SITEBORNE SIGNAL MODULE

FeedbackAtlas — customer-language intelligence

Organizes approved surveys, questions, and feedback into recurring themes with source links and minimum sample rules.

Try the demonstration Add to a project brief

BUSINESS OUTCOME

What this capability changes.

Reveal repeated questions
Identify content gaps
Preserve source evidence

Best for

Service businesses, Product teams, Content teams

AI: optionalStatus: active

From $2,000

Care from $200/month

Illustrative allowance usageAvailable

PLAIN LANGUAGE, TECHNICAL PROOF

What FeedbackAtlas does.

Groups repeated customer comments into themes while keeping every theme connected to the original words.

Why it is valuable

Teams can see what customers repeatedly struggle with without treating a tiny sample as a market truth.

A real-world use

Support comments about setup time, account ownership, and reporting become three traceable themes with sample counts and source excerpts.

FOR TECHNICAL READERS

How the control model works

Source-linked clustering is combined with minimum-sample rules. Exact tags remain the fallback, and trend language is suppressed when evidence is insufficient.

Failure path: Exact filtering, manual tagging, and source-linked review queue.

INTERACTIVE WALKTHROUGH

Use FeedbackAtlas before discussing it.

Edit the sample feedback—one comment per line—and see which themes have enough evidence to surface.

The interaction uses your input and identifies whether it runs locally, checks this live site, or previews a production control. It does not disguise seeded output as intelligence.

Customer comments — one per line

JavaScript-free explanation

This demonstration enhances inputs in the browser. The module page documents its evidence restrictions, fallback, allowance, and upgrade path without requiring JavaScript.

System view

Required

cloudflare-d1, cloudflare-vectorize

Optional

cloudflare-workers-ai, cloudflare-ai-gateway

Fallback

Exact filtering, manual tagging, and source-linked review queue.

Provider dependencies

cloudflare-d1, cloudflare-vectorize, cloudflare-workers-ai, cloudflare-ai-gateway

Last verified

2026-08-02

Provider limits, terms, model availability, and pricing may change. Production accounts remain client-owned.

Data classification

Approved feedback
De-identified support records

LAUNCH ALLOWANCE

Semantic retrieval

Prototype and low-volume vector storage and query capacity.

Current limit: 30M queried vector dimensions/month and 5M stored vector dimensions.
Expected light use: A small public-content index supporting prototype or low-volume semantic retrieval.
What consumes it: Stored vector dimensions and queried vector dimensions across the account.
Fallback: Keyword search, D1 metadata filters, and curated navigation.
Paid successor: Vectorize paid usage or approved vector adapter
Commercial use: Commercial use permitted.
Data warning: Embeddings may reveal source semantics; classify content before indexing.
Last verified: 2026-08-02

Upgrade triggers

sample size — below configured minimum. Suppress trend claims.
vector usage — 75%. Review index and paid successor.

Conventional fallback: Exact filtering, manual tagging, and source-linked review queue.

View implementation proof

Decision rationale

The implementation favors semantic HTML, static rendering, typed content, and isolated on-demand routes. Optional AI never controls pricing, eligibility, or emergency routing.

Controls

WCAG 2.2 AA target
Provider-neutral adapters
Usage caps and kill switches
Client-owned production accounts
Conventional fallbacks

Verified evidence

The browser demonstration runs from a clean clone without a paid credential.
Kill switch and allowance states are typed configuration.
Primary navigation and contact paths remain independent of this module.

Known limitation

External provider allowances and production behavior require owner accounts and deployment-time verification.