Fix, Rebuild, or Retire? How to Prioritize Your Accessibility Remediation Backlog
- Published on: October 23, 2026
- Updated on: October 23, 2026
- Reading Time: 8 mins
-
Views
Check What the Accessibility Audit Really Covers
Address Blocked Learning While Planning the Permanent Repair
Group Accessibility Findings into Work Packages
Decide When to Fix, Rebuild, or Retire Accessibility Issues
Fix When the Existing Design Can Support Access
Rebuild the Part That Cannot Support the Required Interaction
Retire When Withdrawal Is Viable for Existing Users
Estimate Accessibility Remediation Costs from a Representative Pilot
Sequence Remediation Around Dependencies and Release Dates
Maintain Immediate Assessment Access
Repair the Shared Control
Correct the EPUB Production Template in Parallel
Complete the Supplementary Title’s Withdrawal
Agree Who Can Approve Treatment Funding and Completion
Define Accessibility Acceptance Criteria Before Delivery Starts
Stop the Next Release from Recreating the Backlog
FAQs
An accessibility backlog can contain hundreds of findings without showing how many changes the product needs. The same inaccessible assessment control may appear in dozens of tickets. A heading problem may recur across a publishing catalog because one export template keeps reproducing it.
When I scope remediation for an EdTech product or learning content, I start by identifying what prevents learners from completing their work and where those barriers originate. That helps engineering and content teams decide what to repair, what needs redesign, and what can be withdrawn.
To build a funded plan, we also need to establish the delivery order, estimate the work, and agree how the result will be accepted.
Check What the Accessibility Audit Really Covers
Before estimating an inherited backlog, match the findings to the current product and content versions. Reproduce significant barriers, consolidate duplicate reports, and identify anything the original audit did not cover.
Confirm the accessibility requirements against which the work will be evaluated, including the relevant standard, version, and level, supported environments, and customer commitments. An old audit of a previous release cannot tell you everything about today’s experience.
For each group of findings, establish where the change belongs: in a source file, authoring template, shared component, exported package, or delivery platform. Record who controls that layer.
This is important for when a content team can edit an EPUB but cannot change the reader displaying it. It also prevents you from estimating separate repairs for defects caused by the same component.
Address Blocked Learning While Planning the Permanent Repair
Consider an illustrative publisher with three problems: learners cannot submit a required assessment using a keyboard, several EPUB titles contain recurring structural errors, and an older supplementary title is scheduled for withdrawal.
The assessment needs an immediate response. Assign someone to establish an accessible way for affected learners to complete the task while the team investigates the permanent correction. An instructional specialist should check that any temporary alternative preserves what the assessment measures.
Give that arrangement an owner and a review date. Continue investigating and estimating the durable repair in parallel.
W3C’s guidance on interim accessibility repairs prioritizes complete key tasks and known barriers. A barrier affecting a small group can still prevent a required learning activity; usage volume alone should not push it down the queue.
Once urgent access needs and binding deadlines are identified, compare the remaining work using reach, shared causes, dependencies, and available capacity.
Group Accessibility Findings into Work Packages
A ticket describes a problem. A work package should define a change that a team can estimate, deliver, and validate.
For the assessment, that package might cover the shared answer-selection control, the item types using it, and the product versions receiving the correction. For the EPUBs, it might cover a faulty export template and the affected titles that need regeneration and review.
Keep the individual findings linked to these packages so that nothing disappears during consolidation. Check variations before assuming that one solution will work everywhere.
A single course may need several treatments. Its book might need repair, an interaction might need redesign, and an obsolete supplementary activity might be withdrawn. Choose the treatment at the level where the change is needed.
Decide When to Fix, Rebuild, or Retire Accessibility Issues
Fix When the Existing Design Can Support Access
Repair is appropriate when a correction can remove the barrier while preserving a useful, maintainable asset or component.
For an EPUB, that could mean correcting heading structure and navigation, reviewing reading order, and validating the result in the intended reader. Missing original authoring files do not automatically rule out repair; investigate what can be corrected safely in the existing package.
Also establish how future updates will be handled. A repair to the exported file may need repeating if the production process continues to generate the same defect.
Rebuild the Part That Cannot Support the Required Interaction
A rebuild can be confined to an interaction or component. An inaccessible control does not, by itself, establish the need to replace an entire course or platform.
For a dragging activity, first test whether accessible alternatives can be added to the existing control. WCAG’s dragging guidance distinguishes a single-pointer alternative without dragging from keyboard access. Both need consideration.
Prototype a replacement when the current implementation cannot reliably support the required operation or when repeated patches would be difficult to maintain. Have accessibility, engineering, and instructional specialists review whether learners can select, review, change, and submit their answers without changing the intended skill or scoring.
We’ve done this before for a science games and simulations project, where we worked across over 60 activities using templated solutions, automated and manual testing, and testing with people with disabilities. This shows why a reusable solution deserves investigation. Each activity still needs checks for its particular content and behavior.
Retire When Withdrawal Is Viable for Existing Users
Low usage or obsolete source files should trigger investigation into an asset’s future.
Before recommending retirement, check whether customers still assign the material, whether it remains necessary for a course, and what commitments continue. Establish whether learners need an accessible replacement or a supported way to finish existing work.
Product and account owners should confirm the transition, with legal input where commitments need interpretation. Include links, integrations, learner records, and support arrangements in the withdrawal plan.
A future retirement date still leaves a present access need. The treatment plan must cover that period.
Estimate Accessibility Remediation Costs from a Representative Pilot
I would test the proposed treatment on a small, deliberately varied sample before committing to a large batch.
For publishing work, include straightforward material alongside likely sources of complexity, such as equations, diagrams, and unusual layouts. For a shared component, test the different item types and product configurations that depend on it. Include important cases that may break the proposed solution.
Record the effort for diagnosis, implementation, expert review, rework, and validation. Separate work done once from work repeated for each asset.
For example, suppose a pilot indicates 12 hours to correct and validate a template, followed by 3–5 hours per comparable title for content-specific corrections and review. Ten titles would require 42–62 hours before release integration and wider regression testing. These assumed figures show the calculation; a live estimate must use measurements from your material.
Keep complex titles outside that range until they have been assessed. State assumptions and unresolved dependencies alongside the estimate.
Compare repair and replacement over the asset’s remaining useful life. Repeated repairs, repeated testing, and support for an old component all consume capacity. Replacement has its own integration and rollout costs. Where uncertain assumptions could change the recommendation, fund a bounded investigation before approving the full program.
If a platform replacement is already planned, it is reasonable to question spending twice. Check its actual delivery date against the next required use of the content. A targeted repair or validated alternative may still be needed while the new platform is being built.
Sequence Remediation Around Dependencies and Release Dates
Return to the illustrative publisher. Suppose the pilot confirms that the assessment control can be repaired and that the EPUB defects originate in an independent export template.
I would propose the following sequence:
Maintain Immediate Assessment Access
Keep the agreed temporary arrangement in place while the permanent correction is developed and validated.
Repair the Shared Control
Prioritize deployment to the active assessment, then roll the validated change into other affected products with checks for their variations.
Correct the EPUB Production Template in Parallel
If separate people can do this work, regenerate and review affected titles in batches ordered by current assignments and delivery commitments.
Complete the Supplementary Title’s Withdrawal
Remove it only after the transition checks are complete, maintaining access arrangements for anyone who still needs it.
The available people determine which work can run together. Several engineers cannot compensate for an unavailable specialist whose review every package requires.
For a third-party player or component, obtain a vendor response and expected delivery date before treating its correction as scheduled work. If that date misses the learner need, product and accessibility owners must evaluate an alternative delivery route or replacement.
Use reach and commercial importance to compare work within those access and delivery constraints. Keep unresolved barriers visible, with a responsible owner and the next review date.
Agree Who Can Approve Treatment Funding and Completion
Accessibility, engineering, and content operations contribute different evidence. Make their responsibilities explicit before delivery starts.
The accessibility lead should define the barriers and evaluation requirements. Engineering and content owners should establish what can be changed, its dependencies, and the implementation effort. Product owners should resolve scope, roadmap, and retirement decisions with the relevant commercial teams.
Name the budget approver and the person responsible for accepting the completed work. Agree who resolves a disagreement, such as a recommended component replacement that conflicts with the product roadmap.
For each package, retain the chosen treatment and rationale, affected assets, implementation owner, dependencies, effort range, target release, and acceptance requirements. Keep temporary access arrangements linked to the permanent work.
A useful funding request identifies which packages the proposed spend will deliver and what remains uncertain. If the evidence supports only a pilot, ask for that pilot with a clear decision it must resolve.
Define Accessibility Acceptance Criteria Before Delivery Starts
Agree the evidence needed to close each package while estimating it.
For the shared assessment control, that should include the applicable accessibility requirements and the complete learner task. Check answer selection, review, submission, saved responses, scoring, and feedback in the agreed environments.
Record the product version, browser or reader, assistive technologies, and test results. Include instructional review where a change could affect the skill being assessed. Involve people with disabilities early when evaluating a new interaction, while there is still time to change the design.
Define which checks apply to every asset and where a justified sample can be used. A validated template does not settle whether every title has correct image descriptions or reading order.
Tools can support these checks, but W3C’s evaluation guidance makes clear that knowledgeable human evaluation remains necessary. Passing the agreed checks for one package also does not establish conformance across an entire product.
Retest the deployed result and relevant neighboring functions before closing the work. Retain any unresolved findings with their owners and next actions.
Stop the Next Release from Recreating the Backlog
A remediation package should include the upstream correction wherever the team controls it.
That may mean changing an authoring template, updating a conversion script, adding checks to a component’s release process, or improving instructions for content contributors. Assign responsibility for keeping those changes in place as content and products evolve.
This is also where I would assess automation. On suitable EPUB and HTML work, MagicA11y supports structural checks and repeatable corrections, with issues requiring judgment referred for human review. It is an internal toolkit used by Magic’s delivery teams.
The pilot should establish whether that approach reduces the total delivery effort for the material in scope, including review and rework. Complex interactions, instructional decisions, and content-specific descriptions still require the right expertise.
Keep work in-house where your team has the skills and capacity. Use specialist support for decisions or interactions the team cannot confidently resolve, and additional delivery capacity once the method is validated.
At Magic EdTech, we can start with your current audit, a few representative assets, and the learning or release commitments at risk. Our role is to help establish a feasible treatment, the effort needed to validate it, and the first packages worth funding before scaling the work.
FAQs
Break out the barriers that interfere with learning and deadlines from the rest of the backlog. Next, group tickets by their root causes so that a single component-level problem is not competing against itself in a multitude of individual tickets. Finally, evaluate the remaining work packages for impact, interdependencies, and resources needed to verify them.
Conduct a representative pilot project to estimate the effort. Identify the one-off effort for change from versus what will have to be done on an asset-by-asset basis. Budget for the effort identified based on your delivery rates. Consider separately any effort for integration and regression testing. Exclude untested and complex assets from the estimate.
Judge the automation on the entire process of distribution rather than the quantity of corrections it makes. When comparing the pilot effort, account for human review, network, and verification. Even if the automated process of correcting is faster, it should not be considered better if it results in further correction efforts. Keep the complex interactions, decisions made, and content-based descriptions with the experts.
Compare the schedule of delivery of the replacement with the next time the content will be needed. Plan a targeted fix or validated alternative depending on which comes up sooner.
Don’t treat a dependent vendor fix that does not address our issue as a confirmed fix. The product and accessibility owners need to consider the dates from the vendor, compare them with the next required usage date for the learner, and consider an alternative delivery mechanism. Keep the alternative delivery mechanism, owner, and date of next review visible until we get a permanent solution confirmed.
It is important to distinguish between two issues – uncertainty regarding the approach to take and simple capacity of delivery. When the issue is which approach will work, then Magic EdTech can help you with reviewing the audit and some of the sample assets in order to come up with a practical approach and the validation effort. Then you can increase your delivery capacity.
Get In Touch
Reach out to our team with your question and our representatives will get back to you within 24 working hours.
