Skip to content
Get Ranked On TopIllustrative case studySignal plan

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

Abakor POS · Point-of-sale & inventory SaaS · Sofia

How a 45-person POS company got 1,200 hidden help-centre pages into Google

From 64 to 1,164 docs pages indexed

10 months · Bulgaria · Signal plan · Real URLs · Server-side rendering · XML sitemaps · Hreflang

Docs pages in Google's index

Month 0 → month 10 · of 1,200

64 → 1,164

Four rods, sixteen fixes

01The challenge

A thousand answers nobody could find.

The help centre was a JavaScript app with hash URLs, so 1,200 docs pages were never indexed, and the marketing site ran duplicate language versions with no hreflang.

Shop owners search the moment a till misbehaves: how to connect a fiscal printer, how to move stock between two shops, why a barcode label prints blank. Abakor had written an answer to almost every one of those questions, in five languages. But the help centre was a single-page app that routed with hash fragments, so to Google the whole library was one empty page. On the marketing site, each language lived at two addresses with no hreflang, and searchers in Bucharest kept landing on the Bulgarian pricing page.

What Googlebot fetched from the help centre
  • abakor.example/help#/getting-started/first-sale
  • abakor.example/help#/inventory/stock-transfers
  • abakor.example/help#/payments/fiscal-printer
  • abakor.example/help#/integrations/webshop-sync

GET /help

200 · text/html

<div id="app"></div>

<script src="/help/app.js"></script>

1,200 docs pages → one indexable URL

Everything after the # stays in the browser: it is never sent to the server, and Google drops it.

The marketing site, before
00 hreflang tags

5 language versions, each reachable at two addresses, and no hreflang between any of them.

02Diagnosis

Four rods, all at zero.

Routing#001


1 URL

for 240 articles in five languages


Every article lived at /help#/section/article. Hash fragments never reach the server, so crawlers saw a single address, /help, with 1,200 docs pages behind it. Old links shared in forums pointed at that same one URL.

Finding 1 of 4

Rendering#002


0 words

of article text in the HTML Googlebot receives first


The server sent an empty <div id="app"> and a 1.9 MB script bundle. Navigation was built from click handlers, not <a href> links, and missing articles answered 200 with a “not found” message instead of a 404.

Finding 2 of 4

Index coverage#003


64

of 1,200 docs pages indexed


Search Console listed 64 docs pages as indexed: old static exports and PDFs from a previous help-desk tool. Four of them were out of date and gave the wrong fiscal-printer settings.

Finding 3 of 4

Languages#004


0

hreflang annotations across five language versions


The marketing site served English at / and /en/, Bulgarian at /bg/ and ?lang=bg, Romanian on a separate subdomain, and Greek and Serbian at two paths each: about 1,800 URLs for 180 pages, all competing with each other.

Finding 4 of 4

03The strategy

Four rods, one bead at a time.

Built on our Signal plan: the technical engagement for sites where crawling, rendering and architecture decide what search engines can see.

  1. Rod 1

    Real URLs

    Path routing replaced the hash, section by section: /help/inventory/stock-transfers instead of /help#/inventory/stock-transfers. A small script redirects old hash links to their new path, and every article got a self-referencing canonical.

  2. Rod 2

    Server-side rendering

    Articles, category pages and the changelog are now prerendered to HTML at publish time and hydrated afterwards, so the first response carries the full text. Navigation became real <a href> links and missing articles return a true 404.

  3. Rod 3

    XML sitemaps

    One XML sitemap per language with accurate lastmod dates, regenerated on publish, plus breadcrumbs and related-article links, so every article sits two clicks from the help-centre home.

  4. Rod 4

    Hreflang

    Each article and marketing page links its four other language versions with reciprocal hreflang and x-default to English. Duplicate addresses were 301-redirected to one URL per language, and the ?lang= parameters retired.

04Count the index

Flick the beads. The abacus totals what Google can index.

Each rod is one technical fix and each bead one step of it. Push beads up to the beam and watch the modelled count of indexed docs pages change.

Rod 1: Real URLs

One bead per help-centre section moved from hash URLs to real paths.

  • Getting started (60 articles) · not yet
  • Inventory (70 articles) · not yet
  • Payments & fiscal (60 articles) · not yet
  • Integrations (50 articles) · not yet

What Googlebot gets for one article

Requested
abakor.example/help#/inventory/stock-transfers
Indexed as
/help
First HTML
0 words of text · 0 crawlable links
Sitemap
not listed
Hreflang
linked languages: 1 of 5

Docs pages indexed

64of 1,200

64 of 1,200 docs pages indexed

The same total, read on a soroban

Indexed by language

  • English13
  • Bulgarian13
  • Romanian13
  • Greek13
  • Serbian13

Beads on the other rods barely move the count until the URLs are real: start with the first rod.

Model: each converted section counts its articles in five languages, scaled by how much of the page is rendered on the server, how well it is discovered and whether its language is linked by hreflang. Illustrative, not a forecast.

05The engagement

10 months, rod by rod.

  1. Month 1

    Audit

    Crawls with and without JavaScript, render tests on every template, three months of server logs and a page-by-page reading of index coverage.

  2. Months 2–3

    Real URLs

    Path routing shipped one section at a time, with the hash-redirect script and canonicals. Inventory went first: it had the most search demand.

  3. Months 3–5

    Server rendering

    Prerendering at publish time for articles and categories, links rebuilt as anchors, true 404s. The docs writers kept the same editor.

  4. Months 5–6

    Sitemaps & links

    Per-language XML sitemaps regenerated on publish, breadcrumbs and related articles, sitemap submission and monitoring.

  5. Months 6–7

    Hreflang & duplicates

    Reciprocal hreflang on the docs and the marketing site, duplicate language addresses redirected, the Romanian subdomain folded into /ro/.

  6. Months 8–10

    Recrawl & monitoring

    Weekly index-coverage checks per language, log files to confirm Googlebot was rendering the new pages, and fixes for the stragglers.

06Results

The index finally adds up.

Docs pages indexed

64 → 1,164

of 1,200 docs pages · 97% indexed

Organic clicks to the help centre

+668%

1,900 → 14,600 clicks a month

Organic trial sign-ups a month

140 → 310

+121% · from search, mostly via docs pages

Searches landing on the wrong language

31% → 3%

Across five markets, marketing site and docs

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

Docs pages indexed, month by monthOne bead = 50 docs pages
04008001,2006412331045700671,010891,16410Month of the engagement
Organic clicks to the help centreclicks a month
05,00010,00015,000123456789101,90014,600Month of the engagement
Searches landing on the wrong language, by market

Queries that now land on a docs page

QueryMarketBeforeMonth 10
свързване на касов апарат с НАПBG—#3
configurare casă de marcat gestiune stocRO—#5
πρόγραμμα αποθήκης για μικρό κατάστημαGR#48#4
program za kasu i magacinRS—#2
pos stock transfer between shopsCY—#7
barcode label printing posBG#61#6

— = not in the top 100. Queries are shown as searched, in their own language. Modelled positions, not guaranteed.

Is your help centre invisible too?

JavaScript docs, hash URLs, language versions competing with each other: start with the Signal plan, or tell us what search engines can't see.

← All case studies