Methodology

What sigcue collects

sigcue reads public discussions through official APIs: GitHub issues and comments today, and other permitted sources as they are enabled. It does not scrape pages, use session cookies, or read private data.

Each record keeps its original source, community, identifier, and timestamps so any claim on the site can be checked against the source.

How a record becomes evidence

Every record is read for the same structured facts: who is affected, the workflow involved, what failed, any workaround, and the consequence. Records that cannot be described that precisely are not used.

Summaries on this site are written from those facts. The original excerpt and a link to the source stay attached to the evidence.

What counts as an independent report

Records are matched by source identifier, thread, canonical link, and near-duplicate text. Several comments inside one discussion are one report, not many.

Independent reports are separate people or threads describing the same failure.

How signals are grouped

Reports are grouped when they share the affected people, the workflow, the failure, and the desired outcome. A shared tool name or a broad category is never enough on its own.

Related concrete signals can also roll up into a broader problem, which stays linked to each specific signal underneath it.

What the evidence labels mean

Insufficient: fewer than two independent reports. The problem may be described well without being shown to recur.

Early: at least two independent reports, with coverage still thin.

Developing: at least three independent reports across two communities or two weeks.

Strong: at least five independent reports across two communities and thirty days, with consistent failure descriptions and counter-evidence reviewed.

Evidence that weakens a signal

Reports that resolve, contradict, or qualify a problem are recorded and shown on the signal page rather than dropped. A reported workaround is not treated as a fix.

What sigcue does not claim

A public complaint does not prove market size, willingness to pay, or that a business would work. Nothing here is a recommendation to build a particular product.

Public discussions also overrepresent some groups and omit purchasing context, offline work, and people who never post. Use this to ask better questions, not to skip talking to people.

Why a person approves publication

Nothing is published automatically. A reviewer reads the signal, the evidence, and the limitations, and decides whether it can be shown. Evidence is never attached to a signal without that review.

Demo data

Some markets contain seeded examples used to demonstrate the interface. They are labeled “Demo data” and are never counted in live metrics or shown as published evidence.

Corrections and removal

When a source record is deleted or becomes unavailable, it is marked and excluded from future metrics, and the signal is recalculated. Correction and removal requests are logged for review through the deployment workspace.