☑ = Remove from taxonomy🔴 Critical = Target brand as code🟡 Warning = Likely filler/junk
This is a quick preliminary check. A more thorough review is available after Step 1 & 2.
Row Selection
📋 Step 1 Complete — Dictionary Pass
📋 This is a sample run. After taxonomy review you'll choose how many rows to process for the full pipeline.
🌱 Seed Scaffold
Scanning your survey data with GPT-5.2…
I'll use the category context you provided to shape the taxonomy seeds. This takes 15–30 seconds.
Review the proposed parent dimensions below. Uncheck any you want to remove, or click a name to edit it.
These seeds guide Step 0's taxonomy builder — they are suggestions, not constraints.
📋 Configure Intake Mapping
Configure mapping before starting a run. Fields marked * are required.
⚡Smart Analysis
Template:(score: )
Column
Role
Target
Conf
🔍Transcript Question Mapping
Review the detected interviewer questions. Adjust types and brand targets as needed, then apply.
#
Question
Type
Target
⚡Structured Research Extract Detected
CSV detected — ingestion is enabled. Fill in all ingest fields below.
💬 Category Context
Tell me about the category being studied. This context helps the LLM interpret responses accurately — it won't change the prompt rules.
Category text open-end
Brand density open-end questions
Brand assignment columns
ColumnBrand labelFile value 🔒×
Brand value map (raw phrase → short name)
Map each brand's text column. The column name usually contains the brand name.
Conversational follow-up questions
Chat-style columns where an interviewer bot asked follow-up questions per brand. Brand is derived from system prompts in the conversation.
Pick the respondent response/transcript column(s). Hold Ctrl/Cmd to select multiple. If left empty, prefix matching is used (legacy).
Selected columns are treated as user response columns. The paired system/prompt columns will be inferred automatically from adjacent or similarly-named columns.
Select conversation columns above, then click Detect Brands.
Brands are extracted from System: prompts in the selected columns. Check the brands you want to include as brand density targets.
Prefix-based detection (legacy)
📋 Universal Drivers Config
Drivers file loaded — configure open-ended columns from universal drivers.
Open-ended columns from drivers
📊 Analysis Fields
Select additional fields to include in analysis. Breakout columns create separate summary sheets by segment.
Passthrough fields are added to the long file for custom analysis.
Hold Ctrl/Cmd to select multiple. Best for low-cardinality fields like Gender, Region, Age Group.
Select which values count (e.g. top-2 box: pick "4" and "5")
These values will be used as the numerator for the percentage calculation.
Adds metric rows to Summary sheets. Does NOT add columns to association sheets.
Category CSV question
Map the category-level choice/response column and offer-amount column. This is the CSV question asked about the overall category (not about a specific brand).
Offer amount must be the numeric $ value shown. If your offers are split across columns (e.g., LFHP2_1…LFHP2_7), select any one — siblings are auto-collapsed.
Dep coding — values that mean GIVE UP / CANCEL (dep=1)
Map each brand to its choice/response column and offer-amount column for CSV $ value computation.
What should I pick?
Column-targeted: each brand has its own dedicated choice and offer-amount columns (e.g., X21_csv.LinkedIn, X22_csv.Indeed). Row-targeted (monadic): all brands share ONE choice column and ONE offer-amount column. Each respondent answered the CSV question for only one brand; a separate column identifies which brand was assigned.
Choice / response column — the column that records which option the respondent chose (e.g., "Keep" vs "Give up", or 1/0 accept). Offer amount column — the numeric column that contains the $ amount shown in the question (e.g., the X2 bid value: $1, $7, $55, $400…).
Brand
Choice / response col
Offer amount col ($)
All brands share the same dep/price columns. CSV is computed per brand using only respondents assigned to that brand.
Select any one sibling if offers are split across columns — siblings are auto-collapsed.
Dep coding — values that mean GIVE UP / CANCEL (dep=1)
Brand assignment method
Column whose cell value indicates which brand was assigned (e.g., "CVS", "Walgreens")
Brand value mapping — map raw column values to canonical brand names (optional)
Assignment column
Brand
Value
Pick the unaided open-end column where respondents mention brands. We'll compute which brand was mentioned first.
The text column where respondents typed which brands come to mind (unaided awareness).
Add each brand to track. "Other/None" is always included automatically.
Next Step
How do you want to proceed?
Build Seed Scaffold — discover themes from your data and build a new taxonomy. Reuse / Upload Taxonomy — import from another project or upload taxonomy files, then skip to Step 1.
📜 Reuse / Upload Taxonomy
Select a project whose taxonomy you want to import. Step 0 (taxonomy build) will be skipped entirely.
No projects with taxonomy found.
Upload your taxonomy files to skip automatic taxonomy building. Step 0 will be skipped entirely.
1Taxonomy Structurerequired
CSV with 6 columns: theme_id, theme_label, parent_id, parent_label, code_id, code_label
JSON mapping phrases/words to code IDs for dictionary matching
No file chosen
— OR —
Uses code labels as initial phrases. You can add synonyms later.
3Lookup Indexoptional
Auto-generated if not provided. Only upload if you have a custom lookup.
No file chosen
📁 New Project
Creates a new profile and output directories. Lock and runs are per-project.
📋Confirm Run
Current Project State
🌊 Add New Wave
Project State
🌊 Add New Data
Upload new data to add to this project. Existing taxonomy will be reused (no taxonomy review).
⚠ Append new wave requires an existing master long file. Only Full reprocess is available for this project.
!
Label your new data *
Each wave needs a unique label so results can be tracked separately.
Your existing waves:
⚠ This wave label already exists. New rows will be de-duped against existing rows with the same wave + resp_id.
Choose file...
Choose file...
Generate Report from Workbook
1
Settings
2
Outline
3
Generate
Detecting...
Google Slides: a PPTX file will be generated. Import it into Google Slides via File > Import.
(optional)
Upload a previous report (PPTX, PDF, or Google Slides export) to guide the outline structure.
The LLM will mirror its organization and section flow while populating with current data.
Choose file...
Generating slide outline from workbook data...
This may take 30-60 seconds
Review the outline below. Use the chat to request revisions before generating.