Elsom Cellars: Design Once, Edit Forever

Overview

A fresh design, and a menu that keeps changing

Elsom Cellars is a winery that had been making its menus in-house with tools like Canva. They asked me to redesign their food and wine menus using the logos they already had. I designed them in Figma: an 11×17 tabloid food menu printed front and back, and a one-sided wine list. Both use the Elsom wordmark, a watermark, and a five-weight type system pairing Barlow Condensed with Cormorant Garamond.

The design was the easy part. The hard part was what happened every time a price or menu item changed. The winery wanted to update its own menus without calling me, and without breaking the design. I researched the tools they already knew, found that their "locks" don't actually lock, and shipped a spreadsheet-to-print pipeline that keeps design and content in separate hands.

First Pass

The obvious build, and why I stepped back

My first brief was simple: a site where someone signs in, edits the text on the menu but never the formatting or typography, and exports a PDF. I kept the feature set deliberately small. I rebuilt the Figma artboards as a code template that matched them column for column, with the real typefaces embedded in the exported PDF.

One correction came early, and it was on me for not being clearer. The first version treated the page as a PDF editor. What I actually wanted was a plain form for editing text, with the PDF generated only at export time. Once that was fixed, I restyled the interface with Material Design patterns in Elsom's palette.

It worked. But it also came with a sign-in, a save endpoint, and a server-side content store. That's three things someone would have to host and keep running indefinitely, for a menu that changes a few times a year. Before committing to that, I wanted to know whether I should be building anything at all.

Reframing

The questions that changed the scope

I paused the build and treated it as a product question rather than an engineering one. Leveraging the Compound Engineering skills with Claude, I worked through the problem before writing any more code. Four questions reshaped the requirements:

Research · Buy Before Build

Auditing the tools they already knew

I evaluated design tools and menu-specific products against one core requirement: staff can edit text but cannot break the layout.

Vendor pricing and documentation, August 2026.

Then I learned the winery already paid for Canva. That removed the cost argument, but cost had never been the deciding factor. No Canva plan, at any tier, stops an editor from changing the layout. So Canva became the documented fallback: a no-cost way out if the new approach didn't work for them.

I also set a baseline. At a few small edits a year, me making the changes in Figma costs maybe an hour or two annually. Any solution had to beat that, both in money and in maintenance.

The Decision

Content in a spreadsheet, design in Figma

I recommended moving the menu content into a Google Sheet, a tool the winery already knew how to use. My template, layout logic, fonts, and PDF export stay exactly as built, and they read from that sheet. Everything with an ongoing cost gets deleted: the sign-in, the save endpoint, the content store, and the server itself. The result is a static page on free hosting.

Goals

Answering the four questions

Each question from the reframing became a goal. The spreadsheet approach met all four.

The Designs

Guardrails for editing without a designer

With no designer reviewing the output, the tool itself has to catch mistakes.

Content lives where they already work

Each row is a Section or an Item, picked from a dropdown, and the order in the sheet is the order on the menu. No cell controls font, color, size, or column, so the design can't be changed from the sheet.

A preview they can trust

My one must-have was a preview before export. The preview and the PDF read from the same design values, so what the winery sees is exactly what prints.

Food menu, front and back, rendered from the spreadsheet.

A gate instead of a guess

When content runs past the page, the preview names the column that's too long and disables export. A deliberate override is there if they need it, so they're never stuck.

Live, October 2026: the fall wine list ran long, and the gate caught it before it reached the printer.

Layout rules learned from real content

Column placement is calculated from the content, so nobody has to choose it. I tuned the rules against the real artboards, with a separate set for the one-sided wine list.

Failures that explain themselves

Every problem gets a plain-language message instead of a blank page. If the sheet can't be reached, a row can't be read, or a column is missing, the page says what's wrong and holds the export.

Outcome

Smaller to maintain, safer to use

Both menus are live as a static site on Vercel. The winery edits one spreadsheet with a tab per menu, opens the page, reads the preview, and exports the print-ready file. I stay the designer of record in Figma, and content changes no longer come through me.

Reflection

What I'd carry forward