Classify address families using examples
Collect the base category, one filter, multiple filters, different sorting and pagination. Include invalid values and different orderings of the same selection. Note what changes the collection, what changes order and what serves a distinct search task.
Fictional example: a box category and a shared material selection need not be independent search landing pages. A special selection with a genuine buyer task needs justification and maintenance. Make that an editorial decision rather than counting technically possible combinations.
Document one coherent technical model
Specify permitted parameter or path formats, normalised ordering, duplicate selections and the links generated by the interface. Keep crawl control, indexing decisions and duplicate canonicalization distinct. A tag is not one solution for every address type.
Consult Google's current faceted-navigation guidance when deciding responses to empty or invalid combinations. Record a model appropriate to the site and test it. If a crawler cannot fetch an address, do not assume it reads that page's HTML instructions.
| Address family | Decision required | Evidence to collect |
|---|---|---|
| Base category | Intended stable destination | Content, response, links and canonical |
| Shareable visitor selection | Represent state without automatic page production | Direct opening and actual access model |
| Approved filter landing page | Distinct purpose and maintenance | Specific content and coordinated signals |
| Sorting only | Same selection in another order | Working state without conflicting destination |
| Invalid or impossible combination | Clear response at the requested address | Actual status and useful reset route |
Exercise ordering, duplicates and empty selections
Open the same combination with parameters in different orders. Check that result and intended URL follow the policy. Try a duplicated filter and an unknown value. Record the actual server response; a polished error panel does not reveal its HTTP status.
Test an empty selection through the interface and by direct address. Distinguish temporarily empty catalog data from a combination that does not exist in the model. Reset should help visitors without a misleading redirect to a generic error page.
Verify discovery and publication after changes
Confirm products and approved categories have real links independently of anyone first choosing filters. The sitemap should reflect intended indexable destinations rather than automatically listing every combination. Include an unaffected selection as a control.
Save tested URL, state, response, visible links, canonical and applied crawl rule. External tools may later show how a search engine actually handles those addresses where access exists. Local tests establish the sampled implementation, not future search-engine choices.