Caption programme self-assessment: a scored 40-question checklist for L&D teams to evaluate WCAG 2.1 AA compliance across video libraries, live training, vendor accountability, and governance

Why a scored self-assessment, not a yes/no checklist

A yes/no compliance checklist asks “do you have captions?” and produces a binary answer. The auto-captions compliance status post explains why the yes/no answer is almost always misleading: the LMS has captions (technically yes), but they are auto-generated without accuracy review, without a glossary applied, and without any QA process (functionally no, for WCAG 2.1 AA purposes). A binary checklist cannot capture this distinction. A scored assessment can.

The scored approach also produces a gap priority. If an L&D team scores 8/10 on coverage, 6/10 on accuracy, 4/10 on vendor accountability, and 2/10 on live training captioning, the next dollar and the next hour of time goes to live training captioning — not to further improving coverage, which is already strong. A yes/no checklist applied to those same domains would show four green checks and tell the L&D team nothing about priority.

Finally, the scored assessment produces a benchmark for the annual programme review. Running the same assessment each year and comparing domain scores over time shows where investment has moved the needle and where progress has stalled. That trend line is the evidence a CLO needs to approve continued investment in the caption programme — and the evidence the L&D team needs to justify requesting it.

How to use this assessment

Before you start, a few ground rules for getting an accurate score:

• Score what is actually implemented, not what is planned or in progress. If a vendor contract amendment is drafted but not signed, the answer is 1 (in progress), not 2 (fully implemented). Inflating scores to reflect intentions produces a score that does not reflect actual risk exposure.
• Score based on all content and all training programmes, not your best examples. If you have excellent captions on your most recent onboarding video but 200 uncaptioned videos from 2019, the coverage score reflects the total state of the library, not the most recent addition to it.
• Complete each domain independently. The domains cover different infrastructure layers — policy, technology, vendor, process, and governance — and a weakness in one does not imply weakness in others. Score each on its own merits.
• Note the date of your assessment. The assessment is a point-in-time snapshot. Plan to re-run it in six to twelve months to measure progress. The enterprise LMS caption audit methodology describes how to structure the library audit that supports several of these questions.

Scoring:

• 0 — Not started: No process, policy, or implementation exists for this item.
• 1 — Partially implemented or in progress: The item is under development, inconsistently applied, or implemented for some but not all content/events/vendors.
• 2 — Fully implemented: The item is documented, consistently applied, and would be defensible to an external reviewer.

Domain 1: Caption coverage

Caption coverage answers the first question any regulator will ask: what percentage of your training content has captions? The caption compliance programme build guide recommends auditing your content library before writing policy, because the audit reveals the scale of the gap. Domain 1 assesses whether you know your coverage state and whether your policy is closing the gap over time.

Maximum domain score: 10

Q1.1: All training video content added to your LMS in the last 12 months is captioned, and captions have been reviewed for accuracy at or above 99% (not auto-generated with no review). 0 = less than half of new content is captioned or captioned only at auto-generated accuracy; 1 = 50–90% of new content is captioned with accuracy review; 2 = more than 90% of new content meets the accuracy standard.

Q1.2: A written policy exists requiring captions on all new training video before publication in the LMS — not a verbal norm or shared understanding, but a documented standard that applies to all content producers including instructional designers, subject-matter experts recording screencasts, and external producers. 0 = no written policy; 1 = draft policy or informal norm; 2 = written policy in force with a named owner.

Q1.3: You know the percentage of your total video library — including back-catalogue content produced before your current caption policy — that currently has WCAG-compliant captions (not auto-generated captions with no review, but reviewed captions at ≥99% accuracy). 0 = unknown; 1 = rough estimate based on content age or production source; 2 = audited and documented with caption status for each item.

Q1.4: All compliance-mandatory training content — ADA and EEO compliance training, HIPAA workforce training, OSHA safety training, anti-harassment training, mandatory annual certifications — is captioned regardless of when it was produced. 0 = no, some compliance-mandatory content is uncaptioned; 1 = partially (recent compliance content captioned, older content not); 2 = yes, all compliance-mandatory content has reviewed captions.

Q1.5: All external or customer-facing training content — external LMS academy, customer onboarding video, channel-partner certification training, public-facing product tutorial — is captioned to WCAG 2.1 AA standards. 0 = no; 1 = partially; 2 = yes.

Domain 1 score: ___ / 10

Domain 2: Caption accuracy

Coverage without accuracy is not WCAG compliance. WCAG 2.1 SC 1.2.2 requires captions that are “equivalent” to the audio — and the DCMP (Described and Captioned Media Program) protocol, which most regulatory guidance references, sets a 99% accuracy floor. The caption QA methodology guide covers how to measure accuracy using a reference transcript comparison. Domain 2 assesses whether you are measuring accuracy and whether specialised vocabulary is being handled correctly.

Maximum domain score: 10

Q2.1: You measure caption accuracy using a defined methodology — not by watching a video and estimating whether it “seems right,” but by comparing caption text against a reference transcript and calculating an error rate (errors per 100 words, or total errors ÷ total words × 100). 0 = accuracy not measured; 1 = informal spot-check viewing; 2 = systematic methodology with reference transcript comparison.

Q2.2: Your accuracy target is 99% or higher, consistent with WCAG 2.1 AA and DCMP requirements. 0 = no defined accuracy target; 1 = target defined but below 99% or stated as a range; 2 = 99%+ target is written into policy and/or vendor contract.

Q2.3: You have documented accuracy data for at least a sample of your current captioned library — not just a vendor claim of accuracy, but your own measurement of the vendor’s output on your content. 0 = no accuracy data; 1 = one or two informal samples; 2 = documented accuracy data by content type, vendor, or production batch, filed where it can be produced on request.

Q2.4: Specialised vocabulary specific to your organisation or industry — product names, SDK terms, regulatory acronyms, clinical terms, legal Latin, equipment names, trademarked terms — achieves the same accuracy rate as general speech in your captions. 0 = significant jargon errors identified and not resolved; 1 = glossary exists or is in use for some content types; 2 = glossary is applied systematically to all content with specialised vocabulary, and accuracy in jargon-heavy content is measured separately.

Q2.5: When a caption error is identified — by an employee complaint, QA review, or accommodation request review — a documented process exists for logging the error, escalating to the vendor or correcting in-house, updating the caption file in the LMS, and confirming the fix with the person who identified it. 0 = no error-handling process; 1 = ad hoc; 2 = documented process in place.

Domain 2 score: ___ / 10

Domain 3: Technical WCAG conformance

A caption file can pass a binary “captions present” check in the LMS while still failing WCAG 2.1 AA on technical grounds: poor synchronisation, all-caps formatting, missing speaker identification, or a player that cannot render the caption track on mobile. The WCAG 2.1 AA caption requirements specify that captions must be “synchronized with the audio and available for all prerecorded audio content” — but the technical requirements for what synchronisation means, how formatting affects equivalence, and how platform rendering affects accessibility are not typically evaluated in a content audit focused on coverage. Domain 3 assesses the technical layer.

Maximum domain score: 10

Q3.1: Captions are synchronised to the audio across your captioned library — the text appears within approximately two seconds of the corresponding speech and does not remain on screen after the speech ends. Synchronisation has been verified, not assumed. 0 = significant sync issues known or synchronisation not verified; 1 = generally synchronised but not systematically verified across the library; 2 = synchronisation verified across the captioned library using a sample-based QA process.

Q3.2: Captions display in mixed case (not all-caps), with correct punctuation and sentence-end breaks, with maximum line length maintained for readability, and at a reading speed that allows comprehension (generally below 250 words per minute for caption display). 0 = significant formatting issues (all-caps, no punctuation, run-on lines) known across the library; 1 = formatting issues present in some content; 2 = formatting standards applied consistently and verified.

Q3.3: In multi-speaker content where speakers are not visually distinguishable — audio-only training, talking-head interviews with similar framing, voice-over with multiple narrators, simulated dialogue training — speaker identification labels are used so that the viewer following only the caption track can identify who is speaking. 0 = speaker identification not used in any multi-speaker content; 1 = used in some multi-speaker content; 2 = applied consistently in all multi-speaker content where speakers are not visually distinguishable.

Q3.4: Captions are delivered as a closed caption track (a toggle-able CC track) rather than open (burnt-in) captions in all content where the LMS player supports it — or, where burnt-in captions are the only available format, a screen-reader-accessible text transcript is also provided so that the caption text is machine-readable. 0 = all or most captions are burnt-in with no alternative text equivalent; 1 = mixed (some CC tracks, some burnt-in); 2 = all content has CC tracks, or all burnt-in content has an accompanying accessible text transcript.

Q3.5: Captions render correctly in all environments where your learners access training: desktop LMS browser view, mobile LMS app (iOS and Android), embedded player in company intranet or documentation, and any offline download mode offered by your LMS. 0 = rendering not verified on any platform; 1 = verified on desktop only or on some platforms; 2 = verified across all primary learner access environments.

Domain 3 score: ___ / 10

Domain 4: Live training captioning

Live instructor-led training — in-person classroom, virtual classroom (Zoom, Teams, Webex), or synchronous cohort — has a different accessibility infrastructure requirement from a recorded video library. The real-time CART captioning guide covers how to procure and deploy CART services; the accommodation request process for live training is covered in the live course accessibility statement guide. Domain 4 assesses whether your live training operations have the minimum infrastructure in place to deliver on an ADA or Section 504 captioning accommodation request for a live training event.

Maximum domain score: 10

Q4.1: LMS course listings for live instructor-led training courses state the accommodation request deadline for captioning — the specific number of business days before the event that a request must be received — and name the intake contact (person or monitored inbox). 0 = no accommodation information in live course listings; 1 = some live courses have this information, others do not; 2 = all live course listings include the accommodation request deadline and intake contact.

Q4.2: The stated accommodation request deadline matches the actual lead time your CART vendor requires (standard five to seven business days for professional CART services), plus your organisation’s processing buffer of at least two additional business days for request intake, vendor confirmation, and integration setup. 0 = no deadline is stated in course listings; 1 = a deadline is stated but may not account for vendor lead time plus buffer; 2 = deadline verified against vendor SLA and buffer is explicitly included.

Q4.3: You have at least one CART provider under contract or on an approved vendor list who can provide live captioning services for your training sessions, with a written SLA that covers lead time, captioner qualifications, accuracy commitment for live delivery, and post-event transcript availability. 0 = no CART vendor identified; 1 = CART vendor known but no contract or SLA in place; 2 = CART vendor contracted with SLA.

Q4.4: Your event facilitators have a written protocol for what to do when CART is not working during a live session — a decision tree that covers at least: display-side configuration failure, feed-stopped failure, audio quality degradation, and provider unreachable. Facilitators have received this document before running a live session with CART. 0 = no protocol; 1 = informal guidance or verbal briefing only; 2 = written decision tree, distributed to facilitators before events with CART in use.

Q4.5: Post-event accessible alternatives — a captioned recording of the session or a verbatim CART transcript — are available within a stated, written timeframe for employees who missed a live session or who experienced a CART disruption during the session. The timeframe is communicated in the course listing or the accommodation confirmation. 0 = no post-event accessible alternative provided; 1 = provided ad hoc when an issue occurs; 2 = consistent process with a written delivery timeframe, communicated to all live course registrants.

Domain 4 score: ___ / 10

Domain 5: Vendor accountability

The vendor SLA and contract review checklist covers what to include in a vendor contract. The vendor audit rights and examination evidence guide covers what regulators expect to be produced from vendor documentation. Domain 5 assesses whether your current vendor relationship — or your primary tool if you are self-serving — provides the accountability infrastructure required to defend caption quality in a regulatory review. Most mid-market L&D teams discover, when they read their vendor contract carefully, that none of the five items below are included.

Maximum domain score: 10

Q5.1: Your caption vendor contract or service agreement includes a documented accuracy SLA — a minimum accuracy percentage (explicitly 99% or above) with a defined measurement methodology, not a vague commitment to “high quality” or “professional accuracy.” 0 = no accuracy SLA in contract; 1 = accuracy commitment present but measurement methodology is undefined or below 99%; 2 = ≥99% accuracy SLA with defined methodology.

Q5.2: Your contract includes explicit audit rights — the right to request vendor processing logs, reference transcripts, and accuracy verification records for your content, on reasonable notice, for the retention period defined in the contract. 0 = no audit rights clause; 1 = informal understanding or verbal commitment; 2 = explicit audit rights clause in the signed contract.

Q5.3: Reference transcripts — the verbatim source text used to verify caption accuracy — are retained by the vendor for a defined minimum period and are accessible to you on request. The retention period is at least three years, consistent with EEOC charge investigation timelines and OCR complaint investigation timelines. 0 = no retention clause or retention period unknown; 1 = retention clause exists but period is under 12 months; 2 = retention clause with ≥3-year period, explicitly stated in contract.

Q5.4: The contract specifies remediation obligations if vendor output falls below the accuracy SLA — at minimum, re-processing at no additional charge within a defined SLA, with a formal notice and cure process that does not require the customer to file a legal action to get resolution. 0 = no remediation clause; 1 = re-process clause exists but lacks a defined SLA or formal process; 2 = full remediation clause with SLA, notice process, and remedy escalation.

Q5.5: You have independently verified the vendor’s accuracy claims against a sample of your own content — not by relying on the vendor’s internal QA report, but by applying a reference transcript comparison to at least a sample of the content the vendor has processed. This verification has been documented. 0 = accuracy never independently verified; 1 = informal spot-check viewing once; 2 = systematic sample-based verification on vendor output, documented.

Domain 5 score: ___ / 10

Domain 6: Accommodation request process

The accessibility coordinator playbook covers the broader role of the accommodation request function in an L&D context. Domain 6 assesses the operational accommodation request infrastructure specifically: is there a designated intake point, are requests logged in sufficient detail, are requestors receiving confirmation, and is there a documented process for requests that arrive outside the standard window? These are the documents an EEOC investigator or OCR complaint reviewer will request — and if they do not exist, the organisation’s ability to demonstrate compliance is severely limited.

Maximum domain score: 10

Q6.1: There is a designated, named individual or monitored inbox responsible for receiving, logging, and responding to captioning accommodation requests for training content and events — not a generic “contact HR” instruction, but a specific role or monitored channel that is actively staffed during business hours. 0 = no designated intake point; 1 = person or inbox designated informally but no SLA for response; 2 = designated, documented, and with a defined response SLA.

Q6.2: Accommodation requests are logged in writing with sufficient detail to reconstruct the request and outcome for an EEOC or OCR inquiry: date received, employee name and department (where permitted by HIPAA or ADA medical information rules), accommodation type requested, date accommodation confirmed to employee, arrangement made (vendor name or tool, session date), and confirmation reference. 0 = requests not logged; 1 = some logging but fields are incomplete; 2 = systematic logging with all required fields, in a format that is retrievable on request.

Q6.3: Requestors receive written confirmation of the accommodation arrangement before the live training event (or within five business days of the request for on-demand content), including the name of the captioning contact, how to access the caption display, and what to do if the caption is not visible at session start. 0 = no written confirmation process; 1 = confirmation sent sometimes or informally; 2 = written confirmation process for every request.

Q6.4: The accommodation process covers both asynchronous (on-demand video) and synchronous (live training) content, with appropriate timeframes stated for each type in the written process documentation. An employee requesting accessible video on demand and an employee requesting CART for a live session both have a documented process for handling their request. 0 = process covers neither type consistently; 1 = process covers one type (usually live) but not the other; 2 = process covers both, with type-specific timeframes documented.

Q6.5: The process includes a documented decision path for late accommodation requests — requests submitted inside the CART vendor’s lead-time window. The documented decision path includes at least two options (best-efforts fulfilment and a post-event accessible alternative) and defines who has authority to make the call on short notice. 0 = no documented late-request path; 1 = handled ad hoc with no documentation; 2 = written decision path with options and authority defined.

Domain 6 score: ___ / 10

Domain 7: Back-catalogue remediation

Back-catalogue remediation — captioning training video that was produced before your current caption policy existed — is one of the most common gaps in mid-market L&D caption programmes and one of the most frequently surfaced in OCR complaint investigations. The large-scale caption backlog remediation playbook covers how to execute a remediation project. Domain 7 assesses whether your programme has the planning infrastructure in place: do you know what you have, have you prioritised it, and has budget been committed?

Maximum domain score: 10

Q7.1: You have conducted a formal audit of your video library that identifies content lacking WCAG-compliant captions — including content that has auto-generated captions with no accuracy review, which does not meet the WCAG 2.1 AA standard. The audit result is a documented inventory, not an estimate. 0 = no formal audit; 1 = informal or partial inventory (new content only, or titles without reviewing caption status); 2 = documented inventory with caption status for each item, including auto-caption vs. reviewed-caption distinction.

Q7.2: Uncaptioned content has been risk-prioritised using a framework that considers compliance category (mandatory content first), content age and traffic (high-traffic recent content before low-traffic archive), and regulatory context (content in regulated industries with near-term examination risk before general content). 0 = not prioritised; 1 = rough categorisation by content type; 2 = documented priority framework with content ranked or tiered.

Q7.3: A written remediation plan exists for the highest-risk uncaptioned content, with a target completion date, a named owner, and a budget or cost estimate. The plan is not aspirational — it is a project plan with milestones. 0 = no plan; 1 = informal intent or unwritten understanding; 2 = written project plan with date, owner, and cost estimate.

Q7.4: Budget has been identified for back-catalogue remediation, at a volume and per-minute cost that accounts for the accuracy standard required (captioning at 99%+ accuracy is not the same price per minute as auto-captioning with no review). The budget figure is realistic for the volume of uncaptioned content identified in the audit. 0 = no budget identified; 1 = informal estimate discussed but not committed; 2 = budget allocated in this or next fiscal year, at a per-minute cost that reflects actual vendor pricing for reviewed captioning.

Q7.5: Your remediation plan includes a process for content that is updated after captioning: when a training video is revised, corrected, or rebranded, the caption file is updated or re-processed rather than having the old caption file applied to the new audio without review. 0 = update-to-caption lifecycle not considered; 1 = planned but not yet implemented; 2 = process in place, tested, and documented.

Domain 7 score: ___ / 10

Domain 8: Governance and policy

Caption programme governance is what converts individual compliance actions into a sustainable, defensible programme. The caption programme governance policy template covers what a written caption policy should include. The caption compliance reporting to leadership guide covers how to communicate programme status. Domain 8 assesses whether your caption programme has the governance infrastructure to hold up over time — through staff turnover, technology changes, regulatory updates, and regulatory reviews.

Maximum domain score: 10

Q8.1: A written caption policy exists that applies to all training content produced by or for your organisation, regardless of format — pre-recorded video, live training, screen recording, audio-only learning, AI-generated video (Synthesia, HeyGen), and any content produced by external vendors for delivery in your LMS. 0 = no written policy; 1 = draft policy or policy that covers some formats but not all; 2 = written policy in force that explicitly covers all content types, signed by a responsible leader.

Q8.2: The caption policy names a responsible owner — a specific role title, not “the L&D team” generically — who is accountable for caption programme compliance, vendor management, and reporting. The owner is aware of this responsibility. 0 = no named owner; 1 = role implied but not documented in policy; 2 = specific role named in policy with explicit accountability.

Q8.3: Content producers — instructional designers, video producers, facilitators, SMEs who record their own content, and external production vendors — have received documented guidance on the caption requirements that apply to their work. This includes what accuracy standard is required, how to submit content for captioning, and what to do if they have questions about an edge case. 0 = no guidance distributed; 1 = guidance available (e.g., on an intranet page) but not proactively distributed; 2 = guidance distributed to all current producers, with acknowledgment tracked.

Q8.4: Caption compliance status is reviewed on an annual cycle and the results are reported to programme leadership — L&D director, CLO, VP People, or equivalent — in a format that enables resource allocation decisions. The annual review covers coverage status, accuracy data, vendor performance, accommodation request volume, and remediation progress. 0 = no annual review; 1 = informal internal review not reported to leadership; 2 = documented annual review with leadership report.

Q8.5: The caption policy is reviewed and updated on a defined schedule (at least every 24 months) and at any major regulatory trigger — a new OCR guidance document, a WCAG version update, a new state accessibility law, a change in your LMS platform’s caption handling, or a vendor transition. The date of the last review is documented in the policy. 0 = policy has never been reviewed since creation or was created once and forgotten; 1 = reviewed once informally; 2 = scheduled 24-month review cycle documented in policy, with the date of last review on record.

Domain 8 score: ___ / 10

Score calculation and maturity band interpretation

Add up your eight domain scores (each out of 10) for your total score out of 80.

Priority action guide by domain score

Total score tells you your overall maturity band. Domain scores tell you where to act first. Use this guide to prioritise:

If Domain 1 (caption coverage) is below 6/10

Coverage is the foundational domain. No other investment closes a regulatory gap if the content itself is uncaptioned. The first action is to activate a policy requiring captions on all new content (Q1.2) and complete a library audit to understand back-catalogue gap volume (Q1.3). Without this baseline, the other domain investments are optimising a partial programme.

If Domain 4 (live training captioning) is below 4/10 and you deliver any instructor-led training

Live training gaps create the most immediate accommodation failures because they happen in real time, in front of a participant, with no opportunity to repair the failure after the event. A missing CART feed-failure protocol (Q4.4) becomes a visible incident during a mandatory compliance training session. The accommodation request process gap (Q4.1, Q4.2) becomes an ADA finding the moment a participant submits a request and the organisation cannot fulfil it. Address Domain 4 as the second priority after coverage, before back-catalogue remediation work.

If Domain 5 (vendor accountability) is below 6/10

Vendor contract terms are the highest-return compliance investment for most mid-market programmes: a contract amendment adding an accuracy SLA, audit rights, and reference transcript retention provision costs a conversation with the vendor account manager, not additional programme spend. If you are in a contract renewal period, addressing Domain 5 is the single highest-priority action in this guide. The vendor SLA contract review checklist has the specific language to request.

If Domain 7 (back-catalogue remediation) is below 4/10 and you have significant legacy content

OCR complaint investigations routinely focus on high-traffic mandatory compliance content that is uncaptioned — particularly orientation content, annual compliance certifications, and safety training that employees are required to complete. If your Domain 7 score reflects the absence of an audit and a plan, the highest-risk items in your library are unidentified. Completing the library audit (Q7.1) and risk prioritisation (Q7.2) is the first step, even before addressing Domain 5 or 8, if you have significant legacy content.

If Domain 8 (governance and policy) is below 6/10

Without a written policy and a named owner, every improvement made in other domains is fragile. Staff turnover removes the informal knowledge that kept the programme running; vendor relationships are renegotiated without anyone knowing what the previous contract required; new content producers start making videos without knowing that captions are required. Domain 8 improvements lock in the gains from every other domain. If Domain 8 is below 6, completing the written policy (Q8.1) and naming the owner (Q8.2) should happen in parallel with whichever other domain is the highest operational priority.

If Domain 6 (accommodation request process) is below 4/10

The accommodation request process is the most immediate employee-facing compliance layer. A missing accommodation log (Q6.2) is the first thing an EEOC investigator requests. The absence of a written confirmation to the requestor (Q6.3) becomes an issue when a participant claims they submitted a request and received no response. If Domain 6 is below 4, establishing the log and confirmation process is a one-to-two-day implementation effort that should be completed before the next live training event.

After you score: what to do with the results

The self-assessment is not the deliverable — it is the input to a prioritised action plan. A few suggestions for what to do with your score:

Share the domain breakdown, not just the total score

When you report the self-assessment results to leadership, the domain breakdown is more useful than the total score. A total of 48/80 with Domain 5 at 2/10 tells leadership “we need a contract amendment” — a specific, bounded, and achievable action. A total of 48/80 with no domain breakdown tells leadership “we are 60% compliant” — which does not suggest any particular action. The caption compliance reporting guide covers how to frame programme status for leadership audiences who are not specialists in WCAG requirements.

Set a 6-month target score before you report

Before presenting the baseline score to leadership, identify the target domain scores you expect to achieve in six months based on the investments you are recommending. This converts the self-assessment from a compliance exposure report into a programme investment roadmap. Leadership is more likely to allocate resources when the request is “we are at 48/80; here is the plan to reach 60/80 in six months with these investments” than when the request is “we have compliance gaps.”

Re-run the assessment annually as part of the programme review

The annual programme review is the natural home for the annual re-run of this self-assessment. Adding the domain score comparison (year-over-year) to the annual review report gives leadership a longitudinal view of programme progress and creates an accountability mechanism for the programme owner.

Eight failure modes this assessment is designed to catch

1. The auto-caption completeness illusion

The LMS reports 100% of content as “captioned” because auto-captions are enabled by default and counted as captions in the metadata. Q1.1 and Q2.1 catch this: auto-generated captions with no accuracy review do not meet WCAG 2.1 AA and should not be counted as compliant coverage in the assessment.

2. The new-content-only policy gap

A caption policy written in 2024 applies to all new content produced after that date — and leaves 80% of the video library, produced between 2018 and 2023, outside the policy’s scope. Q1.3 and the entire Domain 7 section catch this: the policy date does not determine compliance status for content that already exists.

3. The vendor accuracy claim problem

The vendor contract says “99% accuracy.” The vendor provides a QA report showing 99% accuracy on their own test content. Your content contains product names, SDK terms, and regulatory acronyms that the vendor has never seen. Q2.4 and Q5.5 catch this: vendor accuracy claims must be independently verified on your specific content, not on the vendor’s demonstration material.

4. The mobile rendering blindspot

Captions work perfectly in the desktop LMS browser and break completely in the iOS mobile app, which is where 35% of your learners access training. Q3.5 catches this: technical conformance must be verified across all learner access environments, not just the developer’s testing environment.

5. The live training accommodation vacuum

A learner with a hearing disability registers for a live virtual training session. The LMS course listing has no accommodation information. They email the L&D team two days before the session. The vendor needs seven business days. There is no documented late-request protocol. The session runs without captioning, and the accommodation request becomes an ADA finding. Q4.1, Q4.2, and Q6.5 catch this before it happens.

6. The contract expiry blind spot

The vendor contract with the accuracy SLA and audit rights clause expires. The vendor sends a renewal with standard terms — no accuracy SLA, no audit rights. The L&D team auto-renews because the vendor relationship is working. A year later, a regulatory examination requests accuracy documentation and the vendor has no contractual obligation to provide it. Q5.1 and Q5.2 catch this: domain scores should be re-run at each vendor contract renewal to confirm the new contract terms meet the programme’s requirements.

7. The governance-by-person-not-process problem

The caption programme works well because Jane, the senior instructional designer, personally ensures every video is captioned before it goes live. Jane leaves. Her replacement does not know about the caption requirement because it was never written down. Six months later, 40 new videos have been published without captions. Q8.1 and Q8.3 catch this: programme compliance must be documented in policy and distributed to producers, not held in one person’s institutional knowledge.

8. The back-catalogue arithmetic surprise

The L&D team estimates they have about 50 uncaptioned videos from before 2022 and plans to caption them over the next quarter. The library audit reveals 340 videos, not 50. At a realistic captioning rate of 15 minutes of production time per minute of audio (for reviewed, accurate captioning), 340 videos at an average of 12 minutes each is 1,020 hours of captioning work — roughly six months of full-time captioner time or significant vendor spend. Q7.1, Q7.3, and Q7.4 catch this: back-catalogue remediation planning requires a real audit, not an estimate.

Frequently asked questions

We have a small L&D team of two people. Is a scored self-assessment like this realistic for us?

The scoring is calibrated for the full range of caption programme maturity — a two-person L&D team at a 75-person company should not expect to score the same as a ten-person accessibility team at a large public university. A realistic two-year target for a small L&D team is 40–48/80: written policy, coverage of all new content and highest-risk legacy content, one vendor with a basic accuracy SLA, and a documented accommodation request process. The self-assessment helps a small team identify the three or four highest-priority items rather than trying to achieve every domain simultaneously.

Our LMS vendor says their auto-captions are WCAG compliant. Why would we score Q1.1 below 2?

LMS vendors often describe their auto-caption feature as “WCAG compliant” in the technical sense: the captions are synchronised, toggleable, and formatted. What they do not claim — and what WCAG 2.1 SC 1.2.2 actually requires — is that the captions are “equivalent” to the audio. Equivalence requires accuracy. Auto-generated captions at 80–90% accuracy are not equivalent to the audio when specialised vocabulary is involved. Score Q1.1 based on whether the captions on your content are accurate, not on whether your LMS supports a caption feature.

We do not have a CART vendor currently. Can we still score Domain 4?

Yes. If you do not currently deliver any live instructor-led training, Domain 4 is not applicable — score it zero but note it as N/A and exclude it from your total. If you do deliver any live training, Domain 4 applies even if no participant has yet requested a captioning accommodation. The accommodation infrastructure must be in place before a request arrives, not assembled in response to one.

How does this self-assessment relate to an EEOC or OCR audit? Is it a legal document?

This self-assessment is a programme improvement planning tool, not a legal audit. It is not a substitute for legal advice and does not create legal compliance. What it does is identify gaps that would likely be exposed in an EEOC charge investigation or OCR complaint review, and prompt the L&D team to address those gaps before a regulatory inquiry begins. If you receive a complaint or examination request, consult legal counsel. The self-assessment record should be treated as an attorney-client communication if it is prepared in anticipation of litigation.

We use a SaaS captioning tool rather than a vendor. How do we score Domain 5?

Score Domain 5 against your SaaS tool’s terms of service and the tool’s documentation of accuracy. Q5.1: does the tool’s terms include an accuracy commitment with a defined methodology? Q5.2: do the terms give you the right to request processing logs or accuracy records for your content? Q5.3: does the tool retain the source audio and reference transcripts for your content for at least three years? Q5.4: what is the remediation process if accuracy is below the stated level? Q5.5: have you independently verified the tool’s accuracy on your specific content? Most SaaS self-service captioning tools score low on Domain 5 because their terms of service are not designed for the regulatory accountability requirements of L&D in regulated industries.

We have an accessibility coordinator. Does Domain 6 automatically score high?

Not automatically. Domain 6 assesses the specific documentation and process infrastructure around accommodation requests, not just whether a coordinator role exists. Q6.2 (accommodation request logging with required fields) and Q6.5 (late-request documented decision path) can be absent even when an accessibility coordinator is in the role. Run through each question individually. A coordinator who is excellent at handling requests informally but has not built the documentation layer will score lower on Domain 6 than expected — and that gap is worth surfacing, because the coordinator cannot be the sole repository of accommodation history.

We scored 32/80 on the first assessment. How do we prioritise improvement without overwhelming the team?

At a score of 32/80 (Developing band), focus on two parallel workstreams. The first is infrastructure: write a caption policy (Q8.1), name an owner (Q8.2), establish an accommodation request log (Q6.2), and identify a CART vendor (Q4.3). These are process and documentation actions that cost time, not budget. The second is coverage: identify the compliance-mandatory content in your library (Q1.4) and prioritise getting those items captioned at 99%+ accuracy, even if the rest of the library remains uncaptioned for now. A compliance-mandatory corpus at 100% coverage and a written policy and accommodation log will move your score to approximately 42–48 in one quarter without a large budget increase — and will significantly reduce your regulatory exposure in the content that matters most.