Atlas ConcordatResearch · Analysis · Presentation
Menu

Atlas insight · Research architecture

All insights

Organize the research before you draft

You have found the quotation you need, but not the edition it came from. Two studies appear to disagree, though their populations differ. These problems are easier to address while gathering material than while defending a finished draft. A working research record gives you somewhere to resolve them.

01

Give the collection a working structure

A folder tells you where a document is stored. It may not tell you why you saved it or which claim it supports. Add that information to a source record when the document enters the project.

Check for duplicate files and missing attachments. Keep quotations connected to their page and edition. If a later summary differs from the original record, retain both and note the difference rather than silently replacing one with the other.

02

Begin with the use the work must serve.

Write down what the research must help someone do. An interview brief and a thesis chapter may draw on the same material but need different levels of detail. A novelist may need a record organized around scenes rather than subjects.

Use that purpose to choose the source standard and the format. It also gives you a stopping rule: which unanswered questions would prevent the document from serving its intended use?

  • What must the reader understand, decide, create, teach, or defend?
  • Which claims will carry the greatest consequence?
  • What source standard is appropriate to those claims?
  • Which uncertainties must remain visible in the final form?
03

Preserve provenance before interpretation.

Give each source a stable identifier. Record its origin and the version examined, with a retrieval link or location. Add the claim it may support and any limitation you have found. A later reader should not have to guess which version you used.

That distinction can change the analysis. A contemporary letter and a later recollection have different roles in a chronology. An updated dataset may produce a different comparison. Keep those details in the ledger even if the finished text mentions them only briefly.

  • Stable source identifier and useful title
  • Author and institution, with date and version details
  • Retrieval location and access date
  • The claim supported, its limitations and any conflicting material
04

Separate evidence from what is said about it.

An assertion in a document is not necessarily an established fact. Record what the source directly shows, then distinguish reported claims and your own inferences. Mark what remains unknown.

Also check whether apparently independent sources repeat the same original account. Several citations may amount to one source of support.

05

Build the comparison before the conclusion.

Choose table fields that expose meaningful differences. For studies, those may include population and endpoint definitions. For witness accounts, record the period described. For historical sources, check whether a shared word refers to the same institution.

Leave a comparison unresolved if the sources do not answer the same question. A blank or qualified cell is more useful than a neat but unsupported equivalence.

06

Design the final work product around use.

An executive briefing may need a short summary backed by a separate source ledger. A working team may need the full comparison table. Decide which document each reader needs, rather than giving everyone the entire research file.

Keep references in the shorter version. Reducing the length should not remove the qualifications that keep a finding accurate.

07

Check what the tools produce

Software can help locate material or transcribe a recording. It can also extract data for a comparison. Retain a link to the original so you can check the output before relying on it.

The same rule applies to AI assistance. A generated summary needs checking against the document it summarizes. Save the verification record with the work rather than treating fluent wording as evidence of accuracy.

08

Before drafting

Use these questions to check whether the research record is ready to support a draft.

  • Can every important claim be traced to its best available support?
  • Are reported assertions and your own inferences distinguished from direct evidence?
  • Are unlike sources compared on fields that make their differences visible?
  • Can a new source be added without breaking the structure?
  • Does the final format match the decision or use it must support?
  • Could another careful reader understand how the result was built?

Related standards

Research architecture changes with the field. These established frameworks illustrate the value of transparent reporting and reusable research information.

Research support

Have the material but not the working record?

Describe the collection and the question it needs to answer. Atlas can scope the work of organizing the sources and preparing the document.

Discuss a project