An accessible Excel workbook lets a reader find the correct worksheet, understand tables and charts, follow instructions, and distinguish actual data from decoration. Microsoft Excel Accessibility Checker is an important starting point for multilingual spreadsheets, but it cannot decide whether a translated currency label is accurate, a formula still represents the same business rule, or a color-coded warning is understandable without color. A workbook may appear beautifully localized on screen yet remain disorienting to someone using a screen reader, keyboard, magnification, or braille display.
This practical Excel accessibility checker guide covers checking workbook structure, using meaningful worksheet and table names, identifying header rows, writing chart alternative text, reviewing contrast, removing unnecessary merged cells, and checking translated Excel spreadsheets after publication. The intended readers include educators, translators, finance teams, researchers, health administrators and international organizations sharing files across languages and regions. It separates what Excel can flag automatically from what a person must evaluate in context and explains how to preserve formulas, data types and numerical meaning.
For a useful first test, copy a real workbook into a review folder, open Review > Check Accessibility in a supported Microsoft Excel edition, and inspect both the checker findings and the Accessibility ribbon. Rename vague sheets, identify structured tables, examine visuals, then navigate the result with a keyboard and appropriate assistive technology. Repeat with a locale that changes text length, number punctuation or right-to-left presentation. The final deliverable should still permit a reader to locate the answer, understand the unit and act on the correct value without guessing what a visual cue meant.
- Use the built-in checker, but verify actual workbook tasks rather than treating a clean result as certification
- Keep meaningful sheet names, real header rows, clear table names and navigable data regions
- Translate labels, instructions and alt text without rewriting formulas or silently converting data types
- Check charts, conditional formatting, colors, links, hidden material and exported copies
- Document evidence for each language, device and delivered file
For translation principles beyond spreadsheets, read Master Art of Translation. Related implementation guides are collected in the Translate Like a Pro Series Index. This page concerns readable, operable Excel workbooks; it is not an introduction to translating prose or a guide to modifying SpreadsheetML internals.
What the Excel Accessibility Checker is actually checking
Excel offers built-in accessibility checks and editing tools for potential problems such as missing alternative text, difficult table structure and certain visual presentation issues. In supported versions, Review > Check Accessibility opens checking and related Accessibility ribbon tools. Microsoft also recommends descriptive sheet names, tables with headers and intelligible chart information. Features and exact pane names can vary by platform, license and software version; verify what your installation exposes rather than following an out-of-date screenshot mechanically.
A warning is a request to inspect evidence, not a judgement about every possible user. A green result does not certify accurate translations, screen-reader behavior, keyboard paths or regulatory conformance. Likewise, an automated warning sometimes needs contextual judgement: a decorative image may not require the same description as a chart carrying the only evidence for a policy decision. Treat the checker as one part of a review sequence: identify, repair, test, compare against the source, and record a meaningful outcome.
The most important distinction is between structure, presentation and meaning. Structure includes worksheets, named tables, headers, navigation and the way data regions are organized. Presentation includes contrast, font legibility, size, charts and layout. Meaning includes units, formulas, category definitions, dates, translation, exceptions and the actions a reader must take. Good localization preserves all three, and neither a format copy nor a checker result can replace their combined review.
A safe workbook review from intake to release
Begin by preserving the original workbook and its version. Record the sheet inventory, named tables, pivot tables, charts, external connections, macros if present, protected ranges and important formulas. Establish which cells are editable labels, which are calculated outputs, which contain identifiers and which are inputs. A translation brief should declare what must not change: named ranges, worksheet references, external data, validation rules and numerical data. If you are uncertain whether a visible string is a formula or a label, inspect the formula bar and dependency chain before changing it.
Choose representative tasks instead of attempting to inspect random cells. Ask a reader to find last month’s amount, follow a chart trend, filter a regional table, identify an exception, and locate instructions for the next action. Test those tasks in the source and target languages. This catches problems that a scan of warnings alone misses, especially when a workbook contains twenty near-identical worksheets or a dashboard dominated by colored shapes.
Run the checker after a structural baseline and again after translation. Review all reported issues; use the tool’s suggested repair only when it fits the data’s actual purpose. Perform a manual keyboard route with selection, sheet switching, filtering and reading. For critical files, ask a reader familiar with the target language and assistive setup to attempt the real decision without help from the workbook author.
Verify that each localized copy has the expected file name and revision date. Open the delivered copy in the environment its audience actually uses. An XLSX downloaded from a repository, an emailed attachment, a browser-based spreadsheet preview and a PDF export may behave differently. Retest the delivery format; a successful check in the editing environment is not evidence about a converted document or inaccessible downstream portal.
Thirty workbook situations worth testing
Sheet1 versus a real destination
A budget workbook arrives with Sheet1, Sheet2 and Final2 tabs. An analyst using a screen reader hears generic names and cannot tell which is approved. Rename sheets as Instructions, Monthly Spend and Approved Summary, keeping references intact and checking anything that depends on sheet names.
Switch among tabs using keyboard navigation, hear each unique label, and confirm existing formulas calculate exactly as before. Compare the Arabic and German versions for equivalent meaning without translating fixed machine identifiers. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
A table with a decorative first row
The first row contains a colored slogan, while the actual headers begin on row four. A reader who assumes row one names each column will misinterpret the data. Make the data range into a genuine structured table with an explicit header row; place the title separately.
Use navigation to confirm headers precede values and that sorting retains the correct column relationships. Translate the heading vocabulary, but leave identifiers and cell types unchanged. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Merging cells across the reporting grid
A weekly attendance file uses merged blocks for visual grouping, leaving holes in the navigational grid. Remove unnecessary merges, use consistent unmerged cells and express grouping through explicit labels or well-designed table structures.
Traverse row by row and confirm no student or date is skipped. For Japanese and French copies, check whether expanded labels wrap without recreating merged-cell barriers. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
A dashboard chart labelled only Chart 1
A bar chart compares monthly vaccination coverage, but its default title tells no one what is being measured. Give it a descriptive title and concise alternative text explaining what the chart means, not merely that a chart exists.
Ask a reviewer to identify the leading category, time period and comparison without seeing the bars. Recheck localized axis labels and units so a translation does not imply percentages are counts. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Conditional formatting as the only warning
Red fills indicate overdue invoices and green fills mean paid, but the status cell is empty. Add an explicit Status column with values such as Overdue, Pending and Paid, and use color only as a reinforcement.
Filter by text status and ask a color-blind or screen-reader user to identify action items. Translate state names consistently and verify every rule still refers to the intended status. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
A workbook built as one immense sheet
A country-wide survey places instructions, input form, lookup tables and conclusions many screens apart on one worksheet. Break information into sensibly named sections or worksheets, with a clear contents and return path.
Ask a new reader to find the required field without a mouse. Confirm separate localized versions retain the same logical route even when text expands. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Hidden sheets containing active assumptions
Forecast results depend on hidden assumptions and reference values. Hiding cells may be operationally reasonable, but readers can mistake a visible forecast for raw fact. Add a visible explanation of inputs, scope and limitations.
Check that important assumptions and revision dates are discoverable through accessible summary text, while preserving legitimately restricted material and calculation integrity. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
An image of a timetable
A school publishes a photograph of a bus schedule inside Excel. The times cannot be searched, filtered or navigated as cells. Recreate the timetable as structured text and data, retaining the visual image only when necessary.
A reader must find the correct route and departure independently. Check 12-hour and 24-hour display conventions without shifting the underlying time values. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Dates that become different days
A cell displayed 03/04/2027 might mean 3 April or March 4. Translate the visible format only after identifying the underlying date serial and audience convention; use an unambiguous written date in high-stakes instructions.
Inspect raw values and formulas before and after localization, then ask a reader to report the same calendar day. Do not let a comma or slash substitution silently create a different date. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Decimal and thousands separators
A data table shows 1,250.75 in one locale and a convention like 1 250,75 in another. Changing punctuation as text can turn a number into a string and break SUM or comparisons. Use appropriate number formats instead.
Check calculations and cell data types before release. Distinguish grouping conventions from decimal conventions in the target region and explain any unavoidable ambiguity. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Currencies without a clear basis
A spreadsheet displays $400 but includes readers in Singapore, Australia and the United States. Add ISO currency codes where relevant and disclose whether any conversion occurred; the same symbol is not a currency policy.
Review a sample record with a local finance expert. Confirm that labels say SGD or USD as appropriate and that an unchanged value is not presented as a converted amount. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Right-to-left labels beside numbers
A bilingual Arabic-English workbook contains right-to-left headings adjacent to Western digits, formulas and parentheses. Set suitable text alignment and inspect reading order and visual grouping rather than reversing numbers manually.
Navigate the same cells with assistive technology and compare values character by character. Verify that minus signs, currency symbols and reference codes retain their intended order. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Formula results copied as literal text
An editor pastes translated explanatory strings over computed cells because a dashboard appears to contain words. This destroys links to the underlying data. Separate human-readable labels from formula outputs and confirm the calculation layer remains protected.
Change a sample input to see whether dependent cells still update. Log all cells intentionally changed in translation and investigate any unexpected formula edits. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
An unclear hyperlink labelled Here
The Instructions sheet says ‘Click here’ three times without identifying each destination. Replace ambiguous anchor text with purpose-specific phrases such as View data definitions or Open the submission portal.
Move through links with a keyboard or screen reader and identify every destination from the label alone. Verify regional URLs are actually valid before changing them. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Alt text that repeats the visible filename
A campus map embedded as a picture has alt text ‘map_final_v8.png’. Describe the route or essential geographic relationship, or provide equivalent structured text nearby when too much detail is needed.
A reader should be able to determine the building and accessibility entrance without seeing the image. Translate names carefully and preserve official place names when they are not translated. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Tables used as layout devices
A project overview uses a giant cell grid to position logos and decorative text, making navigation tedious. Separate presentation from the data grid; keep semantic data tables for values and simple text for instructions.
Count key presses to reach the action required and compare that route across localized workbooks. Do not expand the grid merely to fit longer translated decorative wording. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Empty rows hiding the end of data
An export includes thousands of blank formatted rows, so the apparent used range is much larger than the actual survey. Remove unused formatting where safe and explain the intended data boundaries.
Use keyboard navigation to identify the real end of the list and confirm filters cover all actual rows. Reopen the saved file to ensure no formulas or validation ranges were cut inadvertently. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
No declared units in a science sheet
A column named Temperature mixes values with no indication of Celsius or Fahrenheit. Add units to column labels and document conversions separately; do not invent a conversion simply because a new language is used.
Recalculate a known sample and have a subject specialist verify the unit. Translate the column description but preserve numeric values unless conversion is explicitly in scope. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
A pivot table loses its explanatory labels
A pivot dashboard uses default field names such as Sum of Amount and Row Labels. Replace vague visible descriptions with reader-focused wording while respecting pivot field integrity and refresh behavior.
Refresh the pivot from a clean copy and confirm localized descriptions remain interpretable. Check expanded or collapsed categories with keyboard controls. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Slicers without enough context
An interactive sales dashboard presents slicers labelled Region and Product but never states whether selections hide other data. Add a short explanation of active filtering and an obvious reset path.
Ask a reader to identify filtered totals versus whole-population totals. Test names and focus behavior under localization and confirm that missing categories are not described as zero sales. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Data validation error in the wrong language
A translated application uses target-language labels but retains an English validation message explaining the permitted code. Update the human message, preserve the actual validation constraint and test invalid input.
Try a boundary value and a deliberately wrong value; verify that the error explains a repairable action in the target language, including the accepted date or number format. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
A total row mistaken for a record
A finance table places Total under the first ordinary data column with no distinct structure. Explicitly mark summaries and ensure the table design lets the user recognize totals rather than count them as a customer.
Sort and filter the table with the summary visible. Verify sums stay correct and screen-reader output makes it clear which values are aggregated. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Charts with truncated translated legends
A long translated series label wraps or is cut off in a chart legend. Redesign the chart dimensions or use a shorter approved label, with a full accessible description nearby.
Open at standard and magnified zoom, check all series can be distinguished by more than color, and validate abbreviations with a target-language reviewer. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Icons carrying compliance status
A regulatory checklist uses ✓, ! and ✕ as the only status descriptors. Add written equivalents in adjacent cells, with a legend explaining whether the symbol denotes complete, review or rejected.
Check audible output of each row and confirm that punctuation and glyph fallback do not cause ambiguous pronunciation in other scripts. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
A protected template with unreachable instructions
The worksheet protects formula cells, but the input guidance lives in a floating shape that assistive navigation may miss. Put instructions in a predictable text location, label editable cells and test protection behavior.
Complete the template using keyboard-only interaction. Confirm that protected cells cannot be edited inadvertently and that every required field remains reachable. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
A very wide comparison matrix
A price comparison uses forty vendors as columns and forces readers to remember headers over a long horizontal route. Consider a normalized data table and filters, or provide a concise comparison summary.
Give a reader a concrete question such as which vendor meets criterion B under the budget; check whether they can answer it without visual scanning. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
A workbook shared as PDF
A translator checks XLSX successfully but sends a PDF made through a print process. Navigation and table semantics may differ after export. Validate the actual PDF independently for reading order, language and text access.
Ask a reader to locate a row and its header in the delivered PDF. Keep the XLSX accessible as an alternative when the PDF cannot support the intended interactive task. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
External links to vanished regional pages
A workbook includes translated help links, but the target pages redirect to an unrelated language or no longer exist. Confirm destinations for each locale instead of applying a generic domain replacement.
Follow links without author privileges in an ordinary browser and make sure the promised document actually appears in the expected language. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Version labels missing after a revision
Two regional offices have copies with identical filenames but different rules. Add an accessible revision line with version, effective date, scope and owner; preserve it in the file and release notes.
A reviewer should determine which file supersedes which without inspecting metadata tools. Use the same date interpretation in every locale. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
A high-stakes workbook shared without reader testing
An emergency allocation sheet has clean automated checks but users cannot determine which category gets priority because the translated legend changed the meaning. Conduct task-based review with a target-language reader and the actual assistive setup.
Record the observed interpretation and correct the source of ambiguity. Release only when the reader can repeat the intended decision rule and trace the values used. The acceptance record should identify the workbook version, the worksheet or cell range reviewed, the audience task, and any limitation that still affects interpretation. A file is not meaningfully accessible merely because its screen appearance is consistent; the user must recover the same relationships, permissions, and necessary instructions. If the case fails in the localized copy but passes in the source, treat it as a release-blocking localization regression rather than dismissing it as a device preference.
Ten practical habits for translators and spreadsheet owners
- Maintain an untouched source and log intentional edits to text, formats and data validation messages.
- Classify every translatable item as a label, instruction, metadata, visual description, sample value or machine-controlled identifier.
- Require subject-matter approval before changing financial units, dates, measurement systems or business terms.
- Keep formulas, named ranges, structured references, pivot sources and external connections intact unless a separately reviewed engineering change is authorized.
- Use descriptive table and worksheet names that identify the function, not the production stage.
- Explain color states in words and test meaningful icons and chart descriptions.
- Use the checker before translation, after translation and once more on the actual release copy.
- Include a keyboard and screen-reader task in representative languages; do not rely on screenshots alone.
- Verify links, validation prompts, revision labels and downloadable versions for every target audience.
- Keep a concise issue log showing the problem, user impact, repair, evidence and remaining limitation.
A useful review sheet can contain columns for workbook identifier, source locale, target locale, worksheet, range or object, issue, affected user task, proposed repair, reviewer, check method, result and release decision. Separate defects in the language itself from defects in structure or calculations. This avoids a common problem: changing one sentence improves readability but unknowingly destroys a reference that another team depends upon. It also makes future revisions less expensive because the next editor can retest the places with demonstrated risk.
Frequently asked questions
Does a clean Excel Accessibility Checker result mean the workbook is accessible?
No. It shows that the supported automated checks did not identify certain issues in that file at the time of review. It does not establish translation accuracy, complete keyboard functionality, screen-reader comprehension, accessible export, valid calculations or user success. A person must still perform realistic tasks, ideally in the target language and the delivery environment.
Should every picture have the same kind of alternative text?
No. A decorative picture may need to be ignored by assistive reading, while a chart carrying a trend or decision needs an explanation of its meaningful content. A complex chart may require both concise alt text and a nearby accessible table or longer description. The right description communicates why the object is present, not the filename or every pixel.
Can I translate worksheet names freely?
Choose reader-friendly names, but check references, macros, external links and organizational naming rules first. Some workbook systems use sheet names programmatically. A safe localization process maps and tests such dependencies rather than assuming each tab is only visible decoration. When names must remain stable as machine identifiers, add clear localized guidance elsewhere.
Does localizing a number require changing the value?
Usually not. A numeric value can be displayed according to local separator or currency-format conventions while retaining its underlying type. Actual currency or unit conversions are separate operations requiring authorized formulas, exchange-rate assumptions, rounding policy and review. Never reinterpret a figure simply because the number looks unfamiliar in its target locale.
What is the fastest useful manual test?
Ask an unfamiliar reader to open the correct sheet, locate an identified row, report a number with its unit, explain a chart finding and identify the next action. Observe whether the task works without relying on visual position. That short route often reveals missing headings, vague labels, misleading translations and inaccessible graphics that automated checks cannot judge.
How should a translated spreadsheet be delivered?
Provide an accessible, clearly versioned copy in the format readers need, with a brief note about supported interaction, known limitations and whom to contact if a representation fails. If a PDF summary is provided, test that PDF separately. If the workbook requires a particular application or capability, state the dependency rather than promising universal equivalence.
What should be treated as a release blocker?
Any defect that prevents a reader from identifying the right records, interpreting an important value, completing an essential action, understanding a safety instruction or preserving the calculation’s meaning deserves urgent resolution. Severity depends on the task and audience. A missing decorative description is not the same risk as a mistranslated dosage unit or an unannounced hidden filter.
A concise release decision
The most dependable workbook is not the most elaborately styled one. It is the file in which information relationships remain legible, the labels describe their real purpose, the computations still work, and the language is appropriate to the person making a decision. Use the Accessibility Checker as an early warning system, use human testing as a reality check, and publish only the version you have actually reviewed.