Frequently asked questions.

Sigblaze is a market intelligence product. It researches your market, competitors and customers, checks them again on a schedule, raises a signal when something meaningful changes, and turns reviewed evidence into actions for marketing, sales, product and strategy. Sources, claims, contradictions, insights and signals are stored in the workspace, so the next question starts from what is already known.

Boutique B2B agencies running research for several clients, fractional CMOs and GTM consultants, B2B SaaS teams, and founders or product marketers who research competitors and markets on a recurring basis.

General AI tools are good for one-off questions and exploration. Sigblaze is built for recurring team research: it stores workspace context, keeps sources and claims, shows contradictions, preserves lineage and lets a team reuse approved intelligence in later research and actions.

You state a research goal, and the product turns it into a bounded plan of tasks and queries. Each task inherits the workspace's company, product, ICP, market and competitor context, so one vague prompt does not have to carry everything.

Yes. Ask answers from the workspace's stored context, claims, insights, signals and sources, including what you approved or rejected, and it can answer questions such as “what changed this week?”. When the stored evidence is not enough, it says so instead of presenting general model knowledge as workspace evidence.

Yes. Each client or project can have its own workspace with separate context, sources, claims, insights and actions. Evidence is not merged across workspaces.

In the private alpha: positioning, campaign direction, content, sales battlecards, outreach, product opportunities and reports. Each action keeps a trace back to the insight, claims and sources behind it, and none is sent or published automatically.

Providers are configured on the server, and the specific model and search providers are not listed publicly yet. Provider keys are never exposed to the browser.

Not yet. Research over uploaded documents is a later capability. Today the product retrieves public web sources.

Not yet. White-labelled, client-ready reports are a later capability. Approved intelligence can be exported as Markdown and JSON today.

A claim is one factual statement extracted from a source. It stays linked to the evidence that supports, contradicts or adds context to it.

The disagreement is kept. A claim with contradicting evidence is shown as contested, and the system does not silently pick a preferred fact.

An insight is an interpretation written from stored claims and evidence. It keeps the claim relationships, confidence and freshness that produced it, and a person reviews it before it is reused.

Public web pages and search results, retrieved through a provider adapter on the server. Each stored source keeps its type, address and retrieval date. Pages that need a login are not read, and dedicated social, app-store and video sources are later.

Each claim keeps a link to the source excerpt it came from, and further sources are attached as supporting, contradicting or contextual evidence. You can open any claim and see which sources sit behind it.

Yes. Insights and actions are reviewed by a person before they are reused or exported, and what you approved or rejected is available to Ask.

The product says so. Ask reports that the workspace does not hold enough evidence and can list what is missing, and an insight with thin support carries a low confidence label.

Yes, in the private alpha. Monitor re-checks the public pages and search queries on a watchlist daily, weekly or on demand, compares each check with the last stored snapshot and raises a Signal when something meaningful changed. It does not read pages that need a login, and email or chat alerts are a later capability.

A Signal is a detected development, such as a price change, a removed free plan, a launch, a headline rewrite or a rise in customer conversation. It carries what happened, the before and after evidence, why it matters, a confidence label (low, medium, high), an importance rating (low to critical) and recommended actions.

By comparing two stored snapshots of the same page. The comparison is deterministic, so a price that moved or a line that appeared is a fact about the two snapshots. A model may reword a signal for your context but cannot add facts, and its output is rejected if it cites something that does not exist.

A Brief is a digest of a period's signals: what happened, why you should care and what to do. It can be daily, weekly, monthly or on demand, ranks changes by importance and confidence, and links every item to its evidence.

No. Research is deliberate: you ask a question and it is investigated. Monitor keeps watching and reports changes. Both write to the same intelligence graph, so Ask can answer from either.

Pricing is not final. The private alpha is for design partners, and no plan names, prices or limits have been set.

Research, Signals, Monitor, Brief, Ask, evidence-linked actions, the intelligence graph and Markdown and JSON export are in the private alpha. Email and chat alerts, uploaded-document retrieval, CSV and PDF export, white-labelled reports, integrations and billing are later.

Approved intelligence can be exported as Markdown and JSON. CSV and PDF export are later capabilities.

Workspaces belong to an organisation, and database access rules limit reads and writes to that organisation's members. No compliance certifications, SLAs or data-residency options are claimed yet; the security page lists what is and is not covered.

No. Marketing analytics on this site never receive research goals, sources, claims, insights, actions or Ask content, and the access-request form does not ask for proprietary research material.