A restaurant menu hard to read on mobile does not necessarily mean the restaurant needs a new website. Start by repeating the journey from its official website and its public business listing. Try to find one dish and its price at two viewport widths. Then identify the problematic version and who can fix it. If you cannot establish those details, ask a question instead of sending a diagnosis.
This guide is for small web agencies that want to propose a useful, limited intervention. It includes a verification protocol, a recording template, three fictional cases and criteria for checking the delivered work. It assumes no conversion uplift. The examples explain a method, not observations made at real restaurants. The technical references were checked on 22 November 2026.
Turn the observation into a reproducible record
The most valuable evidence is a record the restaurant can reproduce without you. Note the date, location name and address, source, exact route, two widths, and the result for both the dish and its price. Add a screenshot only if it clarifies the observation, limited to the relevant area. Avoid copying the whole menu or sharing contact details, reviews or photos of people unrelated to the finding.
Use this template in your working spreadsheet:
- Restaurant and version: name, address, stated service and menu origin if identifiable.
- Source and route: starting page, link label, final address and observation date.
- Task tested: identify a dish and its price, or record that no price is displayed.
- Width one / width two: browser, orientation, result and observed difficulty.
- Verification: repeat check, legible alternative version and remaining unknowns.
- Decision: make no proposal, ask a question, or offer a limited fix.
- Acceptance criterion: task the restaurant can repeat after the change.
Fill in the decision last. If the route works on the website but not through a gallery image, distinguish the image issue from the primary journey. If the problem is only reproduced when following an old link still shown prominently, the fix may be the destination rather than the menu. If several findings coexist, start with the one that affects the most direct route the restaurant controls, rather than the one that makes the most dramatic demonstration.
A record does not have to prove an impact on sales. Without a measurement designed for that purpose, you can describe a harder task, not a number of lost customers. Show the difference between what you can see and what you infer. That lets the restaurant correct your interpretation. A document may be intended for printing, a set menu may only be available at one time of day, or the team may already be preparing an update.
Three fictional cases: no pitch, a question, or a fix
Fictional case one: the gallery is old, but the official menu works. The photo view contains a menu image that is difficult to read, but the official link leads to a page that clearly shows the dish and price at 360 and 390 CSS pixels. The image displays an old date with no further context. Decision: do not propose a replacement menu on this evidence alone. You can record the unconfirmed source of the image and close the review. The restaurant does not necessarily control that photo.
Fictional case two: the right route depends on the service. The site shows a lunch set menu, and the public listing links to a dinner menu in a PDF. Both may be legitimate. The document looks small at 360 CSS pixels, but the restaurant does not clearly indicate which version it wants to feature. Decision: ask one short question about the services and intended links. Do not replace a document on the assumption that it is old, and do not price a fix before its scope is clear.
Fictional case three: the authoritative menu is a scan that is hard to read. The website and public menu link open the same document. At both widths, enlarging the view sends the dish name and price into separate areas. The restaurant confirms this is the intended version and authorises the agency to prepare a fix. Decision: propose a readable text-based menu page, get approval for one sample section, then add the agreed sections. The printable file can remain available as an option if it serves a confirmed purpose.
These cases do not form a universal severity scale. They illustrate why the decision depends on source, reproduction and control over the links. You can observe a real problem and still recommend selling nothing if the affected version is secondary. Conversely, an incorrect link can justify a small intervention even if the document it should point to is already easy to read.
Propose the smallest fix that solves the journey
Choose one primary fix. A hard-to-find button can be renamed and repositioned on the homepage. If the link opens the wrong version, correct the destination with the owner’s agreement. If the official document is the problem, propose presenting sections, dishes and prices as text on a web page the team can edit. Keep the PDF for printing if that use is real. A QR code pointing to the same scan does not solve the reading problem.
You can use optical character recognition to recover text from a scanned document. The W3C describes this technique for scanned PDFs. Read the resulting text in its intended reading order and check names, prices, symbols and sections with the restaurant. Do not fill missing information from an automatic suggestion. All menu information, including details about ingredients and allergens, must remain the restaurant’s responsibility. Do not infer anything from dish names alone.
A useful proposal states the number of sections, the links you will handle, the source of the supplied content and who will review it. Distinguish preparation on a test environment, approval of the sample section and authorisation to put it live. Changes to the public listing belong in the scope only if the owner wants them and the necessary access is authorised. Do not promise when a third-party service will display an update.
Define what is excluded as well: rebranding, booking setup, changing prices or updating listings across multiple locations. Price any later expansion separately. For first contact, keep to a reproducible route and an offer without pressure. To adapt your message, see the guide to agency outreach email templates. To select businesses for manual review, see SprintLead’s company research page. The tool does not replace manual checks: verify the issue in the public journey before relying on it.
A short message and a check the restaurant can repeat
You can adapt this message if the source and observation are confirmed: “Hello, I opened the menu using the button on your website. On a narrow screen, I had to zoom and move the page to match a dish with its price. Is that the version you want people to see? If so, I can show you a sample for one section without rebuilding your site.” Remove any detail you did not observe. The purpose is to start a conversation, not announce work already decided on the restaurant’s behalf.
Once the work is agreed, plan validation before changing the links. Have the supplied names and prices reviewed and ask the owner to choose a sample section. Check the menu button at 360 and 390 CSS pixels, then repeat the initial task: find the dish and its price without losing their relationship. Confirm on a real device if possible, and give the team instructions to change a price or remove a dish. A page that reads well but nobody knows how to edit may become another competing version.
Prepare a proportionate fallback. Before replacing a link, record its previous destination so you can restore it if the release has a problem. Keep an optional printable document if it has been approved. After release, check the routes you actually changed. Do not promise that third-party photos or external copies will disappear. Record what changed, what did not, and who will check the next update.
Frequently asked questions
- Is every PDF a bad mobile menu?
No. Start by checking the journey and the task. A clear, easy-to-use file can be sufficient for some uses and convenient to print. A scan that requires panning to match a dish with its price presents a different problem. A text-based web page can also make some updates easier, provided the team can edit it.
- Can I pitch a fix if only an old photo is unreadable?
Not on that evidence alone. First check the official destination and the route from the public listing. If the main menu is legible and current, a secondary photo does not automatically justify paid work. If sources conflict, ask the owner which version should be authoritative instead of attributing an image of unconfirmed origin to the restaurant.
- Do two widths prove the menu is accessible?
No. They document a task at two specified widths. They do not cover every device, zoom level, navigation method or assistive technology. Carry out further checks if they are in scope, and do not present the procedure as an accessibility certificate. It is designed to describe and compare a specific journey.
A good proposal starts with simple evidence: the right version, a reproducible route, and a task that fails or becomes unnecessarily difficult. If the official menu works, stop there. If the source is ambiguous, ask. If the problem is confirmed, propose the smallest fix the restaurant can review, accept and maintain. That is more useful than pitching a redesign before you have found the right menu.


