Data sources

Source principles

TrendRise should use official APIs, approved APIs, licensed data providers, public data, user-authorized connectors, or user-uploaded exports. Private, gated, unauthorized, or unclear-rights data should not be used.

• A source signal is evidence, not a guarantee of demand, profit, or future performance.
• Scores and recommendations are hypotheses derived from source context.
• Customer-visible copy should not claim an official source integration unless that integration is actually approved and active.

Keep raw signals separate

Raw source results should stay separate from scored opportunities. A Reddit post, marketplace listing, trend term, ad-library result, or uploaded revenue row is source evidence; the opportunity score is an interpretation built from that evidence.

• Raw snapshots should preserve source, query, region, filters, freshness, and result context.
• Scored opportunities should explain why an item was promoted, watched, or rejected.
• User-private connector data must stay workspace-scoped; reusable public-source cache can be shared when appropriate.

Current source families

The current source strategy combines recent attention, pain language, marketplace intent, distribution fit, ad-presence clues, and later revenue evidence.

• Use now: Google Trends RSS, Reddit public/API-compatible pain search, cached user-entered research searches, and internal Trend Radar evidence.
• Optional when configured: YouTube Data API, approved marketplace discovery, and controlled experimental source adapters.
• Use later: authorized Google/Meta Ads performance data, compliant Amazon replacements, user-connected Stripe/Shopify/Gumroad-like revenue data, and uploaded exports.

Shared cache comes first

External provider calls are a platform cost. TrendRise should check local workspace data and shared source cache before making a fresh provider request. If another user recently searched the same public query, a fresh matching cache entry can be reused.

• Cache keys include source, normalized query, region or marketplace, filters, and result type.
• Research quick searches can use shorter cache windows; deeper or less volatile scans can use longer windows.
• Fresh paid or limited provider calls should be plan-gated, allowance-gated, or explicitly confirmed.

Request logging and cost control

Source pulls should be logged whether they hit cache, miss cache, are disabled, fail, or return no results. TrendRise records request status, result count, estimated provider cost, source, query, and metadata so plan margins and source reliability can be reviewed.

• The protected source-search route returns normalized records to the app and keeps raw provider payloads server-side.
• Disabled sources should return a clear warning rather than silently pretending data exists.
• Expensive sources should stay capped until usage, cost, and compliance behavior are understood.

Experimental sources are guarded

Some source adapters are experimental research signals, not durable public integrations. These tests are environment-gated, cached, capped, and disabled unless explicitly enabled.

• Marketplace source adapters are internal unless an approved platform integration or user-authorized import is configured.
• Meta ad-library evidence can show ad presence or creative patterns, not ROAS, profit, or conversion rate.
• TikTok evidence can show public distribution and audience-reaction signals, not proof of product-market fit or sales.

Generated summaries need review

Generated summaries can organize source evidence, but they can miss nuance, over-compress context, or make weak signals sound cleaner than they are. Review important claims before using them in customer-facing copy, regulated categories, paid ads, or creator outreach.

When sources are unavailable

Source availability, freshness, pricing, quotas, and provider rules can change. When a source is unavailable, TrendRise should show an honest disabled, stale, cached, or no-results state rather than filling the gap with unlabeled demo data.