How SIGNLD cites its sources

Every claim in a SIGNLD Decision Brief carries an evidence link back to the specific source records it was derived from. Each link names the system, the record, and the timestamp the data was read. This is what makes a brief traceable rather than a summary you have to take on faith.

By Christopher Rafter · · 8 min read · Product explained
How SIGNLD cites its sources

In this article

What does an evidence link actually point to

A Decision Brief has four parts: the finding, the evidence, a confidence score, and a recommended action. The evidence section is not a paragraph of context. It is a list of links, and each link resolves to a specific record in a specific connected system. Read the mechanics of this structure in /how-it-works.

For the wider context, see our overview of what SIGNLD is.

If a brief states that overdue invoices rose in a region, the evidence links point to the actual invoice records behind that number, not a rounded summary of them. Clicking a link takes you to the record as SIGNLD read it, including the system it came from and the field values that fed the finding. This is deliberate. A finding without a way to check the underlying record is an opinion. A finding with a link to the record is an answer you can verify.

How does SIGNLD know which record produced a figure

SIGNLD builds a Knowledge Graph from every connected system, resolving the same customer, job, invoice, or account across sources so the same entity is not counted twice under two names. Each node in that graph keeps a reference back to its origin: which system it came from, which table or object, and which record identifier. Read more on how the graph is structured in /concepts.

When a Decision Brief is generated inside a Topic, the answer engine does not paraphrase the graph from memory. It walks the specific nodes and edges that support the finding and attaches their origin references as citations. That is why a Decision Brief for one Topic can name the CRM record and the accounting record it drew from, even when the finding combines both.

Why does the timestamp matter

Every record read into the Knowledge Graph carries the time it was read from the source system. A Decision Brief inherits those timestamps and shows them alongside each piece of evidence. This tells you not just what the record said, but how current that read was when the brief was produced.

Timestamps matter because business systems change continuously. A deal can close, an invoice can get paid, a ticket can get reassigned, all within hours of a brief being generated. Showing the read time lets a reader judge whether the evidence is still fresh or whether it is worth re-running the question.

How does traceability connect to confidence scoring

Confidence scoring is not a separate, opaque number bolted onto the brief. It is built in part from how much of the underlying data is actually covered by traceable, recent evidence. A finding backed by many current records across systems that agree with each other scores differently than one built from a thin or stale set of records.

Source coverage and freshness are two direct inputs into that score. If a Topic only has a partial read of a connected system, or if the most relevant records were read a while ago relative to how fast that data changes, the confidence score reflects that gap rather than hiding it. For the full mechanism behind the number itself, see How SIGNLD confidence scoring works.

What happens when a source record changes after a brief is issued

The brief itself is a snapshot. It shows what the evidence said at the timestamps recorded when it was generated. If the underlying record changes afterward, for example a deal moves stage or a payment posts, the original brief does not silently update itself.

Asking the same question again inside the Topic produces a new brief with new timestamps, reading current records rather than the ones from before. This keeps the historical brief intact as a record of what was known at that point, while giving teams a way to re-check the answer whenever they need current evidence.

Does this work the same way for spreadsheets

Yes. Spreadsheets are a first-class source in SIGNLD, not a lesser one. A spreadsheet uploaded or connected to a Topic is read the same way any other system is read, and the rows behind a finding are cited the same way a CRM or accounting record would be, down to the sheet, the row, and the read timestamp.

This matters for teams whose most current numbers live in a working spreadsheet rather than a governed system. The evidence trail treats that spreadsheet row with the same specificity as a database record from an 800-plus-system connector.

Can a team member check the citations without technical help

Yes. The evidence links inside a Decision Brief are meant to be opened by the person reading the brief, not routed to an analyst. Because connections are read-only, opening a citation carries no risk of altering the source record, so anyone with access to the Topic can click through and verify a figure directly.

This is part of why SIGNLD can return a first answer within minutes of connecting a system. The traceability is generated as part of building the brief, not added afterward as a manual audit step.

Key takeaways

  • A Decision Brief has four parts: the finding, the evidence, a confidence score, and a recommended action.
  • Every record read into the Knowledge Graph carries the time it was read from the source system.
  • Confidence scoring is not a separate, opaque number bolted onto the brief.
  • If they ask at different times, the records read and their timestamps can differ because source systems keep changing.
  • See the evidence links behind a real Decision Brief on your own systems.

FAQ

Does every sentence in a Decision Brief have its own citation

The core claims, meaning the finding and any figures it depends on, carry evidence links. Framing language around the finding does not need its own citation because it makes no factual claim on its own.

Can I see which system a number came from without opening the record

Yes. The evidence section names the source system next to each link, so you can see the origin before deciding whether to open the underlying record.

Does a low confidence score mean the citation is wrong

No. A lower score usually reflects thinner source coverage or older timestamps, not an incorrect citation. The record cited is still accurate, the score is describing how complete or current the evidence behind it is.

Can two people get different evidence for the same question

If they ask at different times, the records read and their timestamps can differ because source systems keep changing. Asking the same question again pulls current records rather than reusing an earlier snapshot.

Do evidence links work for connectors added after setup

Yes. Any system connected to a Topic, whether added on day one or later, is read into the Knowledge Graph with the same origin references, so its records can be cited the same way.

Related reading in this series: Is SIGNLD secure and SIGNLD week one: what to expect.

Try SIGNLD free

See the evidence links behind a real Decision Brief on your own systems. Try SIGNLD free or read /how-it-works for the setup steps.