Skip to content
Get Ranked On TopIllustrative case studySignal plan

A fictional business. The figures are a modelled scenario, not a client result.

CARPA BISTRO · LJUBLJANA, SLOVENIA

How a Ljubljana river-fish bistro stopped showing Google a blank board

+156% organic table bookings

7 months · Europe · Signal plan

  • Server-side rendering
  • Core Web Vitals
  • Restaurant schema
  • Render monitoring

Today’s specials

Organic table bookings / month
+156%
“restaurant ljubljana” position
Not in top 100 → #7
Mobile LCP (p75)
5.8 s → 1.9 s
Words in the raw HTML
38 → 1,240

60 covers · private room for 14

01THE CHALLENGE

A full kitchen behind a blank board.

Menu, opening hours and bookings were drawn by JavaScript in the browser, so Googlebot saw a near-empty page and the bistro fell out of “restaurant Ljubljana”.

Carpa’s menu changes with the morning catch: trout from the Soča valley, carp and pike from the ponds along the Sava. The site had been rebuilt as a single-page app, which made the daily board easy for the kitchen to update and close to invisible to search engines. Googlebot fetched an empty shell, queued it for rendering and often indexed it before the menu ever appeared. Phones waited almost six seconds for the terrace photo. Over one spring the bistro slid from page one for “restaurant Ljubljana” to nowhere at all, and online bookings came almost only from guests who already knew the name.

WHAT GOOGLEBOT FETCHED · HOME PAGE

<!doctype html>
<html lang="sl">
<head><title>Carpa Bistro</title></head>
<body>
  <div id="app"></div>
  <script src="/js/app.js"></script>
</body>

Loading…_

38 words

in the raw HTML. Not one dish, opening hour or address.

MOBILE FIELD DATA, P75

  • LCP 5.8 s
  • INP 420 ms
  • CLS 0.31

02DIAGNOSIS

Four tickets from the render audit.

  1. TICKET 1/4RAW HTML

    38 words

    The server sent an empty container and a script tag. Menu, hours, address and the booking form were all built in the browser, after the JavaScript had downloaded and run.

  2. TICKET 2/4INDEXING

    18 of 41

    Only 18 of 41 URLs were indexed. Seasonal menus, the private dining room and the wine list sat in “Crawled – currently not indexed”: to Google they were copies of the same empty shell.

  3. TICKET 3/4JAVASCRIPT

    1.4 MB

    One 1.4 MB bundle for every page: the whole menu archive, a map library and a third-party booking widget, all parsed before anything could be tapped. INP at p75 was 420 ms.

  4. TICKET 4/4CORE WEB VITALS

    5.8 s

    Mobile LCP at p75 was 5.8 s. The terrace photo was only requested after hydration, with no size set, and the layout jumped twice (CLS 0.31) as the menu and the cookie bar arrived.

03THE STRATEGY

Four courses, served in order.

Built on our Signal plan: the technical engagement for sites where crawling, rendering and speed decide whether anyone finds you.

Technical SEO

  1. 1

    Server-side rendering

    COURSE 1

    Every page now renders on the server: the day’s board, opening hours, address and the private dining room arrive as finished HTML. The kitchen still updates the menu from its tablet; the page is rebuilt within a minute.

  2. 2

    Core Web Vitals

    COURSE 2

    The terrace photo is preloaded, sized and served as AVIF. The booking widget waits behind a lightweight facade until a guest taps it, the bundle is split per page and hydration is limited to the parts that are actually interactive.

  3. 3

    Restaurant schema

    COURSE 3

    Restaurant, Menu and OpeningHoursSpecification markup generated from the same data as the page, seasonal closures and the private room’s capacity included, so Google never reads two versions of the truth.

  4. 4

    Render monitoring

    COURSE 4

    Every night a job fetches the key URLs raw and rendered, compares word counts and headings, and alerts the team if a release ever ships a blank board again.

04WIPE THE BOARD

Raw HTML or rendered page? Drag the cloth and see what Googlebot gets.

Right of the cloth: the old client-rendered page, exactly as fetched. Left of it: the same URL with server rendering switched on. Each mark on the rail is one fix, and the readouts change as the cloth passes it.

Drag the cloth across the board, or focus it and use the arrow keys.

READOUTS · MOBILE P75, MODELLED

Nothing wiped yet

Client-side rendering only: an empty shell and a script tag.

LCP
5.8 s
Poor
INP
420 ms
Needs work
CLS
0.31
Poor
Words in the raw HTML38 words

FIXES ON THE RAIL

Nothing wiped yet. LCP 5.8 s, INP 420 ms, CLS 0.31, 38 words in the raw HTML.

05MONTH BY MONTH

7 months, one board at a time.

  1. MONTH 1

    Render audit

    All 41 URLs fetched raw and rendered, the indexing report read line by line, field and lab baselines on mid-range phones.

  2. MONTH 2

    Server rendering

    Server rendering built and tested on staging, the kitchen’s menu editor left exactly as it was, launch before the May terrace opening.

  3. MONTH 3

    Core Web Vitals

    Hero preload and sizing, the booking facade and the bundle split, with field data watched template by template.

  4. MONTH 4

    Schema and hours

    Restaurant, Menu and opening-hours markup generated from the menu data; Business Profile hours brought into line.

  5. MONTHS 5–7

    Season and monitoring

    Nightly raw-versus-rendered checks through the summer peak; new pages for the private room and the riverside lunch menu.

06RESULTS

A full board, and the tables to match.

ORGANIC TABLE BOOKINGS / MONTH
+156%
“RESTAURANT LJUBLJANA” POSITION
Not in top 100 → #7
MOBILE LCP (P75)
5.8 s → 1.9 s
WORDS IN THE RAW HTML
38 → 1,240

Modelled scenario for a fictional business — not a client result. Rankings and traffic are never guaranteed; see our Terms.

Organic table bookings per month150 → 384 a month
100200300400150Before148Mar156Apr205May262Jun318Jul356Aug384Sept
restaurant ljubljana: Google positionNot in top 100 → #7
page one110100+BeforeMar64Apr31May17Jun11Jul8Aug7Sept

Core Web Vitals, mobile field data (p75)

LCP5.8 s → 1.9 s
INP420 ms → 150 ms
CLS0.31 → 0.04

The white chalk line marks Google’s “good” threshold.

Queries back on the board

QueryBeforeMonth 7
restaurant ljubljana100+#7
restavracija ljubljana#58#5
fish restaurant ljubljana#23#2
ribja restavracija ljubljana#19#1
private dining ljubljana100+#4
kosilo ob ljubljanici#41#3

Queries as typed, in Slovenian and English. Positions are modelled; “100+” means outside the top 100.

Is your menu only there after the JavaScript runs?

If your pages are built in the browser, Google may be reading a blank board. Start with the Signal plan, or tell us what your site is built on.

← All case studies

carpabistro.example