All articles
SEO & GrowthNovember 9, 2026 6 min read

Faceted Navigation Without the Index Bloat: A Rules Engine That Keeps Google Happy

Faceted navigation is where programmatic sites quietly burn crawl budget and dilute rankings. Here's the rules engine we use to decide which facet combinations deserve an index, a noindex, or a hard block.

Faceted navigation is the single biggest source of index bloat we see on ecommerce and marketplace audits. The filters your merchandisers love — colour, size, brand, price band, rating, availability — multiply into millions of URLs that Googlebot will happily chew on, then ignore. The fix is not a blanket noindex. It's a rules engine that treats each facet combination as a decision, not a template default.

Why the usual advice fails

Most SEO guides give you three options for facets: canonicalise everything to the parent category, noindex everything with a query string, or block via robots.txt. Each of those is wrong often enough that we've stopped recommending any of them as a default.

Canonicalising everything throws away real demand. "Waterproof hiking boots size 11" is a query people type. If your /hiking-boots?waterproof=1&size=11 canonicalises to /hiking-boots, you've handed that query to a competitor who built a dedicated landing page.

Blanket noindex is slightly better but still leaks equity. Google will crawl the URL, render it, discover the directive, and drop it — but not before spending budget. On sites with 50k+ products, we've seen noindex facets eat 30-60% of crawl in log files.

robots.txt blocks prevent crawling entirely, which sounds great until you realise blocked URLs can still accumulate backlinks and appear in SERPs as ugly stubs. They also hide canonical signals from Google.

The answer is a decision tree per facet combination, enforced at render time.

The four buckets

Every facet URL on a programmatic site should resolve to exactly one of these states:

  • Indexable: self-referencing canonical, included in sitemaps, linked from internal nav.
  • Canonicalised: renders normally for users, but <link rel="canonical"> points to a stronger parent URL. Not in sitemaps.
  • Noindex, follow: renders normally, <meta name="robots" content="noindex, follow">. Link equity flows through but the URL doesn't compete.
  • Blocked: robots.txt disallow. Reserved for parameters with no SEO or UX value (session IDs, sort orders, tracking).

The job of the rules engine is to assign one of these four states to every URL your filter UI can generate, and to do it consistently as your catalogue changes.

What actually determines the bucket

Four signals, in priority order:

  1. Search demand: does the facet combination map to queries with non-trivial volume in GSC or a keyword tool?
  2. Inventory depth: are there enough products to make the page useful? A facet with two SKUs is a thin page.
  3. Combination order: is this a single facet, a 2-facet combo, or a 3+ stack? Index value drops sharply after two.
  4. Facet type: some facets (brand, category, material) carry semantic weight. Others (in-stock, on-sale, sort-by) carry none.

The rules engine

Here's a simplified version of what we deploy. In production this lives in a shared library called from the template renderer, the sitemap generator, and the internal linking module — three places that must agree.

type FacetState = 'index' | 'canonical' | 'noindex' | 'blocked';

interface FacetUrl {
  category: string;
  facets: Record<string, string[]>;
  productCount: number;
}

const SEMANTIC_FACETS = new Set(['brand', 'material', 'colour', 'size', 'style']);
const JUNK_PARAMS = new Set(['sort', 'view', 'page_size', 'utm_source']);
const MIN_PRODUCTS = 6;

function classify(url: FacetUrl, demand: DemandLookup): FacetState {
  const activeFacets = Object.entries(url.facets)
    .filter(([k, v]) => v.length > 0);

  // Rule 1: any junk param present -> block
  if (activeFacets.some(([k]) => JUNK_PARAMS.has(k))) {
    return 'blocked';
  }

  // Rule 2: multi-select within a facet -> noindex
  // e.g. colour=red,blue is user-useful but SEO-dilutive
  if (activeFacets.some(([, v]) => v.length > 1)) {
    return 'noindex';
  }

  // Rule 3: thin inventory -> noindex, follow
  if (url.productCount < MIN_PRODUCTS) {
    return 'noindex';
  }

  // Rule 4: only semantic facets considered for indexing
  const semantic = activeFacets.filter(([k]) => SEMANTIC_FACETS.has(k));
  if (semantic.length !== activeFacets.length) {
    return 'canonical';
  }

  // Rule 5: depth limit
  if (semantic.length === 0) return 'index'; // bare category
  if (semantic.length === 1) {
    return demand.hasVolume(url) ? 'index' : 'canonical';
  }
  if (semantic.length === 2) {
    return demand.hasStrongVolume(url) ? 'index' : 'canonical';
  }
  // 3+ semantic facets stacked -> canonical to 2-facet parent
  return 'canonical';
}

The demand lookup is where this stops being generic. We back it with a weekly job that pulls GSC data into BigQuery, aggregates impressions per facet combination, and flags combinations above a configurable threshold. A facet URL is "demand-backed" if it has impressions, a plausible query match, or matches a curated keyword list from the merchandising team.

The canonical target matters

When a URL is classified canonical, you need to decide what it points to. The naive answer is the bare category. The better answer is the strongest ancestor in the facet graph. /boots/hiking?brand=salomon&waterproof=1&size=11 should canonicalise to /boots/hiking?brand=salomon&waterproof=1 if that exists and is indexable, not all the way back to /boots/hiking. You're preserving topical specificity.

We compute this by walking the facet set, dropping one facet at a time in reverse priority order, and picking the first ancestor that classifies as index.

Internal linking has to match

The single biggest mistake we see is a rules engine that classifies URLs correctly but a filter UI that still emits crawlable <a href> links to every combination. Googlebot follows those links, discovers the noindex, and you're back to wasted crawl.

The rule: if a facet URL is noindex or blocked, the filter UI must not expose it as a crawlable anchor. Render it as a button with a data-* attribute and handle navigation in JavaScript, or add rel="nofollow" and data-nosnippet. Reserve real <a href> tags for URLs you want crawled.

Sitemaps follow the same rule. Only index URLs belong in sitemap.xml. We generate sitemaps from the same classification function, which eliminates drift.

Measuring whether it worked

Two dashboards, both wired to GSC and your log pipeline:

  • Crawl share by state: percentage of Googlebot hits going to index, canonical, noindex, blocked, and 404. You want index above 70% within a few weeks of deploying the rules.
  • Impression share by state: of your total GSC impressions, what percentage come from index URLs? If canonical URLs are getting impressions, Google is ignoring your canonical and you have a mismatch to investigate.

In our experience, cleaning up facet classification on a mid-sized ecommerce site moves indexed URL counts down by 40-80% and organic sessions up within 6-10 weeks. The drop in indexed URLs scares stakeholders. Show them the impression chart.

Common traps

Price filters: almost always noindex. They look like semantic facets but nobody searches "hiking boots £80-£120". Exception: round-number price bands in categories where "cheap X" or "X under $50" is a real query.

Availability filters: always blocked or noindex. In-stock status changes hourly and you don't want indexed URLs flipping to empty pages.

Pagination inside facets: rel="canonical" on page 2+ pointing to page 1 is wrong if page 1 can't contain all products. Use self-referencing canonicals and let Google decide. We covered this in more depth on the blog.

Category + brand pages: these are often your highest-value facet URLs. Treat them as first-class landing pages with hand-written intro copy, not generic templated filler. That's the line between a thin facet and a real page.

Where we'd start

Pull a week of raw Googlebot log lines and group hits by URL pattern. Count how many of those hits landed on URLs that are currently noindex, canonicalised to a parent, or returning thin results. That number — the percentage of crawl spent on URLs Google will never rank — is your baseline. Then write the classification function, wire it into your template, sitemap generator, and filter UI in the same PR, and ship it behind a flag so you can measure the before and after in GSC. If you want a second pair of eyes on the audit, our team does this kind of work under technical SEO services.

#SEO#Programmatic SEO#Ecommerce#Technical SEO

Want a team like ours?

72Technologies builds production software for the kind of teams who actually read this blog.

Start a project