Skip to content
Get Ranked On TopIllustrative case studySignal plan

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

Station KP-DUB · Dublin

Crawl trace · months 1–9

Kernelpost · Courier dispatch SaaS, 60 staff · Dublin

How a dispatch SaaS got 478 invisible pages indexed by switching to server-side rendering

118 → 596 pages indexed

9 months · UK & Ireland · Signal plan

  • Server-side rendering
  • JavaScript rendering audit
  • Log-file analysis
  • Internal links & sitemaps

Crawl magnitude

M 5.2 → M 2.7

Pages indexed by month 9

596 / 640

01The challenge

A website that only existed once JavaScript had run.

Kernelpost's React marketing site rendered only in the browser, so feature and integration pages reached crawlers empty and 522 of 640 were never indexed.

Kernelpost sells dispatch and route-planning software to courier firms across Ireland and the UK. Its marketing site was a single-page React app: quick to build, pleasant in a browser and almost blank to a crawler. Every URL returned the same 1.2 KB shell and a 1.9 MB script. Googlebot had to queue each page for rendering, and the integration pages, the ones prospects search for by name, waited on a catalogue API that often answered too late. Sales kept hearing the same line: “we couldn’t find you for Shopify”.

What a crawler received first

Raw HTML1.2 KB · 9 words
<!doctype html><title>Kernelpost</title><div id="root"></div><noscript>You need to enable JavaScript to run this app.</noscript><script src="/static/js/main.4f9c1e.js"></script>
Rendered DOM, when rendering finished41 KB · 610 words

522 of 640 pages never indexed

02Diagnosis

Four tremors on the record.

Event 1 / 4

1.2 KB

Rendering

All 640 URLs returned the same shell: an empty root div, a noscript line and the bundle. Titles, meta descriptions and canonical tags were injected by JavaScript, so the raw HTML pointed every page at the home page.

Event 2 / 4

18%

Indexing

Search Console listed 412 URLs as “Discovered – currently not indexed” and 110 as “Crawled – currently not indexed”. Only 118 pages, mostly the home page, pricing and older blog posts, made it into the index.

Event 3 / 4

66%

Render failures

A rendered crawl with Googlebot-like limits failed on 422 of 640 pages: 307 finished blank and 115 timed out while the partner catalogue API took 6–9 s to answer.

Event 4 / 4

312

Internal links

Navigation used click handlers instead of real links, and the integration directory built its tiles in a filter script. To a crawler, 312 integration pages had no inbound <a href> at all, and the only sitemap listed just 140 URLs.

Search Console coverage
  • Indexed
  • Discovered – currently not indexed
  • Crawled – currently not indexed
03The strategy

Four channels, one quieter trace.

Built on our Signal plan: the technical engagement for sites where crawling, rendering and architecture decide what gets found.

  1. CH 1

    Server-side rendering

    The marketing site moved to server rendering on the same React components: no redesign, no URL changes. Feature, industry and guide pages are pre-rendered at build time; the 312 integration pages are generated from the partner catalogue and refreshed hourly, so the HTML is complete before a single script runs.

  2. CH 2

    JavaScript rendering audit

    Every URL was fetched twice, raw and rendered with Googlebot-like timeouts, and the two were compared: missing titles, canonicals, headings, links and copy. That comparison became the acceptance test for the rebuild. A page shipped only when its raw HTML matched its rendered version.

  3. CH 3

    Log-file analysis

    CDN logs showed where Googlebot spent its fetches before and after the switch. Requests for bundles and API calls fell, HTML requests for integration pages rose, and a hydration error that blanked the pricing table on mobile was caught in the first week.

  4. CH 4

    Internal links & sitemaps

    Click handlers became real <a href> links, the integration directory got crawlable category pages, and one sitemap of 140 URLs became five, one per section, with lastmod dates from the CMS. No page is an orphan any more.

04Read the drum

Flip the rendering switch and watch the tremors settle.

Each row of the drum is one section of the site, traced by a crawler fetching its pages. Red bursts are pages that rendered blank; black bursts are renders that timed out. Switch between client-side and server-side rendering, then pick a row to read it.

Rendering mode
  • Blank render
  • Render timeout
  • Clean fetch

CSR · Whole site

1.2 KB of HTML

Server-side rendering doesn't change what the pages say. It changes when a crawler can read it: in the first response, not after a queue, a 1.9 MB bundle and an API call.

Crawl magnitude

M 5.2

Our own log scale of failed renders, not a Richter reading.

Pages
640
Indexed
118 of 640
Blank renders
307
Timeouts
115
Words in raw HTML · average per page9

What Googlebot gets on the first fetch

<!doctype html><title>Kernelpost</title><div id="root"></div><noscript>You need to enable JavaScript to run this app.</noscript><script src="/static/js/main.4f9c1e.js"></script>

The server sends an empty shell. Googlebot has to queue the page, run the bundle and wait for the API; when that stalls, it keeps a blank page or nothing at all.

Client-side rendering · Whole site: 422 failed renders, 118 of 640 pages indexed.

05Month by month

9 months, from tremor to quiet.

  1. M1422 failed118 indexed
  2. M2418 failed124 indexed
  3. M3409 failed131 indexed
  4. M461 failed236 indexedSSR live
  5. M524 failed401 indexed
  6. M615 failed512 indexed
  7. M712 failed556 indexed
  8. M811 failed581 indexed
  9. M910 failed596 indexed
  1. Month 1

    Audit

    Raw-versus-rendered crawl of all 640 URLs, 12 weeks of CDN logs, Search Console coverage, and a list of every integration sales had been asked about by name.

  2. Months 2–3

    Rebuild

    Same components, rendered on the server. Titles, canonicals and structured data moved into the HTML, links became links, and integration pages were generated from the partner catalogue.

  3. Month 4

    Switch

    Released on the same URLs in one evening, with a rollback ready. No redirects needed; five new sitemaps submitted that night.

  4. Months 5–6

    Recrawl

    Logs watched daily as Googlebot fetched the site again. Two hydration bugs fixed; mobile LCP on integration pages down from 3.8 s to 1.6 s.

  5. Months 7–9

    Compound

    With indexing solved, the thinnest integration pages got real setup guides and cross-links, and every new page was indexed in about two days instead of three weeks.

06Results

The needle went quiet. The pipeline didn't.

Pages indexed

118 → 596

of 640 pages, month 9

Organic demo requests / month

+174%

38 → 104 a month

Non-brand organic clicks / month

+252%

3,100 → 10,900 a month

Time to index a new page

23 → 2 days

Median, from publishing to indexed

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

Pages indexed, by monthKernelpostProjected without SSR
0160320480640SSR live596118M1M2M3M4M5M6M7M8M9
Organic demo requests per month
38M137M241M349M463M578M689M797M8104M9

Rankings for what prospects actually type

QuerySectionMonth 1Month 9
courier dispatch softwareHome & pricing#34#6
shopify courier integrationIntegrations100+#4
route planning software irelandFeatures#19#3
pharmacy delivery softwareIndustries#61#7
multi drop route plannerGuides & blog#27#5

Modelled positions on Google in Ireland. Queries shown as typed.

Is your site blank until JavaScript runs?

If your pages depend on client-side rendering, start with the Signal plan, or tell us what a crawler sees on yours.

← All case studies