Architecture advice describes how to build a site correctly. The common situation is inheriting one that grew for eight years without anyone owning the structure.
Map what exists before designing what should
Export every URL, group by pattern, and count pages per section. The output is usually a surprise: whole sections nobody remembered, two competing taxonomies, and a category with four thousand pages that generates no traffic.
Click depth is the metric that translates
Pages many clicks from the homepage get crawled less and rank worse. It is a concrete, measurable property that engineers can act on, unlike "improve the architecture", and it usually points straight at pagination and missing hub pages.
Restructure in stages, never at once
A full re-architecture is a migration with all the associated risk. Sequencing it — hub pages first, then internal linking, then URL changes only where genuinely necessary — captures most of the benefit at a fraction of the exposure.
Leave URLs alone unless the change earns its cost
Restructuring navigation and internal links can be done without changing a single URL, and that captures a large share of the value with none of the redirect risk. Change URLs only where the structure itself is the problem.