How the Score is Calculated

Your Cassian Score™ is a weighted composite of five layers. Each layer measures a distinct aspect of your store's health. Every time a scan runs, Cassian™ evaluates every accessible page on your store — using a mix of deterministic checks and AI analysis — and feeds the results into this model.

Here's exactly what's inside each layer.

---

Layer 1 — Technical Health (30%)

Technical Health is the heaviest layer, and intentionally so. Technical problems are the most likely to affect your customers directly — broken links, missing metadata, and misconfigured crawl settings hurt both search visibility and user experience.

Cassian runs deterministic checks on every crawled page:

• Broken links — Internal links returning 4xx or 5xx errors. Broken links frustrate customers and signal low-quality content to search engines.
• Meta titles — Checks for missing titles, titles under 30 characters, titles over 60 characters, and duplicate titles across pages.
• Meta descriptions — Checks for missing descriptions, descriptions under 70 characters, descriptions over 160 characters, and duplicates.
• Schema markup — Validates the presence and structure of Product, Breadcrumb, FAQ, and Organisation structured data. Missing schema reduces your store's visibility in search results and AI shopping experiences.
• Image alt text — Images without descriptive alt text are flagged. This affects both accessibility compliance and SEO.
• SSL & HTTPS — Flags mixed content warnings and any page-level SSL issues (such as images or scripts loading over HTTP on an HTTPS page).
• Canonical tags — Validates that canonical tags are present and point to the correct URLs. Incorrect or missing canonicals can cause search engines to index duplicate pages.
• Redirect chains — Multiple consecutive redirects (A → B → C) slow down your store and dilute link equity. Cassian flags chains of two or more hops.
• robots.txt — Validates that your robots.txt file isn't accidentally blocking search engine crawlers from important pages.
• XML sitemap — Checks that a sitemap exists, is valid XML, and contains your key pages (homepage, product pages, collection pages).
• llms.txt — Checks for the presence of an llms.txt file at your store's root. This is an emerging standard that tells AI shopping assistants what your store sells. Missing llms.txt is one of the most common quick wins.
• Core Web Vitals signals — Cassian notes performance signals observed during the crawl. For detailed daily Lighthouse scores, see the PageSpeed page.

Technical Health and LLM Readiness checks are entirely deterministic — they run as instant, rules-based analysis rather than AI evaluation. This means they're free, consistent between scans, and unaffected by AI provider availability.

---

Layer 2 — Content Quality (25%)

Cassian AI reads the actual text on your pages — not just the structure — and evaluates it for quality, completeness, and clarity. This layer is about whether the words on your store are doing their job.

What Cassian AI checks:

• Spelling and grammar — Flags errors in product titles, product descriptions, collection page copy, and blog posts. Even one or two errors on a product page can undermine customer trust.
• Thin content — Product pages with very short or absent descriptions are flagged. Cassian uses a threshold around 150 words of meaningful prose — not including spec tables or bullet lists of attributes. Thin descriptions hurt both conversion rates and search rankings.
• Duplicate content — Product descriptions that are identical or near-identical across multiple pages. Common when product variants share descriptions or when catalogue content is copied from a supplier.
• Heading hierarchy — Every page should have exactly one H1 tag. H2 and H3 tags should follow a logical hierarchy. Cassian flags missing H1s, multiple H1s, and heading structures that skip levels (e.g., H1 → H3 with no H2).
• Readability and clarity — Flags content that reads as unusually complex or uses jargon inconsistent with an e-commerce context. Cassian AI evaluates this in the context of your product category.
• FAQ suitability — Evaluates whether FAQ content is structured in a way that's likely to surface in AI-generated search answers. This overlaps slightly with LLM Readiness — FAQ quality matters for both layers.

---

Layer 3 — Translation Quality (20%)

Translation Quality is only active when your store has translations enabled. If your store is monolingual, this layer is excluded and the remaining four layers are reweighted to total 100%.

What Cassian AI checks across translated content:

• Machine translation artifacts — Phrases and sentence structures that appear to be unedited machine translation output. These are often grammatically correct but stylistically off, or use terminology that doesn't match industry norms in the target language.
• Missing translations — Pages, product descriptions, or metafields where translated content is absent entirely. A customer browsing in French who lands on an untranslated product page is likely to leave.
• Variant translation completeness — Product variants (size, colour, material, etc.) that haven't been fully translated. Incomplete variant translations are one of the most common translation gaps in Shopify stores.
• RTL rendering issues — For right-to-left languages such as Arabic, Hebrew, and Persian — flags content that doesn't display correctly in RTL layout contexts.
• Metafield translation accuracy — Compares translated metafields against the source language content to verify completeness and consistency.

---

Layer 4 — Platform Consistency (15%)

Platform Consistency requires an active Shopify connection. Cassian pulls data from your Shopify backend and compares it to what actually renders in your storefront. This catches a category of problem that no other tool finds automatically.

What Cassian checks:

• Product data mismatches — Price, title, or availability displayed on the storefront differs from the data in your Shopify admin. This can happen due to theme bugs, caching, or CDN issues.
• Image completeness — Product variants missing assigned images. Customers selecting a colour or size and seeing no image is a common conversion killer.
• Translation metafield accuracy — Translated content stored as Shopify metafields vs what actually renders in the translated storefront.
• Availability inconsistencies — Products or variants showing as purchasable when they're out of stock, or showing as unavailable when inventory exists.

---

Layer 5 — LLM Readiness (10%)

AI-powered shopping assistants — embedded in search engines, browsers, and retail platforms — are increasingly influencing purchase decisions before a customer ever visits a store. LLM Readiness measures how well your store is positioned for this shift.

What Cassian checks:

• llms.txt presence — A plain-text file at your store's root that tells AI systems what your store sells and how to interpret your catalogue. Missing llms.txt is flagged as a quick win. See the LLM Readiness feature page for instructions on adding one to Shopify.
• Structured data completeness — The breadth and accuracy of Product, Offer, Review, and FAQ schema on your pages. Rich structured data is how AI systems extract product information reliably.
• FAQ schema coverage — Pages with properly marked-up FAQ schema are more likely to surface in AI-generated answers.
• Content extractability — Whether product information can be cleanly extracted from your pages. Heavy JavaScript rendering, content behind interactions, and unusual DOM structures can reduce extractability.
• Brand consistency — Consistent brand naming, product naming conventions, and description patterns across your catalogue. Inconsistency makes it harder for AI systems to build a coherent picture of your store.
• Answer-ready formatting — Product descriptions written in complete sentences, rather than pure bullet-point attribute lists, are significantly more usable by AI assistants answering customer queries.

---

How the weights work

When Translation Quality is excluded (monolingual stores), the remaining four layers are reweighted proportionally so the total still equals 100%. The exact reweighted percentages are calculated automatically — you don't need to configure anything.

When Platform Consistency is inactive (no Shopify connection), the layer still shows in your dashboard but doesn't contribute to or detract from your score. Connecting Shopify will activate it.

---

Common questions

Why is Technical Health weighted so heavily? Technical problems are the most likely to directly impact your store's search visibility and customer experience. A store with a broken sitemap, missing meta tags, and multiple redirect chains is genuinely less healthy than a store with thin descriptions. The weighting reflects the real-world impact of each layer.

What if Translation Quality is enabled but my store is only in English? If no translated content is detected during a scan, Translation Quality will score at or near 100% (no issues found). Your overall Cassian Score won't be penalised, but you are paying for a layer that isn't doing anything. Disable translations in Site Settings to have the weights reallocated to the other layers.

Can the score be 100? Yes. A score of 100 means Cassian found zero issues across all active layers during that scan. It's genuinely achievable with a well-maintained store — though maintaining it requires staying on top of new issues as your store content changes.

What's the difference between Critical, High, Medium, and Low severity? Severity determines how much an issue contributes to score deductions. Critical issues (broken checkout, missing SSL, completely missing meta tags sitewide) have the largest impact. Low severity issues (minor readability suggestions, optional schema types) have minimal score impact. The Issues page always shows severity alongside each finding.

Why did my score drop overnight when I didn't touch anything? Cassian scans your live store. If any page content, product data, or linked resources changed — including third-party scripts, partner feeds, or inventory status — those changes may have introduced new issues. The Issues page will show the date each issue was first detected so you can trace what changed.