Seed is not truth: how NataPulse maps the S&P 500's voice on X without trusting it

A corporate X account posting about a merger is evidence. A CEO’s account posting the same thing might be evidence — or a stale handle, an impersonator, or a personal opinion dressed as disclosure. NataPulse resolves this with a curated registry covering the corporate and executive X accounts of all 503 S&P 500 constituents, governed by one rule that shapes everything downstream: seed is not truth. Being in the registry makes an account a candidate. It does not make it a source.

A registry of candidates, not an allowlist

The registry starts as a seed list: one entry per constituent for the company’s official account, plus the CEO’s account where one is publicly mapped. Every imported entry begins life in an unverified state, explicitly marked as unfit to drive alerts or publishable events until it clears verification.

This is a deliberate posture, not caution theater. X handles are fragile. Companies rebrand, spin off investor-relations accounts, and maintain regional variants. Executives change roles, and handles get recycled or squatted. A list assembled today degrades on its own — so the system treats the list itself as untrusted input and requires each entry to earn status through an explicit verification lifecycle: unverified, verified, rejected, or needs review. Re-importing an updated seed never silently promotes or demotes anything; a changed handle on an already-verified entry is surfaced as a conflict for a human, and editing a verified account’s handle deliberately resets its verification to zero.

Verification is asymmetric — and CEOs never skip the human

Verification checks a candidate handle through an approved API provider interface — never scraping — confirming the account exists, matches the expected identity, and carries platform-level verification. But the bar depends on what the account is.

A corporate account with organization-level verification can be promoted automatically at full confidence. A corporate account with only a standard verification badge scores lower. An account that exists but carries neither lands in the review queue.

A CEO account is different: it is never auto-verified, under any outcome. Even a perfect handle match with complete profile data parks the entry in needs-review, and promotion to verified is always a manual, operator-confirmed action with a recorded reason. Personal accounts concentrate exactly the failure modes a registry exists to catch — impersonation, role changes, personal commentary read as corporate signal — so a human stays in that loop by design.

What verification actually buys: a bounded boost

The payoff for clearing verification is narrow and explainable. When the ingest pipeline sees a post, it matches the author against the registry. A verified corporate or executive source can raise the resulting event’s confidence — a boost that can only ever lift confidence, never lower it — and can resolve the event to the right ticker via the registry’s entity links.

Everything else gets nothing. An unverified or under-review source earns no boost, no entity link, and a rumor flag. A rejected source is annotated as such on the evidence trail and excluded from influence entirely.

The rumor gate is the sharp edge of the design: a post from an unverified account, standing alone, is capped by the social noise ceiling and flagged as rumor. It cannot become a standalone high-confidence event and cannot trigger an alert on its own. To matter, it needs corroboration from another source family — a filing, a news report, market behavior — inside the same cluster. Volume does not substitute for identity: a thousand reposts of an unverified claim are still one unverified claim.

Feeding Social Radar and event scoring

The registry flows into the product in two ways. First, verified accounts are polled directly through a registry-coverage query layer, so official corporate voices are monitored systematically rather than caught incidentally by topic searches. Second, every social event carries its provenance: which registry source matched (if any), its verification status, and the original post reference travel with the event into Social Radar, Live Pulse, clusters, and reports. When an event shows up boosted, the reason is inspectable; when it shows up flagged as rumor, so is that.

None of this is a trading signal. NataPulse produces research evidence with explicit source identity — verification identifies an account, not the truth of every post, and no output here constitutes investment advice.

Why a registry beats ad-hoc scraping

The alternative — opportunistically harvesting whatever accounts look official — fails on exactly the axis that matters for financial evidence: provenance. A registry gives every source a stable identity, a verification state with an audit trail, and a reviewable lifecycle. It makes trust decisions explicit and reversible instead of implicit and forgotten. And it keeps access clean: the registry is built and verified exclusively through approved API interfaces under budget and rate-limit policy, never by crawling pages.

Seed is not truth. But a seed that must earn trust, entry by entry, is how 503 companies’ voices become evidence you can audit.

Sources

• NataPulse Docs — X Pipeline
• NataPulse Docs — Source Reliability
• NataPulse Docs — Social Radar
• NataPulse Docs — Confidence and Corroboration
• X Developer Platform — X API