MaveraMavera Docs
Operator's Guide
Checking…
Use or share this pageSend the page itself, print it, or open its complete context in your AI workspace.

Workshop: plan a customer-language study

Prepare a defensible customer-language analysis without pretending the current Ingest surface builds grounded audiences.

2 min readDocs build Updated Jul 26, 2026
Interactive product walkthrough

Mine customer voice

Upload transcripts, route them, inspect extraction, analyze, and save themes.

Simulation onlyNothing is saved or charged
1
Step 1 of 6Open Files

Use the sidebar to open the workspace file library.

Demo workspace / Mavera workspace
Local simulation
Mavera workspace

Good afternoon

Choose a surface from the sidebar to begin.

Demo workspace0 actions completed
WorkspaceDemo companySafe local simulation
Current surfacefiles InFollow the highlighted action
Recent workspace activityStatus
Audience comparisonReady to review
Customer interview evidenceGrounded
Campaign pre-flightDraft
  1. 1
    Open FilesUse the sidebar to open the workspace file library.
  2. 2
    Upload a sample fileDrop in the included sample interview transcript.
  3. 3
    Choose what it becomesRoute the file to Context, an audience ingest, or analysis.
  4. 4
    Inspect extractionOpen extracted pages and verify what was read.
  5. 5
    Analyze itAsk the sample question against the uploaded evidence.
  6. 6
    Save the resultSave the grounded output to the project.

Use this workshop to design the analysis before uploading anything. The current Ingest page accepts completed research studies and produces source-quoted dossiers. It does not yet convert raw transcripts, tickets, or reviews into a reusable grounded audience.

Current product boundary

Do not upload confidential raw customer text to Ingest. Large-file storage is not private, and dossier/blob deletion controls are not yet available. See Ingest a completed study.

If you already have a completed study

Extract the dossier

Upload the completed report or study bundle in Ingest. Review its purpose, methodology, audiences, stated claims, inferred claims, and source quotes.

Separate observation from interpretation

Treat verified source quotes and clearly reported study findings as human-study evidence. Treat inferred claims as model readings. Do not merge either with new synthetic responses.

Identify the unresolved language question

Ask:

Prompt1 field left
From this completed study, list the exact customer phrases tied to [decision]. Keep verbatim quotes separate from your interpretation. Which language question remains unanswered?
Unfilled variables remain in brackets.

Propose a separate modeled comparison

If a synthetic comparison would help, define the audience and variants separately. Label every response as synthetic and use the result to choose what to validate next, not to estimate market prevalence.

If you only have raw transcripts, tickets, or reviews

Do not use the current Ingest page for this workflow. Prepare the source set offline or with an approved research system until Mavera provides private storage, retention, deletion, and a supported raw-text grounding pipeline.

A useful preparation brief includes:

  • the decision the language should inform
  • source type and collection dates
  • who was included and excluded
  • consent and usage rights
  • stable source identifiers
  • sensitive fields that must be removed
  • the coding framework you want reviewed

What a later grounded workflow must preserve

A defensible implementation should retain source identifiers, distinguish quotes from summaries, expose what was excluded, and keep modeled audience behavior separate from observed customer language.

Workshop: pricing read

45 minutes to a price corridor you can defend — resistance points, framing, and the objection map.

Workshop: campaign pre-flight

The 60-minute check a campaign should pass before money gets spent.

On this page

If you already have a completed studyExtract the dossierSeparate observation from interpretationIdentify the unresolved language questionPropose a separate modeled comparisonIf you only have raw transcripts, tickets, or reviewsWhat a later grounded workflow must preserve