the organisation's current stack, run a functional caption delivery check using a test video and a known-good caption file. The check should be performed in the production environment, not staging, using a standard learner account (not an administrator account, which may display captions even when they are not available to learners due to admin override settings).

The functional check confirms: the caption file loads and displays when playback begins; captions are correctly timed (words appear within 0.5 seconds of spoken audio, synchronisation is maintained throughout the video); captions are toggleable (learner can turn them on and off); caption styling meets WCAG SC 1.4.3 minimum contrast requirements (typically white or yellow text on black background); the caption player does not obscure other player controls or essential video content; and for screen reader testing, the caption transcript is accessible to AT software in the player's accessible mode.

For SCORM packages built in Storyline, Rise, or Captivate: run the functional check in the SCORM player as it appears in the LMS, not in the authoring tool preview. The authoring tool preview displays the content in a controlled environment that may not replicate the SCORM shell as deployed in the LMS. Common SCORM delivery failures include the closed caption button not appearing in the published output, captions displaying only on certain slides and not others, and captions appearing in the authoring tool preview but not in the LMS-hosted SCORM package. The eLearning authoring tools captioning post covers the SCORM delivery configuration for each major authoring tool.

New platform onboarding check

Any video platform added to the organisation's technology stack during the year should be included in the platform audit scope, even if no formal onboarding process included caption configuration. Common platforms added mid-year without caption planning: Loom for asynchronous manager communication; Vimeo for external-facing product training; Microsoft Stream for Teams meeting recordings and company all-hands; internal wikis (Notion, Confluence) with embedded video; sales enablement platforms (Highspot, Seismic) with embedded training content. Each of these creates a new caption delivery obligation if the content includes training video accessible by employees or learners with accommodation needs.

For each platform identified as new during the year, the audit should confirm: (1) whether caption delivery is technically supported by the platform, (2) whether the platform's caption configuration has been completed, (3) whether any content on the platform has been captioned, (4) whether the platform is covered by the current vendor contract scope.

LMS migration verification

If an LMS migration occurred during the year, the platform audit must include a post-migration caption verification check. LMS migrations are the single most common cause of large-scale, sudden caption delivery failures. Video files migrate successfully; caption sidecar files (SRT, VTT) are often not migrated, or are migrated into a file storage location that the new LMS does not associate with the video. The result is a migrated library where all video content appears in the new LMS without captions — even content that was fully captioned and WCAG-compliant in the prior system.

The post-migration verification check should sample 100% of migrated content if the library is under 100 videos, or a stratified random sample of at least 50 videos if the library is larger. For each sampled video, confirm that the caption file is present and loading correctly in the new LMS environment. If caption files are missing, the remediation is not re-captioning (the caption files already exist) — it is re-importing the caption sidecar files into the new LMS from the vendor portal or the organisation's caption archive. The LMS migration caption checklist covers the migration verification process in detail.

Caption format compatibility

Caption format support varies by platform and updates with platform releases. SRT is the most broadly supported format across LMS platforms. VTT (WebVTT) is the preferred format for HTML5 video players and is required by some LMS platforms for correct rendering of positioning and styling. TTML is used in specific broadcast and enterprise contexts. Platform updates can change the preferred or supported format mid-contract. The annual audit should confirm that the current caption file format matches the current format requirements for each platform in the stack. If a platform has updated its format requirements during the year, re-export the affected caption files from the vendor portal in the new format — do not re-caption, as the underlying transcription is unchanged.

Component 5: Budget reconciliation and next-year forecast

The budget reconciliation component converts the programme's operational data into financial language that leadership can act on. It answers two questions: what did the programme actually cost in the year just ended, and what should be budgeted for the coming year? The goal is not to justify the programme's cost — that belongs in the ROI framing conversation with finance — but to ensure the budget request for next year is grounded in actual year-just-ended data rather than year-1 assumptions that may no longer apply.

Actual vs. budget reconciliation

Reconcile actual spend against budget by line item. The standard caption programme budget has four cost categories:

• External vendor cost: Total spend on captioning services, broken down by per-minute captioning, subscription fees, rush processing fees, QA services, and any additional services (glossary management, DCMP reporting). Compare actual against the budgeted amount for each line item.
• Internal labour cost: Total hours spent on caption correction, QA review, glossary maintenance, programme coordination, vendor management, and LMS caption administration — valued at the loaded hourly cost of the staff performing each function. This is the cost that most organisations undercount because it is absorbed into individuals' broader job responsibilities rather than tracked against the caption programme. If internal correction time has declined relative to year 1 (because glossary maturity has improved accuracy), this decline should appear as a measurable cost reduction.
• LMS and platform infrastructure: Any additional captioning modules, caption API access fees, or caption player licensing added during the year that were not in the original programme budget.
• Training and onboarding: Cost of onboarding new staff members to the caption QA process, accessibility training, or vendor liaison training.

Identify variance causes for any line item where actual spend exceeded budget by more than 10%. Common variance causes: content volume growth exceeded forecast (more video produced than budgeted); rush processing fees accumulated because the caption-before-publish gate was not consistently enforced; content revision rates were higher than forecast; new platforms were added that required additional infrastructure spend.

Unit economics

Unit economics make the programme's cost structure visible in a way that absolute spend does not. The three key unit metrics for a caption programme are:

• Cost per captioned minute: Total programme cost (external vendor + internal labour) divided by total captioned minutes. This is the all-in cost of producing one compliant captioned minute of content. It should decline year-over-year as volume grows (vendor volume discounts reduce external cost) and as glossary maturity reduces correction labour. If cost per captioned minute is rising, investigate: vendor pricing increases, glossary maintenance neglect, or declining accuracy that is increasing correction labour.
• Correction labour ratio: Internal correction hours per hour of captioned content. The target for a programme with a well-maintained glossary covering the organisation's full technical vocabulary is under 15 minutes of correction per hour of video. A ratio above 60 minutes per hour indicates a systematic accuracy problem — either the vendor's accuracy has declined below the glossary-supported threshold, or the glossary is no longer covering the vocabulary being used in current content. The hidden FTE cost of caption correction post covers the correction labour model in detail.
• Cost per compliance-verified minute: Total programme cost divided by total minutes verified as compliant via DCMP spot-check. This is a higher bar than cost per captioned minute — it counts only the minutes that have been independently confirmed as meeting the 99% WCAG threshold, not just the minutes with a caption file attached.

Present unit economics year-over-year if this is the second or later annual review. A declining cost per captioned minute is a sign of programme maturity. A rising unit cost requires explanation. A correction labour ratio trend line tells the vendor performance story better than accuracy percentages alone — the vendor may maintain 99% accuracy on spot-checks while the correction labour ratio rises because accuracy on unchecked content has declined.

Next-year forecast

Build the next-year budget forecast using three inputs: the new content production forecast, the back-catalogue remediation plan, and the vendor contract renewal assumptions.

New content production forecast: How many hours of new video are planned for the coming year? Use the historical growth rate as a base (the year-just-ended volume vs. prior year). Common growth rates for L&D teams at growing organisations: 20–40% per year as headcount grows and product complexity increases. If a major product launch is planned, add the projected training content hours for that launch explicitly as a budget line item.

Back-catalogue remediation plan: What is the remaining uncaptioned back catalogue, and what is the plan for remediation? The back-catalogue budget should specify the total hours to be remediated, the prioritisation sequence (Tier-1 compliance obligations first), and the timeline. If the remediation budget is insufficient to complete the full back catalogue in year 2, document which content will remain unaddressed and why, and include that in the compliance risk section of the leadership presentation.

Contract renewal assumptions: If the vendor contract renews in the coming year, what pricing assumptions should the budget use? If renegotiation is planned, budget at the current rate (conservative) and note the renegotiation target separately. If a new vendor RFP is planned, budget at the mid-range of the expected market rate for the current volume level.

The three-year caption programme budget planning guide has a detailed model for next-year forecast construction, including volume discount threshold calculation and per-minute vs. subscription model comparison at different volume levels.

Component 6: Glossary maintenance review

The glossary is the mechanism that makes captioning accuracy compound per-customer over time — but only if it is maintained. A glossary built during the initial programme setup reflects the organisation's product and technical vocabulary at that moment. Over 12 months, products are renamed, SDK versions increment, company acquisitions add new brand names, and regulatory vocabulary evolves. A glossary that is not updated annually is a glossary that is losing coverage relative to the content it is being applied to.

Why glossaries decay

The most common glossary decay pattern is silent: the glossary term count does not decrease, but the percentage of the organisation's current vocabulary that is covered by the glossary declines as new terms are introduced and not added. A product launch in March introduces 15 new feature names, three new SDK component names, and two new integration names — all of which will appear in training videos produced from April onward. If these terms are not added to the glossary before the first training video about the new product is captioned, all 20 terms will generate substitution errors in every caption produced for that product until the glossary is updated. The vendor's accuracy for that content category will decline. The correction team will spend hours correcting the same errors repeatedly on every new product training video. This is entirely preventable with a proactive glossary update process tied to the product launch calendar.

Annual glossary audit

Pull the full glossary term list and compare against four sources:

• Current product naming: Marketing website, product roadmap, release notes. Any product, feature, or integration name that is now different from the glossary entry should be updated. Any new product, feature, or integration introduced since the last glossary update should be added.
• Current technical documentation: API documentation, SDK release notes, developer documentation. SDK component names, API endpoint names, CLI command names, error code identifiers, and configuration parameter names are the highest-density sources of proper-noun captioning failures in engineering training content. These change with every major version release.
• Regulatory and compliance vocabulary: Any new regulatory framework, standard, or acronym that was introduced in the compliance training curriculum during the year. This includes new OSHA standards, new HIPAA guidance, new NIST framework versions, new CMMC levels, new ADA enforcement guidance, and new industry-specific certification vocabulary.
• Acquired brands and new entity names: Any company acquisition or organisational restructure that introduced new brand names, entity names, or product lines into the training vocabulary. Acquired brand names are typically introduced in onboarding training for the integration and in product training for customers of the acquired product — both of which are captioned within weeks of the acquisition close date.

Document the glossary update: terms added, terms retired (sunset products and deprecated SDK components), terms updated (renamed products), total term count before and after the update. This documentation belongs in the programme governance record as evidence that the glossary maintenance process is being actively managed.

Glossary accuracy validation

For a sample of content captioned in the most recent quarter, run an accuracy check specifically on glossary-covered terms, separate from the general DCMP accuracy check. The question is whether the glossary is actually improving accuracy for the specific vocabulary it covers, or whether there are terms in the glossary that are not being applied correctly by the caption engine.

Glossary terms that have high phonetic similarity to common English words are the most likely to be substituted even when the glossary entry is present. For example, a product named "Clario" may be correctly captioned most of the time (because the glossary is applied) but substituted as "clarity" or "clarion" in content with lower audio quality. If glossary-covered terms show accuracy below 99%, the remediation options are: add pronunciation guides to the glossary entry (phonetic spelling that helps the ASR engine apply the correct output), add variant forms of the term to catch common error forms, or flag the specific term for post-production manual review in all new content.

GlossCap's glossary-biased decoding architecture is designed to compound accuracy per-customer over time: each captioned hour where a glossary term is correctly identified and confirmed strengthens the term's probability weight in subsequent caption jobs for that customer. The annual glossary audit should include a review of the confidence metrics for each glossary term — terms with consistently high confidence scores are well-anchored; terms with low or declining confidence scores may benefit from additional pronunciation guidance or context examples.

New vocabulary acquisition triggers

The annual review establishes the glossary baseline and clears accumulated debt. But glossary maintenance should not wait for the annual review. Three trigger events should initiate a glossary update immediately:

• Product launch: Any new product, major feature release, or platform version that introduces new vocabulary to the training content pipeline should trigger a glossary update before the first training video about the release is submitted for captioning. The L&D producer who commissions the training video and the product marketing manager who owns the feature launch are the two sources for the new vocabulary list. Build a standing checklist into the video production workflow: "list any new product names, feature names, or technical terms that appear in this video" — this checklist becomes the glossary update input.
• Company acquisition: Acquisitions introduce new brand names, product names, and technical vocabulary that will appear in integration training and customer training within weeks of the deal close. The glossary update for acquisition vocabulary should be prioritised above routine new-content glossary updates, because acquisition training is often on an accelerated production schedule and accuracy failures on the acquired brand's vocabulary are visible to the employees of the acquired company who are evaluating the new organisation.
• New regulation or compliance framework: Any new regulatory framework introduced to the compliance training curriculum should trigger a glossary update for the regulatory vocabulary. Regulatory vocabulary is a significant source of captioning failures — acronyms like CMMC, ERISA, HIPAA, OSHA 300A, and NIST CSF are high-frequency in compliance training and are also among the most frequently misidentified by generic ASR systems. Adding regulatory framework terminology to the glossary before the first compliance training video on the new framework is a standard programme hygiene step.

Running the annual review meeting

The annual review meeting is a two-hour structured session with a fixed agenda, pre-distributed data, and a defined output. The meeting does not conduct the review — the review happens in the two weeks before the meeting, when data contributors prepare their inputs. The meeting synthesises the findings, resolves any discrepancies, and produces the action plan and leadership presentation outline that leave the room as committed deliverables with owners and deadlines.

Participants

• Accessibility coordinator: Organises the review, distributes the data collection template, runs the meeting agenda, and owns the annual review report output. If the organisation has an accessibility coordinator role, this is the primary owner. If not, the L&D operations lead performs this function.
• L&D Director or VP of Learning: Sponsors the review, owns the leadership presentation output, makes final decisions on action plan priorities and budget assumptions. Required participant for all six components.
• LMS administrator: Presents the content library counts, new platform additions, migration status, and platform delivery audit results. Required participant for Components 1 and 4.
• Vendor manager or vendor account representative: Presents the vendor's accuracy report, SLA adherence summary, and pricing schedule. Required participant for Component 2. Note: the vendor account representative's role in the meeting is data-provision only — the vendor performance evaluation is conducted by the programme team, not by the vendor.
• Finance business partner: Validates the budget reconciliation and next-year forecast against the organisation's financial records. Required participant for Component 5.
• Legal or compliance officer: Reviews and validates the regulatory refresh analysis. Can participate in person for the Component 3 section or provide a written review of the draft regulatory analysis before the meeting.

Meeting structure

Pre-meeting data collection (two weeks before)

Distribute a data collection template to each data contributor two weeks before the meeting. The template specifies exactly what data each contributor should prepare and in what format. Incomplete or late data submissions should be escalated one week before the meeting — not discovered on the morning of the review. A review meeting conducted with incomplete data produces incomplete findings and an action plan with gaps that will not close before the next annual review.

The data collection template should request from each contributor:

• LMS administrator: Total video asset count (current library), new video count (added in the past 12 months), new video captioned count, new video uncaptioned count, content revision log (video files modified after caption delivery date), decommissioned content list, migration status, new platform list.
• Vendor: Annual accuracy report by content category, SLA adherence summary (jobs delivered within SLA / total jobs), SLA breach log with root cause, current pricing schedule, contract renewal date.
• Finance: Actual spend by budget line item (external vendor, internal labour allocation, infrastructure), year-over-year comparison against prior year actuals and current year budget.
• Legal/compliance: List of regulatory changes in the prior 12 months relevant to the organisation's caption obligations, assessment of any pending regulatory changes expected in the coming 12 months.

Post-meeting deliverables

The annual review produces five deliverables, each with a named owner and a delivery deadline:

1. Annual review report (internal): documents all six components, findings, action items, and data used. Owner: accessibility coordinator. Deadline: five business days after review meeting.
2. Leadership presentation (10 slides): summary for the L&D Director to present to leadership or executive sponsors. Owner: L&D Director. Deadline: five business days after review meeting.
3. Action plan: every finding mapped to a single owner, deadline, and priority. Distributed to all owners within 24 hours of the review meeting. Owner: accessibility coordinator. Deadline: same day as review meeting.
4. Updated governance policy: incorporates regulatory changes, policy updates from the legal review, and any process changes agreed in the review. Co-signed by L&D Director and Legal. Owner: accessibility coordinator. Deadline: 10 business days after review meeting.
5. Glossary update: all term additions, retirements, and updates committed to the caption system. Owner: accessibility coordinator or glossary manager. Deadline: five business days after review meeting.

The leadership presentation

The annual review produces two documentation outputs: an internal report that documents every finding and data point in detail, and a leadership presentation that summarises programme status in terms that leadership can act on. The internal report is evidence. The leadership presentation is communication. They serve different purposes and should be prepared separately.

Leadership does not need the DCMP accuracy protocol specification. Leadership needs to know: are we compliant, what does it cost, and what are the risks heading into next year? The 10-slide structure below answers those three questions without requiring the presenter to over-explain the technical details.

The 10-slide deck

Slide 1: Programme health dashboard

A single-page green/amber/red status dashboard across five dimensions: caption coverage (percentage of library captioned), accuracy (% meeting 99% WCAG threshold), regulatory compliance status, vendor performance, and budget adherence. Leadership sees the programme health at a glance before any detail. This slide is often the only one that gets shared outside the L&D department — design it to stand alone.

Slide 2: Coverage — what we captioned this year

Total video library size (hours). New content added in the year. New content captioned. Back-catalogue remediation progress (% of back catalogue now captioned vs. prior year). Uncaptioned items and remediation plan summary. Use a bar chart or progress ring — absolute numbers are harder to read than visual proportions for this audience.

Slide 3: Accuracy — are we meeting the compliance threshold?

Average accuracy rate for the year (from DCMP spot-check). Comparison to 99% WCAG threshold. Trend: year-over-year accuracy if this is the second or later annual review. Content category breakdown if accuracy varies materially by category. One-sentence explanation of any category that is below threshold and the remediation in progress.

Slide 4: Compliance status — where we stand

Which regulatory obligations apply to the organisation (ADA Title I, ADA Title II, Section 508, EAA, state law). For each obligation: status (in full compliance / at manageable risk / at elevated risk). Any open OCR complaints or litigation. Any regulatory changes from the prior year that affected the programme's compliance posture. This slide is the compliance risk summary that legal and the executive team need to see.

Slide 5: Vendor performance

Vendor name. Contract term remaining. SLA adherence % for the year. Accuracy self-reported vs. independently measured. Notable incidents. Contract renewal recommendation (renew / renegotiate / RFP). One line on the outcome of any renegotiation already in progress.

Slide 6: Budget — what we spent and what it cost per minute

Actual spend vs. budget for the year (by line item: external vendor, internal labour). Cost per captioned minute (year just ended vs. prior year). Correction labour ratio (minutes per captioned hour — declining is the target). Any budget overruns and the cause. This slide should show a trend: a mature programme should have declining unit costs over time.

Slide 7: Key achievements

Three to five bullet points. Back-catalogue percentage improvement. Glossary expansion (terms added). New platforms onboarded with caption delivery. Accuracy improvement year-over-year. Any milestone (first year with zero OCR complaints, first year with 100% new content captioned before publish). This slide is the positive framing — it demonstrates programme value, not just programme cost.

Slide 8: Key risks heading into next year

Uncaptioned content categories and the risk they represent. Regulatory changes in force next year that raise the compliance bar. Vendor contract renewal risk. Platform changes planned that will require caption reconfiguration. Back-catalogue pace vs. remediation timeline. Frame risks as manageable-with-action, not as binary failures — leadership needs to know what to approve or resource, not what to worry about.

Slide 9: Next-year plan and priorities

Top three to five programme priorities for the coming year with brief rationale. Back-catalogue remediation target (percentage or hours). New platform onboarding plan. Vendor renegotiation or RFP if planned. Regulatory refresh actions. Glossary update schedule. This slide is the plan that leadership is being asked to endorse, not just observe.

Slide 10: Budget request for next year

Next-year budget request by line item (external vendor, internal labour, infrastructure). Year-over-year comparison. Unit economics projection (if volume grows as forecast, cost per captioned minute should decline to X). Any large one-time items (migration, back-catalogue sprint, new vendor RFP). The approval ask — what leadership action is needed to fund the plan on Slide 9.

Compliance status framing

Leadership presentations about compliance status risk two common framings that are both unhelpful: "we're fully compliant" (which is rarely precisely accurate and creates false certainty) and "we have significant compliance gaps" (which triggers alarm without providing the context for risk assessment). The accurate framing distinguishes between programme status dimensions:

• In full compliance: All mandatory content captioned to WCAG 2.1 AA accuracy; back-catalogue Tier-1 (live compliance obligations) remediated; governance policy current; no open OCR complaints or litigation.
• At manageable risk: Tier-2 and Tier-3 back-catalogue remediation in progress per documented plan; new content captioned before publish (with documented exception process for edge cases); no open complaints; accuracy meeting threshold for all spot-checked content.
• At elevated risk: Open OCR complaint or litigation; significant uncaptioned content with no documented remediation plan or timeline; accuracy below threshold for a content category that is actively delivered to employees with accommodation needs; platform delivery failure affecting learner access.

The leadership presentation should conclude with a clear statement of the programme's current risk tier and the specific actions that would move it from an elevated or manageable risk tier to the next tier up. This gives leadership the information they need to make resource decisions, not just a status update.

Budget framing for financial leadership

CFOs and financial leadership respond to unit cost trends, not absolute spend. Present the caption programme in unit economics terms: cost per captioned minute, year-over-year. A mature programme that is producing 30% more content per year while cost per captioned minute is declining has a positive efficiency story that makes the absolute spend increase defensible. A programme where cost per captioned minute is rising without a corresponding accuracy improvement has a cost management story that needs to be addressed. The caption ROI framing for finance executive audiences post has specific frameworks for presenting caption programme value in CFO-friendly language.

Annual review calendar template

The calendar below is a 12-month template keyed to an organisation with a January fiscal year end. Adjust the month numbers proportionally if your fiscal year ends in a different month. The key constraint is that the annual review meeting should occur approximately 4–6 weeks after fiscal year close, giving data contributors time to prepare complete annual data without slipping into the budget cycle window where the forecast is needed for decisions being made in parallel.

Mid-year check-in (June–July): The mid-year check-in is a shorter event (30–60 minutes) that reviews: content captioning rate for H1, accuracy spot-check for H1, back-catalogue remediation progress against the annual plan, and any action items from the annual review that are behind schedule. It is not a full six-component review — it is a progress gate that confirms the programme is tracking as expected and surfaces any emerging issues before they compound into year-end findings.

Eight failure modes in annual reviews

1. The annual review never happens at all

The governance policy references an annual review but no one owns the scheduling. The accessibility coordinator who built the programme has moved to a different role. The annual review gets scheduled in Q1, deferred to Q2, deferred again to Q3, and quietly dropped when Q4 sprint work begins. The following year, the same conversation happens — "we should do an annual review" — with no prior-year baseline to compare against and no institutional knowledge of what the review is supposed to cover. Fix: schedule the annual review as a standing recurring calendar event at the time the programme governance policy is signed — not after the programme is running. Include a note in the governance policy that the annual review is a programme compliance requirement and that the review date is a standing calendar event maintained by the accessibility coordinator role, not the individual currently in the role.

2. Content audit counts only new content, not revised content

The most common oversight in content library audits. New content added during the year may be 100% captioned — because the caption-before-publish gate is working for new content. But 25–40% of existing content was revised during the year and the captions were not updated. The annual review scope explicitly includes content revision audits alongside new content inventory. Fix: add a content revision audit as a required component of the data collection template that the LMS administrator completes before the annual review. Require the LMS admin to pull a list of all video files with modified dates more recent than the associated caption file date.

3. Vendor performance review uses only vendor-supplied data

The vendor supplies their own accuracy reports. These are internally generated metrics, not independent audit results. A vendor whose internal QC protocol is less rigorous than the DCMP protocol, or who selectively samples easy content for their internal accuracy reporting, will show higher accuracy than independently measured results reflect. Fix: the vendor performance review must include independently conducted DCMP spot-check results alongside vendor-supplied data. The DCMP spot-check sample from the annual content audit is specifically designed for this purpose. Any discrepancy greater than 0.5 percentage points between vendor-reported and independently measured accuracy is a programme finding that requires a vendor conversation.

4. Regulatory refresh is conducted by L&D without legal involvement

L&D teams often conduct their own regulatory review by reading ADA news and checking accessibility blogs. This approach misses enforcement guidance that is published in DOJ or EEOC administrative records, OCR resolution agreements that set precedent for compliance interpretation, and state law changes that are not prominently reported in the general accessibility press. A self-conducted regulatory review is better than no review, but it is not sufficient for an organisation with material compliance obligations. Fix: engage legal counsel for the regulatory refresh component, at minimum for a 30-minute review of the L&D team's draft regulatory analysis. The legal review should check whether any enforcement guidance, resolution agreements, or court decisions issued in the prior 12 months affect the organisation's compliance obligations.

5. Platform delivery audit is conducted in a staging environment

Caption delivery in a staging LMS environment may work correctly while production has a configuration drift — for example, a production LMS upgrade that reset the caption player configuration but did not affect the staging environment. The staging environment is useful for testing new configurations, but it is not a proxy for the production delivery experience. Fix: run the functional delivery check in the production environment, using a real learner account. Administrator accounts in many LMS platforms have override settings that allow them to see captions even when the caption delivery configuration is broken for standard learner accounts. The learner account test is the only test that accurately represents what a learner with an accessibility need experiences.

6. Budget reconciliation is treated as a finance exercise, not a compliance exercise

The budget reconciliation is typically owned by finance and treated as a cost tracking exercise. The compliance dimension of the budget — the unit economics that reflect programme quality, not just cost — is not presented to leadership. Cost per captioned minute rising year-over-year without a corresponding accuracy improvement is a warning sign that should be surfaced to leadership alongside the accuracy and coverage metrics. Fix: present unit economics (cost per captioned minute, correction labour ratio) as programme quality indicators alongside the accuracy and coverage metrics, not as separate finance line items. The correction labour ratio in particular is the most sensitive leading indicator of glossary maintenance gaps or vendor accuracy decline.

7. Glossary maintenance is deferred because "captions seem fine"

The glossary maintenance review is the component most commonly deferred because the compliance failure it prevents is invisible until it manifests as an accuracy decline in a future spot-check. The team's perception that "captions are fine" is based on what they see in the QC workflow — and if the glossary is not covering new vocabulary, the accuracy failure appears in content categories that may not be sampled in the routine QC sweep. Fix: tie glossary maintenance to the product launch calendar, acquisition calendar, and regulatory calendar — not to the annual review alone. The annual review is the catch for any vocabulary added during the year that was not caught at the time. The standing trigger events (product launch, acquisition, regulatory change) are the first-line mechanism for keeping the glossary current between annual reviews.

8. Action plan has no single owners or deadlines

The annual review produces a list of findings. The findings are discussed. The meeting ends. The action plan is drafted as a list of things "we" should do. At the following year's review, the same findings appear on the agenda because "we" never mapped to a specific person with a specific deadline. Fix: every finding from the annual review maps to a single named owner (not a team or a role — a person), a specific deadline, and a priority classification (critical: address within 30 days; medium: address within 90 days; low: address before next annual review). The action plan is distributed to all owners before the end of the day of the annual review meeting. The mid-year check-in reviews action plan progress explicitly and names any items that have slipped.

FAQ

How is the annual review different from the quarterly compliance KPI report?

The quarterly compliance KPI report is an operational cadence — it tracks content volume, caption coverage, accuracy rates, and vendor throughput on a regular reporting cycle. It is forward-looking (what is the programme producing now) and backward-looking for one quarter at a time. The annual review is a structured assessment event that: synthesises four quarters of KPI data into a year-level view, adds vendor performance, regulatory, budget, and glossary components that are not tracked quarterly, and produces a governance update and leadership presentation as outputs. The quarterly KPI report feeds data into the annual review; it does not replace the annual review. Running a strong quarterly KPI cadence makes the annual review faster to prepare (four quarters of data are already documented), but the annual review requires components — regulatory refresh, vendor contract review, glossary audit — that the quarterly report does not cover.

Does the annual review need to happen on an exactly 12-month cycle?

Annual is the minimum cadence for a stable programme. The six trigger events (vendor change, LMS migration, regulatory event, significant accuracy failure, acquisition, major product launch with vocabulary impact) create exceptions where a mid-year review of one or more components is appropriate. The calendar block should be annual and standing. If a trigger event prompts a mid-year review of one component, the findings from that mid-year review are incorporated into the next annual review rather than resetting the annual calendar. Running the annual review on an exactly 12-month cycle allows year-over-year comparison of all metrics on a consistent basis.

Who needs to formally accept the annual review findings?

The annual review report should be formally accepted by the L&D Director or VP of Learning, with a notation that the review was conducted and accepted on a specific date. The governance policy update that results from the review should be co-signed by the same parties who signed the original governance document — typically the L&D Director and Legal or Compliance. Formal acceptance of the review and the policy update creates an evidence record that the organisation did not simply build a caption programme and walk away from it, but actively manages it through a structured annual governance process. This evidence record is valuable in any OCR complaint, litigation discovery, or external accessibility audit.

What if the vendor renegotiation is in progress during the annual review?

Treat the vendor performance review and the vendor renegotiation as separate processes that share data. The annual review provides the vendor performance data (accuracy, SLA adherence, volume summary) that informs the renegotiation. The renegotiation is a contract event; the annual review is a programme assessment event. Document the renegotiation status in the annual review report and note the pricing assumptions used in the next-year budget forecast as "pending renegotiation outcome" rather than confirmed rates. After the renegotiation concludes, update the budget forecast and note the revision in the programme documentation. Do not delay the annual review because renegotiation is in progress — the review data is exactly what the renegotiation needs.

How do we handle content from departments outside L&D in the annual review?

If the captioning policy covers video produced by any department and distributed via company systems, that content is in scope for the annual review. The content library audit should pull video counts across all departments, not only L&D-produced content. Departments that have been bypassing the caption-before-publish gate — knowingly or unknowingly — should be identified in the content audit. The annual review action plan should include outreach to those departments, not as a compliance complaint but as a programme expansion event: the department needs to be onboarded to the caption workflow, their content managers need to be trained, and their upcoming production schedule needs to feed into the next year's volume forecast. The change management rollout post covers the process for expanding caption programme adoption across departments.

What if the DCMP spot-check finds accuracy below 99% for some content?

Below-threshold accuracy findings from the annual spot-check are treated as immediate programme incidents, not as findings to be queued for next year's action plan. The annual review documents the finding: when first identified, what content category was affected, what the root cause is (vendor accuracy decline, glossary gap, content category with high proper-noun density, audio quality issue), and what the remediation action is (re-caption the affected content, expand the glossary, add vendor correction requirement, negotiate SLA amendment for affected content categories). Any content with active compliance obligations — mandatory compliance training, content distributed to employees with documented accommodation requests — should be prioritised for immediate re-captioning if accuracy is below threshold. The QA methodology post has the incident response process for below-threshold findings.

Is there a regulatory requirement to conduct an annual review?

No regulatory framework — ADA, Section 508, EAA, or WCAG — specifies a required annual programme review cadence. The annual review is a programme governance best practice, not a regulatory mandate. However, if an organisation has a captioning policy that references an annual review (as most policy templates do), and that policy has been disclosed to legal counsel, shared as part of a compliance representation, or produced in discovery during an OCR investigation or litigation, failing to conduct the documented annual review could be used as evidence of inadequate programme governance. The organisation represented that it would review the programme annually; the evidence record shows it did not. The governance policy template post discusses the compliance value of programme documentation and the risk of policy commitments that are not followed through.