# DTC site improvement brief

Site: Choose the merchant domain
Category: Home & lighting
Contract: dtc-1.0
Scope: Evidence-led content planning. Not a visibility score or proof of agent checkout.

## Establish the baseline
Run the existing Agent Friendly basic check; keep its run ID, profile, version and sampled URLs. Inspect current products and policies before suggesting changes.

## Four customer tasks
1. I need a warm lamp for a small bedroom.
Page: /guides/bedside-lighting-small-bedroom
Title: How to choose a bedside lamp for a small bedroom
Answer: Explain how dimensions, light temperature and dimming affect a bedside setup. Recommend a named variant only after verifying these facts.
Facts to verify:
- Dimensions and required bedside clearance
- Color temperature in kelvin, by bulb or variant
- Dimming and switch compatibility
- Included bulb, socket and plug region
Limit: Do not infer warm light from a styled photograph. State when a bulb must be bought separately.
Comparison: Warm vs neutral light for bedtime reading (/compare/warm-vs-neutral-reading-light)

2. Can I install a wall light without drilling?
Page: /guides/wall-lighting-without-drilling
Title: Wall lighting for renters: what needs drilling?
Answer: Separate plug-in power from mounting requirements. Describe the actual mounting method, surface limitations and removal instructions.
Facts to verify:
- Mounting method and fasteners
- Supported wall surfaces and load limits
- Cable length and socket requirements
- Removal instructions and possible damage
Limit: Plug-in does not mean drill-free. Never promise damage-free removal without evidence.
Comparison: Plug-in sconces vs portable table lamps (/compare/plug-in-sconce-vs-table-lamp)

3. Will this dining table fit four people in my apartment?
Page: /guides/four-person-table-small-apartment
Title: Measure a small dining space before choosing a table
Answer: Give exact table dimensions, recommended chair clearance and the assumptions behind any seating claim.
Facts to verify:
- Table dimensions and extensions
- Chair and walking clearance assumptions
- Delivery package size
- Assembly and doorway requirements
Limit: Show a worked measurement example, labeled as illustrative. Do not call every four-seat table suitable for a small room.
Comparison: Round vs rectangular tables in small spaces (/compare/round-vs-rectangular-small-space)

4. Which sofa fabric is easiest to clean with a dog?
Page: /guides/sofa-fabric-cleaning-pets
Title: Compare sofa fabrics for cleaning and everyday wear
Answer: Explain the documented care method and fabric construction. Identify any tested durability claim and its source.
Facts to verify:
- Fabric composition and weave
- Published cleaning instructions
- Removable cover and wash limits
- Documented abrasion or stain-test results
Limit: Do not turn easy-clean marketing into a pet-proof guarantee.
Comparison: Removable covers vs fixed upholstery (/compare/removable-vs-fixed-sofa-covers)

## Small launch plan
4 × Use-case guides: One useful answer for each of the four selected shopping tasks. Link the exact qualifying variants and the original evidence.
3 × Honest comparisons: Compare measurable attributes and exclusions. Begin with product types or your own variants; use competitors only with dated primary evidence.
2 × Fit and compatibility guides: Resolve sizing, installation or formulation identity with a worked example and an acceptance check.
1 × Product facts index: A canonical index of product and variant facts that points to the source of truth. Reuse existing product pages instead of duplicating them.
1 × Purchase conditions: Connect delivery, returns and availability to their existing current policies. Do not create duplicate policies.
1 × Material product updates: Record real changes to formulation, compatibility, availability or documentation. Never manufacture weekly news.
Treat this as a content inventory, not permission to auto-publish 12 low-value pages. Reuse existing canonical pages where appropriate.

## Agent-readable implementation
Render the meaningful answer in initial HTML. Produce an equivalent Markdown version from the same source. Link the exact product variant and current supporting policy. Use supported Product/ProductGroup schema and validate it. Publish a sitemap and canonical URLs using the merchant's production origin. Do not invent specifications, review ratings or certifications. If the brand uses Shopify, inspect its eligible agentic channels and catalog using official documentation; do not invent an MCP endpoint or promise support.

## Acceptance tests
For each customer task: retrieve the page without JavaScript; identify the answer, qualifying variant, disqualifying constraint, price/currency source and policy source; record unknowns. Compare a consistent fixture set before/after. Mark browser and checkout actions not tested unless an authorized sandbox run provides evidence.
- Answer coverage: Answered required facts / required facts, with source links. Unverified statements count as unknown.
- Variant accuracy: Correct variant selections / test tasks requiring a variant. Confirm identity against the canonical catalog.
- Unsupported claims: Count recommendations containing claims not supported by published evidence; inspect each one.
- Outcome: Track qualified product visits and consented conversion events separately from the HTML benchmark. Do not infer sales from a score.

## Rollout
Week 1 — Choose the questions: Collect actual support questions and search queries. Prioritize four repeated tasks. Save a benchmark baseline and verify the required product facts.
Week 2 — Publish the answers: Ship the first four guides, linked from products and relevant collections. Publish HTML and plain-text versions from the same content source.
Week 3 — Connect the catalog: Keep variant IDs, prices, currency, inventory and policies consistent. Review supported Shopify agentic channels in Admin; eligibility and checkout routes vary.
Week 4 — Test and improve: Rerun the same shopping tasks, record evidence and unknowns, then remeasure the site. Expand content only where a real unanswered question remains.

## Boundaries
- No guaranteed model recommendations, search rankings or revenue outcomes.
- No published product fact without an identified source. Unknown values remain unknown.
- No competitor claim without dated, comparable evidence.
- No live cart, purchase, account or payment action without explicit authorization.
- Agent discovery files are a convenience, not proof of integration with a model or commerce platform.
- Preserve human usability, original brand voice and the existing checkout.

## Primary references
- Railway agent tooling: https://docs.railway.com/agents
- Shopify product discovery: https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/products
- Shopify catalog interfaces: https://shopify.dev/docs/agents/catalog
- Google product variants: https://developers.google.com/search/docs/appearance/structured-data/product-variants

## Other categories
Home & lighting: /dtc/agent-guide.md?category=home
Fashion & accessories: /dtc/agent-guide.md?category=fashion
Beauty & personal care: /dtc/agent-guide.md?category=beauty