Back to docs
Study Design

Uploading Context Documents

How to add background files that help Koji's research assistant design a sharper brief and better questions.

Context documents give Koji's research assistant a richer understanding of your research landscape. They're background materials that help it design a more informed study: a sharper brief, better questions and a clearer target participant. Think of them as briefing materials you'd hand to a human research assistant before they start drafting your discussion guide. They stay with you and the research assistant. The AI interviewer never reads them during a participant's interview; it works from the study you publish.

Why Context Documents Matter

Without context documents, Koji works with what you tell it in the chat conversation. That's often enough for straightforward studies. But when your research lives within a specific domain — a particular product, a niche market, a complex organizational structure — extra context can make a meaningful difference to how the study is designed.

Here's what context documents help with:

Better Question Design

When the research assistant has access to your product documentation, previous research, or market analysis, it can craft interview questions that reference real features, actual pain points, and genuine use cases rather than generic placeholders. This applies to both open-ended questions and structured question types — context helps it suggest more relevant scale endpoints, choice options, and ranking items.

Without context: "How do you use the product in your daily work?" With product documentation as context: "How does the workflow automation feature fit into your team's daily operations?"

A Sharper Brief

The research assistant uses your files when it drafts the problem statement, the hypothesis, the decision the study should inform and the topics to cover. Those parts of the research brief are what the AI interviewer works from, so a well-informed brief is how your context reaches every interview.

Accurate Terminology

Every domain has its own vocabulary. Context documents help the research assistant use the right terms in the questions and the brief — your product's feature names, your industry's jargon, your organization's terminology. Because the interviewer asks the questions as written in the study, the right words carry through to the conversation.

Better Structured Question Design

When the research assistant has access to your existing data — previous NPS scores, known feature lists, competitive landscape — it can design more targeted structured questions. For example, context about your product's feature set helps it suggest accurate options for multiple-choice or ranking questions, rather than generic placeholders.

Supported File Formats

Koji accepts the following file types as context documents:

FormatExtensionBest For
PDF.pdfResearch reports, whitepapers, product specs
Plain Text.txtQuick notes, raw content, transcripts
JSON.jsonStructured data, API documentation, survey results
Markdown.mdTechnical documentation, README files, knowledge bases

Each file can be up to 25 MB. Word documents aren't accepted directly: save them as PDF first.

You can upload up to 5 files per study. This limit keeps the research assistant focused — too many documents can dilute the signal. Choose the most relevant materials rather than uploading everything you have.

What to Upload

Not all documents are equally useful. Here are the types that tend to have the biggest impact:

High-Value Context Documents

  • Previous research findings — summaries or reports from past studies on related topics. These help the research assistant build on what you already know rather than retreading old ground.
  • Product documentation — feature descriptions, user guides, or product specs. These help it understand what participants are using and what language to use when asking about it.
  • Customer feedback summaries — compiled feedback from support tickets, NPS surveys, or review platforms. These surface recurring themes worth turning into questions or topics.
  • Competitive analysis — information about alternatives in your market. This helps it understand the landscape participants are navigating.
  • Internal strategy documents — product roadmaps, quarterly plans, or team goals. These help align interview questions with strategic priorities, without exposing the documents to participants.

Less Useful Documents

  • Raw data dumps — large spreadsheets or databases without context. The research assistant can't easily extract relevant information from unstructured raw data.
  • Entire codebases — technical implementation details rarely help with qualitative research design.
  • Legal or compliance documents — unless your study is specifically about legal compliance, these add noise without signal.
  • Marketing materials — these tend to describe ideal states rather than reality, which can bias the questions toward your own story.

How to Upload

Adding context documents to your study is straightforward:

  1. Open your study in the design workspace
  2. Click Attach next to the chat with the research assistant, or drag and drop your files onto it
  3. Your files are processed and read by the research assistant
  4. Open the files list at any time to view or delete a file

You can upload documents at any point during the design process — before you start chatting with the research assistant, during the conversation, or after the research brief is drafted.

Tip: Uploading documents before your first message tends to produce the best results, since the research assistant can factor in the context from the very beginning.

How Context Documents Are Used

Context documents are design material. They shape the study, and the study shapes the interview.

During Study Design

The research assistant draws on your context documents when suggesting methodologies, drafting interview questions, defining your target participant, and recommending structured question types. You'll notice more specific, grounded suggestions compared to studies without context.

During Interviews

The AI interviewer doesn't read your context documents. It works only from the study itself: the problem, the hypothesis, the questions, the topics and the guardrails you can see and edit. That's deliberate. Roadmaps, hypotheses and earlier participants' data are the researcher's material, and an interviewer that has read them is more likely to ask leading questions or hint at things participants shouldn't hear.

So anything the interviewer must know goes into the study:

  • A concept, requirement or feature you want reactions to: write it into the question text, or attach an image to the question
  • Terms the participant will use: put them in the brief or the question wording
  • Topics to dig into: add them to the brief's topics, so the interviewer probes on them
  • Things to stay away from: add them as guardrails

Review the brief before you publish with that in mind: if a detail matters for the interview and it only lives in a file, move it into the study.

Best Practices

Be Selective

With a limit of five files, prioritize quality over quantity. One well-written research summary is more useful than five loosely related documents.

Keep Documents Focused

If you have a 50-page report but only the executive summary and findings sections are relevant, consider extracting those sections into a separate document before uploading.

Update When Needed

If your context changes — for example, a new product feature launches between study design and interviewing — upload the new document and ask the research assistant to update the brief or the questions. The interview follows the study, so the change reaches participants only once it's in the brief or the questions.

Check What Made It Into the Brief

After the research assistant drafts the study, read the brief and the questions as if you'd never seen your files. Whatever is missing there, the interviewer won't know.

Context Documents and Study Quality

Think of context documents as a multiplier for study design. A well-designed study without context produces good results. The same study designed with relevant context produces sharper questions and a better-aimed brief. The context doesn't replace good research question writing or thoughtful work with the AI Consultant — it enhances both.

If you're new to Koji, it's perfectly fine to start without context documents and add them as you become more comfortable with the platform. Your first study doesn't need to be perfect — it needs to teach you something useful.

Related Articles