Nobody wants to have the crawl budget conversation
Three agencies told a client to write blog posts. The actual problem was that Googlebot was spending its entire budget on 340,000 permutations of the same eleven gaskets.
A distributor came to us with 40,000 products and 300 organic sessions a month. They had been told, by three separate agencies, that they needed content.
They did not need content. They had the best catalogue in their region and Google could not see it.
Where the budget went
Their filters generated links. Length, thread, material, brand — every combination produced a crawlable URL. Multiply those out and the site exposed roughly 340,000 addresses, nearly all of which were the same handful of products in a different order.
Googlebot has a finite budget for any given site. It was spending that budget diligently working through permutations and never reaching the 38,900 real product pages that had never been indexed at all.
Publishing blog posts into that situation would have added more pages to a site that could not get its existing pages crawled. It would have felt like progress for about five months.
Why this advice is rare
Content is a lovely thing to sell. It produces visible artefacts every month, it is easy to put in a report, and everyone in the meeting understands what they bought.
Canonicalising a faceted navigation produces nothing you can show anyone. There is no deliverable to admire. It requires engineering time you have to argue for, from a team who did not ask for this and have their own roadmap.
So it does not get recommended. Not usually through cynicism — it just loses every pitch to the option with pictures in it.
What actually happened
We cut the crawlable URL space by 94%. Indexation went from 1,100 products to 8,900 and organic sessions multiplied fourteen-fold inside a year.
We deliberately did not chase all 40,000. About three quarters of the catalogue are low-volume variants that would only dilute the pages worth ranking. Indexing everything is not a goal; it is a failure to choose.
Then we spent six weeks on something nobody asked for: piping analytics IDs into their ERP, so a phone order placed three weeks after a search could be traced back to it. Without that the whole project would have been a graph of sessions, and sessions do not pay for anything.
The general shape
If your site is large and your traffic is flat, look at what the crawler is spending its time on before you commission a single word. Server log analysis is dull and it is frequently the entire answer.
The constraint is rarely the thing in the brief.
Spent nine years as a backend engineer before discovering that most SEO problems are engineering problems wearing a marketing hat. Has migrated four ecommerce platforms without losing rankings, and one with, which he will tell you about at length.