Case study — 347,000 product pages that Google cannot read

1The number that matters

340,981

product pages published

keywords ranking anywhere

39%

of traffic is the company name

A catalogue this size should be a search asset — every part number is a query someone types. Instead the site earns fewer visitors than a well-run ten-page business site, and four in ten of those visitors already knew the company.

2The cause, in one test

We fetched 28 URLs the way a crawler does, then opened the same URLs in a real browser.

What a crawler receives

Identical 126,007-byte shell on every URL — home page, category, product, brand page
No title tag
No meta description
No H1
No canonical
Zero words of body text

What a browser shows after scripts run

Title appears — the raw part description
Product specifications and copy appear
Category listings appear
Still no H1 on product pages
Still no Product structured data
Home page still has no meta description

The entire catalogue is client-side rendered. Google can render JavaScript, but it does so in a second pass, on a budget, and it decides how much of a site deserves that budget. At 347,000 URLs, a site asking for rendering on every page is asking for more than it will get — and the evidence is in the ranking count.

The proof that this is the binding constraint and not the market: the site holds #1 for "lab supplies" — 1,300 searches a month. One category where the page carries real text ranks first. The other 3,470 categories do not.

3What else the audit found

Finding	Scale	Why it costs money
Product pages have no H1	all sampled	Even after rendering. The part name — the exact thing people search — is never marked as the page heading.
No Product structured data	catalogue-wide	No price, availability or review markup anywhere. In a category where competitors show price and stock directly in results, this site cannot be shown at all.
Home page: no meta description, H1 reads "Explore available brands", no H2s	1	The most linked page on the domain describes itself to Google as a brand directory.
Category pages: 146 words, description duplicates the H1	3,471	Nothing to rank with. This is the layer that should capture "band saw blades" and "industrial fasteners".
An error page is listed in the sitemap	1	`/UnhandledError` returns HTTP 500 and is submitted to Google for indexing.
Account pages submitted for indexing	~40	Create-account, cart, checkout, quick-order, RFQ. Two were verified live: "create account" and "your password reset link has expired" both return 200 and are marked indexable.
Duplicate URLs inside the sitemap	24	Small, but it is the signal of a sitemap that is generated without validation.
Non-www redirects with a 302, not 301	sitewide	A temporary redirect on the canonical host decision, in place indefinitely.
Images without alt text on the home page	6 of 42	Minor next to the rest, listed for completeness.

4The demand they are missing

US monthly search volume for categories already present in this catalogue.

Search	Per month	Difficulty	Cost per click	Status
welding supplies	3,300	29	$0.80	not ranking
band saw blades	3,100	1	$0.90	#12
safety supplies	1,400	32	$2.50	not ranking
lab supplies	1,300	39	$1.90	#1
industrial fasteners	1,100	0	$1.50	not ranking
mro supplies	1,000	0	$1.50	not ranking
abrasives supplier	200	68	$0.45	not ranking
masking tape bulk	100	0	$0.90	#6
industrial hose clamps	70	0	$1.20	#6
hard hats bulk	40	1	$1.40	not ranking

11,690 searches a month in the head terms alone — before counting the part-number searches that a 340,000-product catalogue should own outright. "Band saw blades" sits at #12 against difficulty 1: a category page with real copy moves that onto page one.

The plan, in order

Serve HTML that contains the page. Server-side rendering or pre-rendering for category and product templates, so title, description, H1, canonical and body copy exist before JavaScript runs. Nothing else in this plan produces a result until this is done.
Give product pages an H1 and Product markup — part name as heading, price and availability as structured data. One template change across 340,981 pages.
Write the top 50 category pages — 300-600 words of buying guidance, specification help and cross-links. Prioritised by the demand table above; "lab supplies" is the proof this works.
Clean the sitemap — remove the 500 error page, the account and checkout URLs, and the duplicates; noindex the account area; make the generator validate what it publishes.
Fix the 302 on the canonical host.
Then scale to part numbers. With rendering fixed and a product template that carries headings and markup, the long tail of 340,000 part numbers becomes reachable — that is where a catalogue of this size eventually earns its traffic.

5Link profile — not the problem

domain rating

859

referring domains (live)

1,020

live backlinks

Authority is respectable for the sector — acquisitions and trade press have earned this domain real links. It is not what is holding the site back.

Competitor	Authority	Traffic	Keywords
Tooling manufacturer	74	7,217	1,137
Small online tool retailer	20	3,381	2,396
Regional supply distributor	16	1,395	1,104
This distributor	31	659	99

The second row is the whole argument. A retailer with authority 20 against this site's 31 ranks for 2,396 keywords and takes five times the traffic. Its pages are server-rendered HTML with headings and text. No amount of link building closes a gap that is caused by pages a crawler cannot read.

Link budget

Item	Volume	Why
Nothing, until rendering is fixed	0	Links pointing at pages with no title, no H1 and no text are spend with no mechanism to pay off.
Afterwards: manufacturer and distributor-locator links	20-40	Brands this distributor carries maintain "where to buy" directories. Relevant, free, and rarely claimed.
Trade press around acquisitions	ongoing	Press-release pages already rank for acquired company names — the pattern works, it simply is not being used deliberately.
Supplier and association directories	10-20	Industry bodies and safety associations; standard, durable B2B citations.

6What is not in this report

No Search Console or analytics access. Everything is from a public crawl, a rendering comparison and third-party search data. Traffic figures are estimates.
28 URLs sampled across every template type, plus full analysis of the 347,177-URL sitemap. A complete crawl of a catalogue this size is a project in itself and would be the first deliverable.
Rendering diagnosis is behavioural, not architectural. We can prove what crawlers receive; naming the framework and the fix inside the stack needs access to the codebase.
Part-number demand not sized. Estimating the long tail across 340,981 products needs a sampling study, scoped separately — the figures above are head terms only.