Compliance Operations · Published 2026-06-12
Caption compliance KPIs and leadership reporting: what to measure monthly, the audit-readiness dashboard, and the metrics that survive a WCAG investigation
There is a layer of caption compliance work that most L&D teams never build: the reporting layer. The audit methodology gives you a baseline. The QA methodology gives you a per-video accuracy measurement. The accessibility coordinator runs the programme month to month. The 90-day programme build gives you the operational framework. But none of these address the question that comes up when a VP of Learning asks for a programme update, when a CLO needs to present compliance status to the board, or when an OCR investigator arrives and asks to see twelve months of documentation: what are the numbers, what do they mean, and what is the paper trail that proves the programme is real? The reporting layer is where the captioning programme becomes legible to people who do not run it. Without it, a programme that is operationally sound looks identical to a programme that exists only on paper. An investigator cannot distinguish between an organisation that runs a rigorous monthly DCMP spot-check and one that just uploads videos with auto-captions if neither organisation has documentation. The paper trail is the programme, from a regulatory standpoint. Building the reporting layer is not a separate project from running the programme — it is a measurement discipline that runs in parallel with operational execution, costs roughly two hours per month when the measurement protocol is defined in advance, and produces documentation that serves three audiences simultaneously: the accessibility coordinator as a management tool, leadership as a programme visibility artefact, and investigators as a compliance trail. This post covers the five KPI categories that constitute a complete caption compliance measurement framework, the monthly measurement protocol that generates the data for each KPI in under two hours, the twelve-row dashboard structure that makes the monthly numbers actionable, the one-slide leadership model that converts programme data into executive visibility, the specific metrics and documentation that survive a WCAG or ADA Title II investigation, LMS-specific reporting notes for the eight platforms where pulling caption inventory data is non-trivial, eight failure modes in compliance reporting that undermine otherwise sound programmes, and a seven-question FAQ covering the reporting questions that L&D accessibility coordinators encounter most often in practice.
TL;DR
A complete caption compliance measurement framework has five KPI categories: library coverage rate (% of the in-scope video library with WCAG-compliant captions — the primary outcome metric), new-content submission rate (% of new videos submitted for captioning before publication — the leading indicator that predicts whether the programme is gaining or losing ground), accuracy pass rate (% of spot-checked videos that pass the 99% WCAG 2.1 AA threshold on a DCMP spot-check — the quality gate), remediation velocity (median days from flagging to completion — the measure of whether the programme can close its own findings), and exception programme health (count and age of open exceptions by category — the measure of whether the governance policy is being administered or ignored). These five categories generate twelve reportable metrics, which compose a monthly dashboard that takes roughly two hours to populate when the measurement protocol is defined in advance. The dashboard has three audiences: the accessibility coordinator (management tool), leadership (programme visibility — use the one-slide, four-metric RAG model), and investigators (compliance trail — the dashboard plus the governance policy plus the monthly measurement log is the complete documentation package that answers the four questions investigators consistently ask). Organisations that report monthly against this framework, retain the dashboard rows for twelve months, and document their measurement methodology have the full investigator-ready package. Organisations that report ad hoc, discard monthly data, or confuse library coverage rate with compliance rate are missing the paper trail that separates a real programme from a paper one under investigation.
Why the reporting layer is distinct from programme operation
Running a captioning programme and reporting on a captioning programme are different activities. Running the programme means ensuring videos get captioned, accuracy is checked, exceptions are administered, the glossary is maintained. Reporting on the programme means converting those operational activities into numbers that are meaningful to audiences who do not run the programme and may not know how it works. The distinction matters because most L&D teams conflate the two: they treat the existence of a programme as sufficient evidence that the programme is working, and they treat the absence of complaints as evidence that coverage is adequate. Both conflations fail under scrutiny. An investigator who asks for programme documentation is not asking whether the programme exists — they are asking whether it can be measured, and whether the measurements have been retained.
The reporting layer is also where the programme generates executive visibility, and executive visibility is what generates programme continuity. Caption compliance is not intrinsically compelling to a VP of Finance or a CLO who is managing twenty competing budget priorities. The ROI framing makes the budget case. The reporting layer makes the continuation case: it gives leadership a mechanism to see that the programme is doing what it was funded to do, that it is gaining ground on the backlog, that the submission rate is holding, that no complaints have arrived. Without monthly reporting, the captioning programme disappears from leadership attention after the kickoff. The next time it surfaces is when something goes wrong — a complaint, an audit, a question from a board member about ADA compliance. At that point, having twelve months of dashboards and documentation changes the outcome of the conversation.
The reporting layer is also the piece that connects the individual operational activities — the quarterly audit, the monthly QA spot-check, the producer submission cadence, the exception log review — into a coherent programme narrative. Without reporting, each of those activities is isolated. The QA spot-check produces a list of accuracy failures; the exception log tracks approved exceptions; the audit produces a coverage count. The reporting layer synthesises these into metrics that answer the question the coordinator and the CLO and the investigator all share: is the programme working?
The gap between having captions and having a programme
The most common failure in caption compliance reporting is the coverage-rate fallacy: the belief that measuring the percentage of videos with captions is sufficient to characterise the compliance programme. Library coverage rate is the right outcome metric, but it is not a sufficient reporting framework by itself, for three reasons. First, it is a lagging indicator: it reflects decisions made one to six months ago, not what the programme is doing right now. An organisation at 94% coverage today could be at 86% in six months if the submission rate has been running at 70% — but the coverage rate will not show that signal for months. Second, it does not distinguish between captions that meet the accuracy standard and captions that do not. A video with auto-generated YouTube captions at 82% accuracy counts as "captioned" in an LMS inventory pull; it fails the WCAG 2.1 AA 99% threshold. Third, it does not capture the programme's ability to sustain itself. A programme at 94% coverage with a 70% submission rate is a programme that will regress. A programme at 85% coverage with a 97% submission rate is a programme that will reach 100% within six months. These two programmes look different only when you measure submission rate alongside coverage rate. The reporting framework below measures both, plus the three additional KPI categories needed to give leadership and investigators a complete picture.
The five KPI categories for caption compliance
A complete caption compliance reporting framework has five KPI categories. Each category addresses a distinct question about the programme's performance. Together they form a measurement set that is sufficient for management reporting, leadership visibility, and investigator documentation. No single KPI from the five is optional: omitting any one of them leaves a gap that is either a management blind spot (invisible programme decay) or an investigator blind spot (missing documentation for a question investigators consistently ask).
KPI Category 1: Library coverage rate
Library coverage rate is the percentage of the in-scope video library that has WCAG-compliant captions — captions that meet the 99% accuracy standard at the word level, are synchronised to within two seconds of the corresponding audio, and include speaker identification where multiple speakers are present. This is the primary outcome metric for the caption compliance programme. It answers the question "what fraction of our obligation have we discharged?" directly.
Measuring library coverage rate requires two inputs: the total count of in-scope videos in the LMS and the count that have compliant captions. The first input comes from an LMS inventory pull (platform-specific notes in the LMS reporting section below). The second input requires a distinction that most LMS platforms do not make automatically: the LMS inventory will tell you which videos have a caption file attached, but not whether that caption file meets the accuracy standard. The caption field in the LMS is binary — present or absent — and a caption file with 80% accuracy satisfies that binary as well as one with 99% accuracy. The compliant-captions count therefore requires cross-referencing the LMS inventory with the accuracy log: videos where the caption file was produced by a WCAG-compliant process (vendor with 99% contractual SLA plus spot-check passing) or where the most recent DCMP spot-check returned a pass result.
Target: 100% for videos published before the applicable compliance deadline (ADA Title II for public universities and government entities, EAA for EU-market content, Section 508 for federal content, ADA Title III for customer-facing content from public accommodations). For newly produced content, the operational target is 100% on the new-content gate (no video published without captions), with a short processing window accommodating the caption-production SLA — typically 24-72 hours depending on vendor turnaround. The dashboard tracks both dimensions separately: historical backlog coverage and new-content coverage. Conflating them produces a single number that obscures which part of the problem is solved and which is not.
Report frequency: monthly. The coverage rate for the backlog moves slowly — a few percentage points per month as backlog remediation closes historical items. Measuring it more frequently than monthly creates reporting noise without useful signal. Measuring it less frequently misses trend data that is important for month-over-month leadership reporting.
KPI Category 2: New-content submission rate
New-content submission rate is the percentage of videos published in the reporting period that were submitted for captioning through the programme's defined process before or at the time of publication. This is the most important leading indicator in the framework: it predicts whether the coverage rate will improve, hold, or decay in future periods. An organisation with a 90% library coverage rate and a 70% submission rate for new content is losing ground: every month, 30% of newly published videos add to the uncaptioned backlog. An organisation with an 85% library coverage rate and a 97% submission rate is gaining ground: the backlog is shrinking, and the new-content pipeline is almost clean.
Measuring submission rate requires a submission log — a record of every video submitted for captioning in the reporting period, with the submission date, the publication date, and whether the video was submitted before or at publication. The submission log is a natural output of the captioning workflow if the workflow is designed to generate it: the LMS can be configured to require a caption ticket reference before a video is published, or the captioning vendor can generate a submission log as part of its monthly invoice. Programmes that do not have a submission log have to reconstruct it from LMS publish timestamps and vendor delivery timestamps — which is possible but adds one to two hours to the monthly measurement process.
Target: 95%+. The 5% allowance accommodates the short exception window — videos that were submitted within 48 hours of publication under an emergency exception, or that were published outside the normal workflow due to a system outage. The 5% is not an accuracy tolerance; it is a process-integrity tolerance. Programmes that run at 90-94% submission rate are operating in a range where the submission gap compounds against the coverage rate over six to twelve months. Programmes below 85% submission rate are losing ground faster than backlog remediation can recover.
The submission rate is the metric that the change management post addresses at the producer level. From the reporting perspective, submission rate below target is a signal that something in the producer adoption layer has broken: either the workflow is generating friction that producers are routing around, or scope ambiguity is allowing producers to self-classify their content as out-of-scope, or accountability is absent and non-submission has no visible consequence. The submission rate gives leadership the signal; the change management framework gives the coordinator the diagnostic to identify which of the three structural gaps is causing the shortfall.
KPI Category 3: Accuracy pass rate
Accuracy pass rate is the percentage of spot-checked videos that meet the 99% WCAG 2.1 AA accuracy threshold on a DCMP Captioning Key evaluation. This metric answers the quality question: of the videos that have captions, what fraction meet the accuracy standard? The library coverage rate measures the quantity dimension of compliance; the accuracy pass rate measures the quality dimension. Both are necessary. A programme at 100% library coverage with a 70% accuracy pass rate is not a compliant programme — it has captions on every video, but 30% of those captions do not meet the standard they need to meet to discharge the legal obligation.
The measurement methodology matters for this KPI because the methodology is what makes the accuracy pass rate documentable. The DCMP spot-check protocol defines a specific evaluation methodology: select a random 60-second segment from the video, count all words in the audio, count all errors (substitutions, insertions, deletions, formatting failures) in the caption for that segment, calculate the word error rate, determine pass (error rate ≤1%) or fail (error rate >1%). The reason the methodology must be documented is that an accuracy claim without a methodology is not an accuracy claim at all — it is a vendor assertion. An investigator who asks "how do you know your captions are 99% accurate?" wants an answer that describes a measurement process with a defined sample, a defined error taxonomy, a defined threshold, and a defined remediation trigger. An organisation that answers "our vendor guarantees 99% accuracy in their contract" has no measurement process and no documentation of independent verification. An organisation that answers "we run a monthly DCMP spot-check on a random 5% sample using the methodology described in our QA policy, retain the results in our accuracy log, and remediate any video that fails within 14 days" has both.
Target: 95% pass rate on the spot-check sample. This means 95% of the videos in the 5% random sample pass the DCMP evaluation — not that 95% of all videos in the library are compliant. The 5% failure rate on the spot-check is the remediation trigger, not a compliance claim. Videos that fail go into the remediation queue. Spot-check results are retained as documentation. Report frequency: monthly (concurrent with the library coverage measurement).
KPI Category 4: Remediation velocity
Remediation velocity is the median calendar days from flagging to completion for items in the remediation queue. The remediation queue includes two types of items: accuracy failures identified in spot-check (videos that have captions but failed the DCMP evaluation) and submission failures (videos that were published without going through the captioning process). Remediation velocity measures the programme's ability to close its own findings. A programme that generates findings faster than it closes them is not self-sustaining: the queue grows, coordinator attention fragments, and the backlog remediation plan falls behind because active remediation is competing with historical remediation for vendor capacity and coordinator hours.
Target: median ≤14 calendar days for accuracy-fail remediation (the caption file exists; it needs to be corrected or replaced — a 14-day turnaround is achievable with most vendor SLAs) and ≤30 calendar days for submission-fail remediation (the video was never submitted; it needs to go through the full captioning process — the 30-day window accommodates vendor queuing and glossary adaptation for domain-specific content). Programmes that run accuracy-fail remediation above 21 days are typically blocked by a vendor SLA issue or an internal approval step that was not accounted for in the governance policy. Programmes that run submission-fail remediation above 45 days are typically blocked by producer cooperation issues (the content owner has not provided the source video file) or by backlog volume that exceeds vendor monthly capacity.
Remediation velocity is the metric that leadership finds most actionable because it has a clear operational lever: if the median is above target, something is blocking the remediation pipeline, and identifying the blocker is a solvable management problem. It is also the metric that investigators find most probative: an organisation that identifies accuracy failures and closes them within 14 days has a functioning quality loop. An organisation that identifies failures and takes three months to close them has a quality loop that does not work.
KPI Category 5: Exception programme health
Exception programme health measures whether the exception procedure in the governance policy is being administered correctly. The governance policy defines four exception categories, an approval authority, duration limits per category, and a mandatory exception log. Exception programme health has two metrics: the count of open exceptions by category and the count of exceptions that have exceeded their approved window.
The count of open exceptions by category is a signal of whether the exception procedure is being used as designed. A healthy exception log has a small number of short-duration exceptions (typically three to eight open at any given time in an organisation producing thirty to fifty videos per month, almost all in the standard-window or extended-window categories). An exception log with twenty open exceptions, half of which are in the maximum-window category (30 days), is a signal that the exception procedure is functioning as a permanent workaround rather than a temporary accommodation — a pattern the governance policy is specifically designed to prevent.
The count of exceptions exceeding their approved window is a binary compliance metric within the governance framework: the target is zero. An exception that has exceeded its approved window is a video that has been operating outside the captioning standard for longer than the governance policy authorises. The governance policy specifies what the response should be (remediate immediately or bring to coordinator for review with documented rationale for extension) — the exception programme health metric is what surfaces the situation so the coordinator can take that action before it accumulates. Report frequency: monthly (the exception log review takes twenty to thirty minutes and is the fastest of the five KPI categories to measure).
Monthly measurement protocol: how to run the numbers in under two hours
The five KPI categories generate twelve reportable metrics. Measuring all twelve from scratch each month takes four to six hours in an organisation without a defined protocol. With a defined protocol and the right LMS exports pre-configured, the same measurement takes under two hours. The difference is almost entirely in preparation: knowing exactly which LMS report to pull, which fields to filter, which counts to extract, and in what order, eliminates the navigation and improvisation time that dominates the first few measurement cycles.
The protocol below is an eight-step process. Steps one through four cover library coverage and accuracy (the outcome metrics). Steps five and six cover submission rate and remediation velocity (the leading indicator and the operational health metric). Steps seven and eight cover exception programme health and dashboard compilation. The steps are ordered to allow parallelisation: steps one and two can run simultaneously with the vendor data pull in step five if two people are involved, or they can run sequentially in the order below for a solo coordinator.
Step 1: Pull the LMS inventory (30 minutes)
Log into the LMS admin panel and export the video content inventory for the reporting period. The export should include: video ID, title, category/course assignment, publish date, caption file status (present/absent), and duration. Filter to in-scope content only — exclude content types that fall outside the programme's coverage mandate (see the governance policy's content gate definition). This is the most time-variable step in the protocol because LMS platforms differ significantly in how they expose caption status data (platform-specific notes below).
Calculate two numbers from the inventory export: total in-scope video count and captioned-video count (videos with a caption file attached). These are inputs to the library coverage rate calculation in step three. Do not calculate the compliant-captioned count yet — that requires cross-referencing with the accuracy log in step three.
Step 2: Pull the accuracy log (15 minutes)
Open the accuracy log — the running record of DCMP spot-check results maintained by the coordinator. The accuracy log should have one row per spot-checked video with: video ID, check date, sample segment (start time, duration), word count, error count, error rate, pass/fail result, and remediation status if fail. Pull all rows from the current reporting period and the prior twelve months.
From the current period's rows, calculate the spot-check sample count and the pass count. From the historical rows, identify any videos that have a recorded pass result from a spot-check within the past twelve months — these are the "compliant-captioned" videos for the coverage rate calculation. Videos with a caption file but no accuracy log entry are captioned but not verified — they should be counted as captioned but flagged for the next spot-check cycle, not counted as compliant.
Step 3: Calculate library coverage metrics (15 minutes)
Using the LMS inventory export (step 1) and the accuracy log (step 2), calculate three metrics:
- Library coverage rate (all captions): captioned-video count ÷ total in-scope video count. This is the headline coverage number that will appear on the dashboard and in leadership reporting.
- Library coverage rate (verified compliant): compliant-captioned count (videos with a caption file AND a passing accuracy log entry within 12 months) ÷ total in-scope video count. This is the more conservative and more defensible compliance number — it is what a programme should be able to show investigators.
- Accuracy pass rate: passing spot-check results this period ÷ total spot-check results this period. If no spot-checks were run this period, record "no spot-check this period" — this is a reportable gap, not a zero.
The delta between the two coverage rates is the "unverified coverage" gap — videos that have captions but have not been spot-checked. A large gap (more than 20% of captioned videos unverified) indicates that the spot-check protocol is not keeping pace with the library growth rate and the sample should be increased.
Step 4: Pull the remediation queue (15 minutes)
Open the remediation queue — the running log of videos flagged for remediation (either accuracy fails or submission fails). The queue should have one row per flagged video with: video ID, flag date, flag type (accuracy fail or submission fail), assigned remediation action, remediation start date, remediation completion date, and days-to-close (if closed). Pull all rows with a flag date in the reporting period or prior and a completion date either in the current period or not yet completed.
Calculate three metrics: open remediation item count (accuracy fails + submission fails separately), median days-to-close for items closed in the current period (accuracy fails and submission fails separately), and oldest open item age in calendar days. The oldest open item age is a supplementary metric that does not appear on the main dashboard but should be surfaced in the coordinator's internal notes — a single item that has been open for 90 days is more significant than ten items averaging 14 days.
Step 5: Pull the submission log (20 minutes)
Pull the submission log for the reporting period — the record of every video submitted through the captioning process. The submission log should have: video ID, submission date, publication date, submission-before-publication flag, and any exception reference if the video was published before captioning under an approved exception. Cross-reference the submission log with the LMS inventory export (step 1) to identify the denominator: all videos published in the period. Videos in the LMS inventory that are not in the submission log (and not in the exception log with an approved exception) are submission failures — they go into the remediation queue.
Calculate the submission rate: videos submitted (through process or under approved exception) ÷ videos published in the period. Note: videos that were submitted on the day of publication count as submitted even if caption delivery occurred after publication, as long as the submission log shows the submission timestamp preceded the publish timestamp. Videos submitted the day after publication are submission failures regardless of how quickly the caption was delivered — the programme's new-content gate applies at submission, not at delivery.
Step 6: Review the exception log (20 minutes)
Open the exception log and review all open exceptions. For each open exception, check whether the exception is still within its approved window (7 days for emergency exceptions, 14 days for standard extensions, 30 days for maximum extensions per the governance policy). Flag any exception that has exceeded its window as a governance policy breach requiring immediate coordinator action. Calculate open exception count by category and the count exceeding their approved window.
For the quarterly exception pattern review (which generates the exception log amendment recommendations per the governance policy), note whether any exception category is generating a disproportionate volume of exceptions — a signal that the content gate may need adjustment or that a specific content type needs a permanent category revision. This analysis does not appear on the monthly dashboard but should be documented in the coordinator's monthly notes and surfaced at the quarterly review.
Step 7: Compile the twelve-row dashboard (15 minutes)
Enter the calculated metrics into the monthly dashboard (template in the next section). Compare current-month values to prior-month values and to target values. Apply RAG status (Red/Amber/Green) based on the criteria below. Save the dashboard row to the compliance log — the running record of monthly dashboard rows that constitutes the twelve-month documentation trail required for investigator-ready documentation.
Step 8: Draft the leadership report (10 minutes)
Using the dashboard, draft the four-metric leadership summary (the one-slide model, described below). For months where a metric is Red or where a metric has declined significantly month-over-month, include a one-sentence explanation of the cause and a one-sentence description of the corrective action underway. The leadership report should be a five-minute read; it should never require the recipient to understand the DCMP spot-check protocol to interpret the numbers.
The audit-readiness dashboard: twelve-row structure
The monthly dashboard is a twelve-row table that captures all five KPI categories in a single view. Its primary audience is the accessibility coordinator — it is the coordinator's management tool, not a leadership deliverable. The leadership deliverable is derived from it (the one-slide model), but the dashboard itself is the operational record. Retain every monthly row for a minimum of twelve months; retain rows for the entire programme history for programmes that have had a complaint or investigation.
The twelve rows are:
| Row | Metric | KPI Category | Target | RAG: Green | RAG: Amber | RAG: Red |
|---|---|---|---|---|---|---|
| 1 | Library coverage rate (all captions) | Coverage | 100% | ≥95% | 85–94% | <85% |
| 2 | Library coverage rate (verified compliant) | Coverage | 100% | ≥90% | 75–89% | <75% |
| 3 | New-content submission rate | Leading indicator | 95%+ | ≥95% | 85–94% | <85% |
| 4 | Accuracy pass rate (DCMP spot-check) | Quality | 95%+ of sample | ≥95% | 85–94% | <85% |
| 5 | Accuracy remediation queue (open, count) | Quality / Remediation | 0 | 0–5 | 6–15 | >15 |
| 6 | Accuracy remediation velocity (median days) | Remediation | ≤14 days | ≤14 days | 15–21 days | >21 days |
| 7 | Backlog remediation queue (open, count) | Remediation | Decreasing | Decreasing MoM | Stable MoM | Increasing MoM |
| 8 | Backlog remediation velocity (videos closed this period) | Remediation | ≥30/month | ≥30 | 15–29 | <15 |
| 9 | Open exception count | Exception health | Low | 0–8 | 9–20 | >20 |
| 10 | Exceptions exceeding approved window | Exception health | 0 | 0 | 1–2 (being resolved) | ≥3 or unacknowledged |
| 11 | Complaints / regulatory notices received | Compliance event | 0 | 0 | — | ≥1 (any) |
| 12 | Measurement protocol executed this period | Process integrity | Yes | Yes | Partial | No / skipped |
Row 12 is the most important row on the dashboard for investigator purposes: it records whether the monthly measurement was executed. A dashboard with eleven Green rows and row 12 showing "skipped" for three of the last twelve months is a programme with documentation gaps. An investigator reviewing the twelve-month trail will ask why the measurement was skipped and whether the programme was operating during those months even without measurement. Row 12 is also the row that the accessibility coordinator most often wants to leave out of the leadership report — it makes the measurement burden visible. It should stay in the coordinator's compliance log regardless of whether it appears in the condensed leadership summary.
Sample monthly dashboard row (illustrative)
The following illustrative row represents a mid-sized L&D organisation at month nine of a captioning programme. The programme launched with 85% backlog coverage; it is now at 94.2%. The submission rate has held above 97% for the last four months. The accuracy pass rate dipped this month because two engineering onboarding videos with high domain-vocabulary density failed the DCMP spot-check on a technical term cluster; both are in remediation. The exception queue is clean.
| Metric | Month 9 | Month 8 | Target | Status |
|---|---|---|---|---|
| Library coverage rate (all captions) | 94.2% | 91.8% | 100% | 🟢 On track |
| Library coverage rate (verified compliant) | 88.4% | 85.1% | 100% | 🟡 In progress |
| New-content submission rate | 97.3% | 96.8% | 95%+ | 🟢 Pass |
| Accuracy pass rate (DCMP spot-check) | 91.7% (11/12 sampled) | 100% (10/10 sampled) | 95%+ | 🟡 Below target |
| Accuracy remediation queue (open) | 2 videos | 0 videos | 0 | 🟢 In remediation |
| Accuracy remediation velocity (median days) | 9.0 days (3 closed) | 7.5 days (2 closed) | ≤14 days | 🟢 Pass |
| Backlog remediation queue (open) | 87 videos | 124 videos | Decreasing | 🟢 Decreasing |
| Backlog remediation velocity (closed this period) | 37 videos | 29 videos | ≥30/month | 🟢 Pass |
| Open exception count | 3 | 5 | Low | 🟢 Pass |
| Exceptions exceeding approved window | 0 | 0 | 0 | 🟢 Pass |
| Complaints / regulatory notices received | 0 | 0 | 0 | 🟢 Pass |
| Measurement protocol executed | Yes | Yes | Yes | 🟢 Pass |
The one notable item in month nine is the accuracy pass rate dip to 91.7%. In isolation, this looks like a quality problem. In context — two engineering onboarding videos with unusual domain vocabulary, both already in remediation, both on track to close within nine days — it is a quality system functioning as designed: it found failures, flagged them, and they are being resolved within the target window. This is the value of trend data: the month-nine pass rate dip reads differently against a backdrop of month-eight 100% and month-seven 95.8% than it would in isolation.
Presenting to leadership: the one-slide model and the 200-word email
Leadership reporting on caption compliance has a different audience and a different purpose than the coordinator's monthly dashboard. The CLO or VP of Learning who receives the monthly report is not trying to understand how the DCMP spot-check works. They are trying to answer three questions: Is the programme working? Is it getting better or worse? Is there anything they need to act on? The reporting format should answer all three questions in five minutes without requiring the recipient to read a methodology document.
The one-slide model
The leadership-facing report is a single slide (or single-screen document) with four metrics, each with a RAG status and a one-sentence trend note. The four metrics are selected from the twelve-row dashboard to represent the four dimensions leadership actually cares about: outcome (coverage rate), process health (submission rate), quality (accuracy pass rate), and trajectory (backlog velocity). The other eight metrics in the coordinator's dashboard are management tools, not leadership visibility tools — they appear in the coordinator's monthly notes and in the compliance log, but not in the leadership slide.
| Leadership metric | This month | Prior month | Status | Note |
|---|---|---|---|---|
| Programme coverage | 94.2% of in-scope library | 91.8% | 🟢 On track | +2.4 points; at current pace, 100% by month 12 |
| New-content hygiene | 97.3% of new videos captioned before publication | 96.8% | 🟢 On track | Fourth consecutive month above 95% target |
| Accuracy quality | 91.7% of spot-checked videos passed | 100% | 🟡 Watch | Two engineering videos failed; both in remediation, on track to close by month end |
| Backlog pace | 37 historical videos captioned this month | 29 | 🟢 Improving | Accelerating; 87 videos remain in backlog |
Below the four-metric table, add a single line: "Next action items: [one or two items maximum, each one sentence]." If there are no action items required from leadership, write "No action required from leadership this period." This line is important: it tells the CLO whether they need to do anything or whether the report is information-only. Most months it will be information-only. On months where there is a Red metric or a complaint, it will have a specific ask (budget for remediation vendor capacity, escalation authority to a department that is not cooperating with producer training, awareness that a complaint has been received and is being responded to per the response protocol).
The 200-word email model
The leadership report can also be delivered as a short email rather than a slide — more practical for programmes where the CLO does not attend a standing monthly review. The email template:
Subject: Caption compliance update — [Month Year]
Programme coverage is at [X]%, up from [Y]% last month. At current pace, full backlog coverage is projected by [month/quarter]. New-content submission rate is [X]% — [above / below] the 95% target. Accuracy spot-check returned [X]% pass rate on [N] videos sampled this month [, with [N] failures in remediation and on track to close within 14 days]. Backlog remediation is [accelerating / holding / slowing]: [N] historical videos captioned this month, [N] remaining. No [/ One / Two] complaints or regulatory notices received this period. No action required from you this month [/ Please see attached for one item requiring your awareness].
Full dashboard available in the [compliance log location] if you need the detail behind any of these numbers.
The 200-word format works because it respects the recipient's time. A CLO who receives this email each month, sees Green status, and takes no action is a CLO who is aware that the programme is functioning without being asked to manage it. A CLO who receives this email and sees a Amber or Red item has a one-sentence explanation and a one-sentence status. They do not need to read three paragraphs to understand what is happening.
Why you do not lead with the compliance law
The temptation in caption compliance reporting is to frame every report in terms of the legal obligation: "We are required to comply with WCAG 2.1 AA under ADA Title II / Section 508 / EAA, and our current coverage rate indicates..." This framing is counterproductive with a leadership audience. It puts the CLO in the position of the accountant reviewing a liability, not the executive receiving a programme update. The ROI framing post covers the budget conversation — that conversation happens at programme inception. Once the programme is running, the monthly report should lead with trajectory, not obligation. "We are at 94.2% coverage and gaining 2-3 points per month" is more compelling than "we are not yet fully compliant with the ADA deadline." The first framing shows a programme that is working and getting better. The second framing shows a liability that is not yet closed. Both describe the same situation, but only the first one maintains executive sponsorship over the long arc of a multi-year compliance programme.
The metrics that survive a WCAG investigation
When a complaint is filed with the DOJ, OCR (Office for Civil Rights), or a state-level accessibility enforcement body — or when a plaintiff's attorney sends a pre-litigation letter — the organisation needs to produce documentation of its captioning compliance programme. The documentation package that answers investigator questions is specific and predictable. Investigators across enforcement contexts ask the same four questions, because those four questions correspond to the four elements of a defensible compliance programme.
The four investigator questions
Question 1: What percentage of your in-scope video content has compliant captions? This is answered by the library coverage rate (verified compliant) from the monthly dashboard. The answer the organisation needs to be able to give is not just the current number — it is the number for each of the last twelve months, showing the programme trajectory. An organisation at 94% coverage with twelve months of upward trend data is in a different legal position than an organisation at 94% coverage that has been at 94% for eighteen months. The trend data demonstrates good-faith remediation effort; the static number does not.
Question 2: What is your process for captioning new content before publication? This is answered by the governance policy's new-content gate and the submission log. The answer needs to describe: who is responsible for submitting content, what the submission process is, what the turnaround SLA is, and what happens when a video is published without going through the process. The submission log provides the documentation that the process is being followed month-to-month — the process description without the log is a policy claim, not a compliance demonstration.
Question 3: What is your accuracy standard, how do you measure it, and how do you remediate failures? This is the most technically detailed of the four questions, and it is where programmes without a documented QA methodology have the most exposure. The answer needs to describe: the accuracy threshold (99% word accuracy at the WCAG 2.1 AA level), the measurement methodology (DCMP Captioning Key spot-check on a defined random sample), the measurement frequency (monthly), the remediation trigger (any video failing the spot-check) and the remediation timeline (≤14 calendar days). The accuracy log is the documentation that this methodology is being applied — monthly spot-check records with video IDs, sample parameters, error counts, pass/fail results, and remediation status.
Question 4: Show me the last 12 months of documentation demonstrating that this process is being followed. This is answered by the twelve-month compliance log — the running record of monthly dashboard rows, the submission log, the accuracy log, the exception log, and the remediation queue. An organisation that has retained all four of these logs for twelve months, plus the governance policy that describes the standards being measured against, has the complete documentation package. An organisation that has the policy but not the logs has a paper programme. An organisation that has some months of data and not others has gaps that investigators will probe.
What investigators look for in the documentation
Beyond the four questions, investigators reviewing documentation look for three specific patterns that indicate whether the programme is real or paper:
Pattern 1: Consistency of measurement.** A compliance log that has entries for ten of the twelve months and is missing February and July is a programme that stopped measuring when something else was more urgent. Investigators treat gaps in measurement as potential gaps in compliance — the organisation cannot demonstrate what it did not measure. A complete twelve-month record demonstrates that the measurement process is institutionalised, not ad hoc.
Pattern 2: Evidence of self-correction. A compliance log that shows all Green metrics every month is suspicious to an experienced investigator — either the programme is performing exceptionally well, or the metrics are not measuring what they claim to measure, or some findings are not being recorded. A log that shows occasional Amber items (accuracy pass rate dipped one month, remediation queue was elevated for six weeks) with documented corrective action is more credible because it demonstrates that the programme is finding and closing its own problems. This is exactly what a quality loop is supposed to produce; its absence in the documentation is a warning sign.
Pattern 3: Traceability from policy to practice. The documentation package should be internally consistent: the governance policy should describe the accuracy standard; the accuracy log should measure against that standard; the remediation queue should reference the videos that failed; the dashboard should aggregate the accuracy log results into the pass-rate metric. An investigator who can trace a single finding — an accuracy failure in the log — through to a remediation record and a dashboard row and a governance policy provision has seen a programme with operational integrity. An investigator who cannot make that trace — because the logs are in different formats, maintained by different people, with different video IDs — is looking at an administrative programme that exists on paper.
The one document that changes the conversation
In the event of an OCR investigation or a plaintiff's attorney letter, the single most useful document the organisation can produce is the twelve-month compliance log — the running record of monthly dashboard rows. This document demonstrates continuous measurement, documents trajectory, shows self-correction, and provides the numeric evidence of good-faith remediation effort. Organisations that can produce this document in response to the initial inquiry change the trajectory of the investigation: they are not fighting over whether the programme exists (it visibly does) but over whether the current coverage level, trajectory, and remediation timeline are sufficient. That is a conversation about programme pace, not programme existence — and the outcome of that conversation is typically a negotiated remediation timeline rather than a consent decree or injunction. Organisations that cannot produce it are defending a programme they cannot demonstrate.
LMS-specific reporting notes
Pulling the caption inventory data that feeds the monthly dashboard requires navigating the LMS admin interface, and the path to caption status data varies significantly by platform. The notes below describe the fastest route to the data you need for the monthly measurement protocol for the eight LMS platforms where L&D accessibility coordinators encounter the most reporting friction.
Kaltura
Kaltura's Caption Analytics report in the admin console gives the most complete view. Navigate to Analytics → Custom Reports → Caption Analytics; filter by date range and content type (exclude live-stream content unless it is in scope). The report includes caption status per video entry. For MediaSpace-hosted content, the consumer-facing caption report may need to be pulled separately. The most common error: pulling the caption-file-present field without filtering for caption-language, resulting in a count that includes auto-generated (unchecked) captions alongside human-reviewed ones. Filter by caption-source type to separate auto-generated from vendor-delivered files. Cross-reference caption-file creation date with the accuracy log to identify which vendor-delivered files have been spot-checked. See the Kaltura captions guide for integration-level detail.
Docebo
Docebo's reporting console exposes video content at the learning-object level. Go to Reports → Learning → Learning Objects; filter by type "Video" and export to CSV. The CSV includes a "closed captioning" field that shows whether a caption file is attached at the platform level. Note that Docebo's caption field reflects only captions uploaded directly to Docebo — video content delivered via external embed (Wistia, Vimeo, YouTube) will show no captioning status in Docebo even if the host platform has captions configured. Pull the external-embed inventory separately and cross-reference with the host platform's caption data.
TalentLMS
TalentLMS does not have a native caption inventory report as of mid-2026. The fastest workaround: pull the Course Content report (Admin → Reports → Content) and export to CSV; filter for video type. Caption status is not included in the export. Cross-reference video IDs against the submission log and the accuracy log manually to determine captioned and verified-compliant status. This is the platform where the monthly measurement protocol takes the most time — allocate an extra thirty minutes compared to platforms with native caption reporting. See the TalentLMS captions guide for upload and workflow notes.
Absorb LMS
Absorb's Content Library view (Admin → Content Library → Video filter) shows caption status as a column in the content grid. Export the filtered view to CSV. The caption-status column reflects whether a caption track is attached in Absorb — cross-reference with the accuracy log to identify verified-compliant videos. Absorb's caption field distinguishes between "auto-generated" and "uploaded" tracks in the content detail view but does not surface this distinction in the bulk CSV export, so the manual cross-reference against the accuracy log is necessary to separate vendor-delivered from auto-generated captions.
Cornerstone OnDemand
Cornerstone's Object Learning and Study Activity (OLSA) reports provide the most complete content inventory. Navigate to Reports → OLSA → Content; filter by learning-object type "Online Course" and look for video objects. Caption status is in the Content Management section of the admin panel, not in the OLSA reports — pull the two datasets separately and join on the learning object ID. The join step is the most common source of error in Cornerstone reporting: the OLSA report uses a different ID scheme than the Content Management export. Map both to the video title as a secondary join key.
Workday Learning
Workday Learning's Content Inventory report (accessible via Reports Worklet → All Reports → Learning Content) includes a "Closed Captioning" field for Workday-native video objects. External video content — content hosted in Vimeo, YouTube, or a CDN and delivered via Workday link — does not appear in the Content Inventory report and must be inventoried separately from the source host. For organisations using Workday as a catalogue layer over externally hosted video, the caption reporting process involves two parallel inventory pulls (Workday native + external host) and a reconciliation step that adds thirty to forty-five minutes to the monthly measurement.
Panopto
Panopto's Admin panel → Usage report includes caption status per session. Export to CSV and filter for caption-enabled sessions. Panopto distinguishes between automatic speech recognition (ASR) captions and imported (professional) captions in the session detail view — the bulk export does not make this distinction cleanly. Run a secondary filter: sessions where the only caption track is the ASR track should be counted as captioned but not verified-compliant unless the ASR output has been reviewed and approved through the Panopto editor. Sessions with an imported track from a verified vendor are candidates for compliant status pending accuracy log cross-reference. See the Panopto captions guide for upload architecture.
Canvas LMS
Canvas's video caption inventory is distributed across its content types: Media Gallery (Canvas Media hosted video), Studio (if provisioned), and instructor-uploaded video in course pages. The Canvas Admin → Media Objects report covers Media Gallery content; Canvas Studio has a separate admin dashboard. Course-page embedded video (YouTube, Vimeo, external iframes) is not captured in either report and must be inventoried via a course-by-course content audit. For universities using Canvas, this distributed structure is the primary reporting challenge: a clean caption inventory requires reconciling three or four separate data sources. Organisations using Canvas who need a clean monthly caption inventory typically assign the LMS admin report pull to the IT or LMS admin team rather than the accessibility coordinator, because the multi-source reconciliation requires admin-level access to multiple subsystems.
Eight failure modes in caption compliance reporting
Caption compliance reporting fails in consistent patterns. The eight failure modes below are drawn from the gap between what programmes document and what investigators and auditors actually need to see. Each failure mode has a specific correction that does not require rebuilding the programme — it requires updating the measurement protocol or the documentation practice.
Failure mode 1: Reporting only library coverage rate
The programme measures and reports only the percentage of videos with captions, treating this as a sufficient characterisation of compliance status. The consequence is leading-indicator blindness: the coverage rate moves slowly and lags the programme's actual health by three to six months. A programme at 90% coverage with a 70% submission rate will not show the submission rate problem in the coverage rate for many months — by the time the coverage rate starts declining, the submission problem has been running for half a year. The correction: add the new-content submission rate as a co-equal metric alongside coverage rate in every report. If the submission rate is below 95%, it is the more urgent problem, regardless of what the coverage rate shows.
Failure mode 2: Measuring accuracy only on complaint
The programme has no proactive accuracy measurement — it measures accuracy only when a viewer reports a captioning problem. The consequence is that the programme has no baseline, no trend data, and no documentation of independent verification of vendor accuracy. When a complaint arrives, the organisation cannot demonstrate that it was monitoring accuracy before the complaint. The correction: implement the DCMP spot-check protocol on a monthly random 5% sample and retain the results in the accuracy log. The first twelve months of log entries are the documentation that the programme was measuring quality before any complaint arrived.
Failure mode 3: Reporting against an undocumented target
The monthly report shows metrics but does not define the target each metric is being measured against, or the target was never formally approved by leadership. When a metric falls below a threshold, there is no agreed-upon basis for determining whether it is a problem. The correction: include the target column in the dashboard (as in the twelve-row template above) and get leadership approval for the targets as part of the programme's initial setup. The target approval is typically a five-minute addition to the programme kickoff meeting — "here are the five KPIs we will track and the thresholds we will treat as pass/fail; are these the right thresholds?" — that prevents a significant ambiguity when the first Amber or Red metric appears.
Failure mode 4: Undocumented measurement methodology
The dashboard shows an accuracy pass rate but does not document how the accuracy was measured. Investigators ask for the methodology because the methodology is what makes the measurement admissible. An accuracy pass rate without a documented methodology is not a measurement — it is a claim. The correction: document the measurement methodology in the QA policy (the DCMP spot-check parameters, the sample size, the error taxonomy, the pass/fail threshold) and reference that policy document from the accuracy log. The accuracy log row should include a "methodology reference" column that links to the policy version in effect at the time of the measurement.
Failure mode 5: Mixing in-scope and out-of-scope content in the coverage calculation
The LMS inventory export includes all video content — in-scope training content and out-of-scope content (informal recordings, archived webinars, pre-programme content). The coverage rate is calculated over the combined inventory, producing a rate that is either inflated (if out-of-scope content happened to be captioned) or deflated (if large archives of clearly out-of-scope content are included). The correction: define the in-scope content population precisely in the governance policy (which content types, which publication-date cutoff, which delivery channels) and filter the LMS inventory export to that population before calculating coverage rate. The scope definition is what gives the coverage rate a defensible denominator.
Failure mode 6: Treating the absence of exception-log entries as "no exceptions"
The exception log has not been reviewed or updated in three months because the coordinator has been managing the backlog remediation. When the monthly report asks for exception programme health, the coordinator reports "no exceptions" because the log is empty — not because there are genuinely no exceptions, but because the log has not been maintained. Meanwhile, four producers have been getting informal verbal approvals to publish videos without captions because their content type is "probably exempt." The correction: make exception-log review a scheduled step in the monthly measurement protocol (step six in the protocol above), not a periodic activity that happens when someone remembers. An empty exception log that has not been reviewed in three months is not evidence of a clean programme — it is evidence of a lapsed process.
Failure mode 7: Presenting single-month snapshots without trend data
The monthly leadership report shows the current month's numbers but no prior-month comparison and no trajectory indicator. Leadership has no basis for evaluating whether a 94% coverage rate is good (if it was 88% three months ago, it is excellent) or concerning (if it was 97% three months ago, it represents a significant regression). The correction: include at minimum the prior-month value for each metric in the leadership report. The twelve-row dashboard template above includes prior-month and target columns for this reason. For annual reviews, include a twelve-month trend chart that shows all five KPI categories over time.
Failure mode 8: Single-point-of-failure documentation storage
The twelve-month compliance log lives in a spreadsheet on the accessibility coordinator's laptop or in a personal Google Drive folder. When the coordinator transitions out of the role, the new coordinator either cannot find the log or finds a version that is months out of date. The three-year compliance trail is effectively gone. The correction: store the compliance log in a shared, access-controlled location (a team Google Drive folder, a SharePoint site, a shared Notion database) with at least two people having edit access and one additional person having read access for business continuity. The compliance log is an institutional document, not a personal work product — it should be treated with the same access controls as any other institutional compliance record.
Frequently asked questions
How often should we report to leadership on caption compliance?
Monthly is the right cadence for most L&D organisations. Monthly reporting keeps the programme visible without overwhelming leadership with data. Annual reporting is too infrequent — it misses the trajectory signals (submission rate trends, backlog velocity) that indicate whether the programme is gaining or losing ground. Quarterly is a reasonable minimum for leadership, but the coordinator should still measure monthly even if the leadership report is quarterly, because the compliance log needs monthly entries for investigator documentation purposes. A practical structure: monthly measurement and coordinator review, quarterly leadership summary (four months of data presented together), annual programme review with twelve-month trend data.
What should we do if our submission rate drops below 95%?
A submission rate below 95% is a signal that one of three structural problems has appeared: workflow friction (producers have to leave the normal publishing path to submit for captioning), scope ambiguity (producers are self-classifying their content as out-of-scope), or accountability absence (there is no monitoring mechanism and non-submission has no visible consequence). The first step is diagnosis: pull the submission log and identify which departments or content types have the lowest submission rates. That will tell you which of the three problems is causing the shortfall. The change management post covers the specific structural counters for each of the three problems. Do not respond to a submission rate shortfall with a reminder email to all producers — that addresses the symptom, not the structural cause, and will produce a one-month improvement followed by return to baseline.
How does the DCMP spot-check work in practice — what are we actually reviewing?
The DCMP Captioning Key spot-check is a structured accuracy evaluation. Select a random 60-second segment from the video (not the first 60 seconds, which are often better-captioned than the rest). Play the audio and read the corresponding caption text simultaneously. Count all words spoken in the audio. Count all errors in the caption: substitutions (wrong word for a spoken word), insertions (extra words in the caption not in the audio), deletions (words spoken but missing from the caption), and formatting errors (speaker identification omitted, significant timing gap, punctuation missing where meaning is affected). Divide error count by word count to get the word error rate. Pass: error rate ≤1%. Fail: error rate >1%. The full QA methodology post covers the complete error taxonomy, the error-type root-cause mapping, and the systematic triage process for identified failures.
Do we need to measure accuracy if we are using a certified captioning vendor with a 99% SLA?
Yes. A vendor SLA is a contractual claim; independent verification is the measurement. The SLA tells you what the vendor is liable for if accuracy falls below 99%. It does not tell you what accuracy the vendor is actually delivering on your content. Most vendor SLAs are measured by the vendor using their internal methodology — which may use a different sample, a different error taxonomy, or a different evaluation window than the DCMP protocol. Running your own independent spot-check on a monthly 5% sample gives you three things the vendor SLA does not: documentation of independent verification (which investigators want), an early-warning system for accuracy drift (which the vendor's internal monitoring is motivated to miss), and a baseline for vendor-performance management (which is useful if you ever need to enforce the SLA). See the vendor contract checklist for SLA clause analysis and enforcement mechanisms.
What is the difference between library coverage rate and compliance rate?
Library coverage rate is a count of videos with caption files attached, expressed as a percentage of the in-scope library. Compliance rate is a more demanding measure: the percentage of in-scope videos that meet the full WCAG 2.1 AA standard — which requires not just a caption file, but 99% word accuracy, synchronisation within two seconds, and speaker identification where multiple speakers are present. The two rates differ wherever the video library contains caption files that do not meet the accuracy standard. In most L&D organisations, the primary source of non-compliant caption files is auto-generated captions (YouTube auto-captions, Zoom auto-transcription) that were attached to videos without human review or accuracy verification. A video with an auto-generated caption file counts toward library coverage rate but not toward compliance rate if the caption accuracy has not been verified. The twelve-row dashboard tracks both — library coverage rate (all captions) in row 1 and library coverage rate (verified compliant) in row 2 — because both have different uses: coverage rate for programme trajectory, compliance rate for investigator documentation.
How should we handle newly acquired content from mergers or LMS migrations?
Acquired content — videos that enter your LMS from a merger, an LMS migration, or a content vendor — should be treated as historical backlog for compliance purposes: they do not go through the new-content submission process (since they were not produced under your programme) but they do enter the in-scope library and need to be remediated. The first step is an inventory: count the acquired videos, determine what caption files exist, run a sample spot-check to assess the accuracy of the existing captions. The spot-check sample should be larger than the routine 5% monthly sample — for an acquired library, a 10-15% sample is appropriate because the accuracy baseline of the acquired captions is unknown. The LMS migration caption checklist covers the full pre-cutover audit sprint and post-migration validation protocol for migration scenarios where caption data is being transferred between platforms.
What documentation do we need to produce if we receive a complaint?
If an OCR complaint or a DOJ letter arrives, the documentation package you need to produce typically includes: the governance policy (the written compliance standard), twelve months of compliance log entries (the twelve-row dashboard rows showing measurement history), the accuracy log for the reporting period (DCMP spot-check records), the remediation queue showing how identified failures were closed, the submission log showing new-content process compliance, and any correspondence with the captioning vendor relevant to the complaint. If the complaint concerns a specific video, also produce: the caption file for that video, the accuracy log entry for that video (or a note that it was not in the spot-check sample for the relevant period), and the remediation record if it was flagged. The organisations that manage OCR investigations most effectively are those that can produce this package within 48-72 hours of the initial inquiry — which requires that the documentation be maintained continuously, stored in an accessible location, and organised in a format that does not require weeks of reconstruction. That is exactly what the monthly measurement protocol and the compliance log are designed to produce.
GlossCap makes the accuracy measurement automatic
The DCMP spot-check protocol and the monthly accuracy log are manageable with two hours per month when the process is defined. They are unmanageable without a vendor whose captions start at 99% accuracy rather than 82%. If your current captioning vendor is delivering auto-generated captions that your team is correcting by hand before the monthly spot-check, the measurement process is papering over an accuracy problem that GlossCap solves at the source: WCAG 2.1 AA compliant captions with your company glossary applied, so engineering terms, product names, and medical terminology come out right the first time. The reporting framework above gives you the dashboard. GlossCap gives you numbers worth reporting.