# Golden Analytics — Full Reference Last updated: July 4, 2026 Most BI tools force a choice: deep analytical capabilities, user-friendly interfaces for non-technical teams, or high-quality visual design. Golden Analytics is built for data teams who are done making that tradeoff. It combines a shelf-based visualization builder, a Canva-style freeform dashboard canvas, and AI that is woven into every step of the workflow — from writing the first query to publishing a polished data story — without requiring SQL, a designer, or a separate AI add-on. ## Who It's For - **Data analysts**: explore datasets, build charts, and answer ad-hoc questions without writing SQL; AI suggests chart types, follow-up questions, and key findings automatically - **BI engineers**: generate dashboards from a text prompt, then refine with a drag-and-drop builder; save transformation recipes and data dictionaries to standardize work across the team - **Business users and executives**: ask questions in plain English and get visualization answers instantly; consume published dashboards and stories without needing to understand the underlying tool - **Data teams**: collaborate in shared Spaces, build a reusable asset library (datasets, dashboards, stories, recipes), and publish polished reports at the cadence the business needs ## Common Use Cases - Connect a CSV, Excel file, or data warehouse and immediately ask questions in plain English without writing any SQL - Generate a complete multi-chart dashboard from a single text prompt ("show me a sales performance dashboard by region and quarter") - Build a board-ready data story (slide-based presentation) from a dataset with AI-suggested narrative structure - Let the Analyst Agent proactively surface anomalies, key drivers, and patterns — without specifying what to look for - Set up a Monitoring Agent to alert via email or Slack when a metric crosses a threshold or data quality degrades on a schedule - Clean, deduplicate, and enrich a messy dataset using the no-code Data Agent; preview row-level changes before applying - Save a sequence of data transformations as a Recipe and replay it automatically on new datasets of the same shape - Share an interactive dashboard with teammates via a shared Space with access control ## How It Differs from Traditional BI Tools - **AI-native, not AI-added**: AI drives the entire workflow from initial query to published story. There is no chatbot bolted onto a legacy drag-and-drop tool — the shelf builder, canvas, chat interface, and data prep pipeline are all designed around AI from the start. - **Design quality as a first-class feature**: dashboards and stories are visually polished by default, with a freeform canvas, rich component types, and infographic-style elements. Good-looking output does not require a designer or post-export cleanup in another tool. - **No SQL required; SQL never blocked**: natural language queries cover the full analysis workflow for non-technical users. Power users can always drop into SQL when needed. - **In-browser data processing**: file uploads (CSV, Excel, Parquet, JSON) are processed entirely in DuckDB in the browser. No data leaves the user's machine until they explicitly save or connect to a warehouse. - **Unified surface**: chat, visualization builder, dashboard canvas, data prep, and presentation builder are a single integrated product — not separate tools connected by exports and copy-paste. --- ## Product Overview Golden Analytics is a browser-based BI platform with the following top-level capabilities: - **Natural language chat** for querying data without SQL, with threaded follow-up conversations - **Shelf-based drag-and-drop visualization builder** using rows, columns, color, size, and trellis shelves for flexible, grid-free data encoding - **AI-generated dashboards** from a text prompt, with style presets and layout modes - **AI-generated data stories** — slide-based presentations backed by live data - **Analyst Agent** — proactive insight discovery across a dataset, structured into confidence tiers - **Storyteller Agent** — AI-assisted slide narrative builder with story structure suggestions - **Data Agent** — AI-powered data preparation, cleaning, enrichment, and transformation with row-level previews - **Monitoring Agent** — scheduled, automated monitoring of dashboards, datasets, and metrics with multi-channel alerting - **Data Dictionary** — per-dataset field documentation that improves AI accuracy - **No-code data transformations** tracked in a non-destructive history, saveable as Recipes - **Freeform dashboard canvas** with independently positioned, sized, and styled components - **Spaces** for team collaboration and shared asset libraries - **Full undo/redo history** on the canvas --- ## Data Sources ### File Upload Supported formats: CSV, TSV, TXT, Excel (.xls, .xlsx), JSON, Parquet. Files are processed entirely in-browser using DuckDB WASM. Queries against uploaded files run locally — no data is sent to a server unless the user explicitly saves the session. File-based datasets support the full transformation pipeline, recipe system, and visualization builder. ### Live Database Connections Queries run against the user's data warehouse in real time. Each query is generated by the AI or the visualization builder and executed against the connected warehouse — no data is copied or cached. Supported connectors: - **Snowflake**: connects via account identifier, username/password or key-pair auth; supports multiple databases and schemas - **Databricks**: connects via workspace URL and personal access token; supports Unity Catalog and legacy Hive metastore - **Google BigQuery**: connects via service account JSON; supports project/dataset/table hierarchy - **Amazon Redshift**: connects via JDBC endpoint; supports IAM auth and password auth - **Microsoft Fabric**: connects via workspace credentials; supports lakehouse and data warehouse queries - **ClickHouse**: connects via host and credentials; supports standard ClickHouse query syntax - **Amazon S3**: queries Parquet or CSV files directly via S3 path; supports IAM role or access key auth - **Salesforce**: connects via OAuth; queries Salesforce objects as tabular data ### Google Sheets Connect by sharing a Google Sheets link. Data is fetched live from the sheet and treated as an in-memory dataset. Supports the same transformation and visualization capabilities as file uploads. --- ## Chart Types 38 chart types, organized by category: ### Bar and Column - **bar**: vertical bar chart, single or grouped series - **stackedBar**: vertical stacked bars, absolute values - **stackedBar100**: vertical stacked bars, normalized to 100% - **divergingBar**: bars that extend in both positive and negative directions from a center axis - **horizontalBar**: horizontal bar chart, single or grouped series - **horizontalStackedBar**: horizontal stacked bars, absolute values - **horizontalStackedBar100**: horizontal stacked bars, normalized to 100% - **lollipop**: bar chart variant with a dot at the end of a thin line - **rangeBar**: bars with explicit start and end values (e.g. schedule or duration data) - **waterfall**: sequential bar chart showing cumulative effect of positive and negative values ### Line and Area - **line**: standard time series or continuous line chart; supports multi-series - **area**: filled area under a line; supports multi-series - **stackedArea**: stacked filled areas, absolute values - **stackedArea100**: stacked filled areas, normalized to 100% ### Part-to-Whole - **pie**: circular chart showing proportions of a whole - **donut**: pie chart with a hollow center; supports a center label - **treemap**: nested rectangles sized by measure value - **funnel**: sequential stages showing drop-off or conversion - **sankey**: flow diagram showing volume moving between nodes ### Distribution - **scatter**: x/y plot of two measures; supports color and size encoding - **boxPlot**: distribution summary showing median, quartiles, and outliers - **histogram**: frequency distribution of a single measure, auto-binned or custom bins - **wordCloud**: text visualization sized by frequency or measure value ### Maps - **map / shapeMap**: shape-based geographic map; regions filled by measure value - **filledMap**: choropleth-style map using geographic region names - **densityMap**: point-density heatmap on a geographic canvas - **choropleth**: explicit choropleth component in the dashboard canvas ### Tabular - **table**: row-column data table with sorting, pagination, and conditional formatting - **pivotTable**: cross-tabulation with row and column dimensions and measure aggregation - **bigNumber / kpi**: single large metric display with optional comparison and sparkline ### Heatmaps - **heatmap**: 2D grid colored by measure intensity - **matrixHeatmap**: heatmap with explicit row and column dimension labeling ### Radial The **radial** chart type supports five visual variants, selectable in format settings: - **radar**: polygon chart with one axis per dimension; useful for comparing profiles - **radialBar**: arc-based bar chart arranged in a circle - **stellar**: spike chart radiating from a center point - **radialLollipop**: lollipop variant on a radial axis - **sunburst**: hierarchical concentric ring chart (like a radial treemap) ### Other - **combo**: mixed chart combining bar and line series on shared axes (dual-axis supported) - **retention**: cohort retention grid showing activity rates over time periods - **gantt**: horizontal bar chart for project scheduling with start/end date encoding --- ## Data Preparation ### No-Code Transformations All transformations are non-destructive — they are tracked in a history log and can be undone at any time. The original dataset is never modified. **String** - Smart Split: split a column by delimiter or regex pattern into one or more new columns - Uppercase / Lowercase / Title Case: normalize text casing - Trim Whitespace: strip leading and trailing spaces - Find & Replace: literal string substitution across a column - Regex Replace: pattern-based substitution using regular expressions **Date** - Extract Year / Extract Month / Extract Day: derive a numeric column from a date field - Parse Date: convert a text column to a recognized date type using format inference or an explicit pattern **Number** - Round: round to a specified number of decimal places - Absolute Value: replace negative values with their absolute value - Calculate Percentage: express a measure as a percentage of a total or another column - Normalize: scale values to a 0–1 range - Clip Outliers: cap values above or below a specified percentile threshold **Type Conversion** - Text → Number - Number → Text - Text → Date **Logical** - Fill Null Values: replace nulls with a literal value, column mean, or forward/backward fill - If/Then/Else: derive a new column using conditional logic **Structural** - Deduplicate Rows: remove exact or partial duplicates - Rename Column: change a column's display name without altering underlying data **AI-Powered** - AI Extract: extract structured entities or patterns from unstructured text (e.g. extract city from an address string) - Enrich: augment the dataset with additional columns inferred from existing data using AI - Standardize: normalize inconsistent values in a column to a canonical form (e.g. normalize country name variants) ### Recipes Any sequence of transformations can be saved as a Recipe. Recipes are replayable against new datasets that share the same column schema, enabling teams to standardize data prep steps across multiple ingestion runs. ### Data Agent The Data Agent is an AI assistant focused exclusively on data preparation. It can be used in two modes: **Chat mode**: The user describes what they want in plain English ("remove duplicate rows", "split the full name column into first and last", "flag rows where revenue is negative"). The agent applies the transformation directly and shows a row-level preview. **Cleaning wizard**: Scans the dataset automatically and surfaces a prioritized list of suggested fixes, grouped by category: - Splits & Extraction (e.g. compound columns that should be separated) - Text Analytics (e.g. columns suitable for sentiment analysis) - Enrichments (e.g. a city column that could be enriched with country or lat/lon) - Nulls, Outliers & Duplicates - Type Corrections (e.g. a numeric column stored as text) Each suggestion shows a confidence level (high/medium/low) and a row-level before/after preview. The user can accept, reject, or modify each suggestion individually. Additional Data Agent capabilities: - **Improve field names**: renames columns to be clearer and more consistent - **Suggest new fields**: proposes derived columns based on patterns in existing data (e.g. "profit margin" from revenue and cost columns) - **Text analytics**: runs sentiment analysis, topic extraction, or NLP classification on text columns ### Data Dictionary A per-dataset field reference that each dataset carries with it. Each field entry includes: - **Name**: the column name - **Type**: dimension, measure, date, or geographic - **Description**: a human-readable explanation of what the field represents Descriptions can be authored manually, bulk-imported from a CSV or spreadsheet, or auto-imported from a URL pointing to existing documentation. The Data Dictionary is injected into AI prompts automatically, improving the accuracy of natural language queries and AI-generated charts. --- ## AI Capabilities ### Chat Interface Users ask questions in plain English. The AI interprets the question, queries the connected dataset (in-browser DuckDB or a warehouse), and returns a visualization or a text summary. After each response: - **Suggested follow-ups** appear automatically: filter by a dimension, show a trend over time, compare segments, break down by a category, or drill deeper into a sub-population - **Threaded conversations**: users can reply to any AI message to continue exploring that specific result in a sub-thread, keeping related analyses grouped - **Visualization suggestions**: the AI recommends a chart type and shelf configuration for each answer; users can accept the suggestion or switch chart types manually ### Analyst Agent The Analyst Agent runs a proactive, unsupervised analysis of the full dataset and returns a structured set of findings organized into three confidence tiers: - **Headlines** (high confidence): the most significant findings — key drivers, standout segments, unexpected values, and clear trends — each accompanied by a suggested visualization - **Supporting** (medium confidence): context and distribution patterns that provide background for the headlines - **Investigate** (low confidence): data quality warnings, ambiguous signals, or patterns worth human review Additional Analyst Agent behaviors: - Insights can be pinned to the workspace for persistent reference while building a dashboard or story - Individual insight explanations can be copied or shared with teammates - Each insight includes recommended next-step analyses and questions to explore - A "Deeper Insights" dialog allows the user to expand any finding into a full sub-analysis ### Storyteller Agent The Storyteller Agent converts a dataset into a slide-based narrative presentation (called a Story). **Story structure suggestions**: the agent proposes narrative frameworks appropriate for the data: - Executive Summary (high-level KPIs and key takeaways) - Rise and Fall (a story arc tracking a metric over time) - AI-generated custom frameworks based on the dataset's content Each structure includes a recommended slide count and a logical sequence of chart types. **Story building**: the agent generates and populates slides with visualizations drawn from the dataset. Users can: - Chat with the agent to refine individual slides ("make this chart a bar chart", "add context about the Q3 dip") - Update the visual theme across all slides at once - Add, remove, or reorder slides - Add speaker notes per slide **Presentation mode**: full-screen presentation view for live delivery. Slides can be exported to PowerPoint, PDF, or Google Slides. ### Dashboard Generation Users describe a desired dashboard in a text prompt. The AI generates a complete multi-chart dashboard with: - Appropriate chart types for each metric - A logical layout with grouped sections - Titles, labels, and a consistent color theme **Layout modes**: structured presets (e.g. executive overview, operational monitoring) or freeform mode where components are placed and the user refines positioning manually. **Style presets**: control the visual theme (color palette, font choices, background) of the generated output. ### Monitoring Agent The Monitoring Agent runs on a schedule, analyzes a target (a dashboard, dataset, or specific metric), and surfaces findings with configurable alerting. **Targets**: - Dashboard: monitors the underlying data powering a published dashboard - Dataset: monitors a connected data source for changes, anomalies, or quality issues - Metric: monitors a single KPI or calculated measure **Schedule options**: - Frequency: daily, weekly, or monthly - Time of day: configurable, with timezone support - Day of week / day of month: for weekly and monthly frequencies - Weekends excluded: option to skip Saturday and Sunday runs **Finding types**: - Anomaly: values that deviate significantly from historical patterns - Trend: directional changes in a metric over time - Threshold breach: values that cross a user-defined or AI-suggested threshold - Data quality: missing values, unexpected nulls, format inconsistencies, or distribution shifts **Finding severity**: critical, high, medium, low, info **Notification channels**: - In-app notification - Email (to the agent owner) - Slack via webhook URL or channel ID **Subscription mode**: the Monitoring Agent can be configured to send reports to external recipients (by email) who are not Golden Analytics users, supporting use cases like weekly digest emails to executives or stakeholders. **Template categories**: anomaly detection, trend analysis, threshold monitoring, data quality checks --- ## Visualization Builder The builder uses a shelf-based drag-and-drop model for intuitive data encoding. **Shelves**: - **Rows**: dimensions or measures placed here define the Y axis or row grouping - **Columns**: dimensions or measures placed here define the X axis or column grouping - **Color**: a dimension or measure drives the color encoding of the chart - **Size**: a measure drives the size of marks (e.g. bubble size in a scatter plot) - **Trellis**: a dimension creates a small-multiples grid, one panel per value **Fields**: each column in the dataset is represented as a field (dimension or measure) in the field panel. Fields are dragged onto shelves or dropped directly onto the chart canvas. **Chart type switching**: switching chart types automatically migrates fields from the current shelf configuration to the best equivalent configuration for the new chart type. Incompatible fields are moved to an inactive shelf rather than discarded. **Format settings** (per chart): - Colors: custom color palettes, per-series overrides, sequential/diverging scales for measures - Labels: show/hide, position, format (number, percent, currency), font size - Axes: titles, min/max, gridlines, tick intervals, dual-axis configuration - Legends: show/hide, position, title - Themes: pre-built visual themes applied to the full chart --- ## Dashboard System Dashboards (called D2) are built on a freeform canvas where every component is independently positioned, sized, and styled. There is no grid constraint. **Component categories and types**: *Core data components* - **chart**: any of the 38 chart types, embedded from the visualization builder - **KPI / bigNumber**: a single large metric with optional comparison value, trend indicator, and sparkline - **table**: a data table with sorting, filtering, and conditional formatting - **choropleth**: a geographic map component *Interactive components* - **slicer / filter**: a dropdown, range slider, or multi-select control that filters all connected components on the dashboard simultaneously *Narrative components* - **annotation**: a text note anchored to a position on the canvas - **callout**: a highlighted text block with a visual accent - **quote**: a pull-quote styled text block - **divider**: a visual separator between canvas sections - **highlight**: a colored background block for grouping or emphasis *Enhanced data components* - **stat-comparison**: two metrics side-by-side with a delta and directional indicator - **progress**: a progress bar or gauge showing a metric against a target - **icon-metric**: a KPI with an icon indicator - **metric**: a compact embeddable metric tile *Creative and visual components* - **image**: a static image uploaded or linked from a URL - **hero**: a large-format banner with headline and supporting text - **gradient-text**: display text with gradient color treatment - **timeline**: a horizontal or vertical timeline of events or milestones - **comparison**: a side-by-side visual comparison of two values or states *Infographic components* - **section**: a labeled, styled container grouping related components - **status-bar**: a horizontal status indicator showing categorical states **Dashboard filters**: filter controls (slicer components) apply across all chart components on the canvas simultaneously. **Dashboard generation**: AI can generate a full dashboard from a text prompt, populating all component types appropriate for the described use case, with a consistent style preset. --- ## Export ### Presentation Export (slide screenshots) Exports the current analysis or dashboard as a slide presentation. Each sheet or dashboard component is captured as a high-resolution image and composed into the output format. **Formats**: - **PowerPoint (.pptx)**: standard presentation file - **PDF**: fixed-layout document - **Google Slides**: appends new slides to an existing presentation, or creates a new one in the user's Google Drive **Options**: - Slide size: 16:9 (default), 4:3, A4 - Quality: standard (1.5× scale), high (2× scale), maximum (2.5× scale) - Include title slide: auto-generated cover slide with dataset and date - Include chart narratives: adds AI-generated chart description text to slide notes - Optimize file size: applies JPEG compression to reduce output size ### Data Export (tabular data) Exports the underlying data from a chart or dashboard as a structured file. **Formats**: - **CSV**: comma-separated text file - **Excel (.xlsx)**: spreadsheet with column headers - **Google Sheets**: writes data to a new or existing Google Sheet in the user's Drive Data export is available from analysis sheets and individual dashboard components. --- ## Collaboration **Spaces**: shared workspaces scoped to a team. A Space contains sessions, datasets, dashboards, and stories. Members of a Space can view and interact with all shared assets. **Explore**: a browsable gallery of all assets shared across a workspace — sessions, dashboards, stories, and datasets — with search and filter. **Sharing**: assets are shared by adding them to a Space or by sharing a direct link with team members. Access control is managed at the Space level. **Export**: any chart, dashboard, or story can be exported in presentation format (PowerPoint, PDF, Google Slides) or as raw data (CSV, Excel, Google Sheets) for distribution outside the platform. **Presence**: real-time cursor and presence indicators show which team members are currently viewing or editing a shared session or dashboard. --- ## Pricing & Plans Golden Analytics offers straightforward per-seat pricing with no additional charges for queries, API calls, or AI features. **Pricing Tiers**: - **Team Plan**: $24/user/month (annual billing) or $30/user/month (monthly billing). Unlocks shared Spaces, team collaboration, and additional data connectors. - **Enterprise Plan**: Custom pricing. Includes advanced security, single sign-on (SSO), granular access controls, and dedicated support. **Trial & Access**: - Free 14-day trial available. No credit card required. - Trial users get full access to all Enterprise features. - After trial ends, account is paused for 30 days. No data is deleted during this period. **What's NOT Charged**: - No per-query fees - No per-AI-call fees - No per-transformation fees - Unlimited dashboards, stories, and data prep operations --- ## Security & Compliance **Data Handling** - Golden does not train on customer data. The AI uses schema information, statistical summaries, and small data samples to improve accuracy — not your raw data. - Data is not sent to third-party LLMs for core functionality. Golden uses a combination of models optimized for specific use cases. - File uploads are processed securely in your account. Live database connections stay within your warehouse — data is not copied or cached. **Compliance & Certifications** - Golden Analytics is currently pursuing **SOC 2 Type II certification**. Full security documentation and compliance details are available on request for Enterprise customers. - Enterprise plans include granular security controls, SSO via SAML or OIDC, role-based access control (RBAC), and audit logging. **Infrastructure** - Encryption in transit (TLS) and at rest - Secure, redundant cloud infrastructure - Regular security audits and vulnerability assessments --- ## Limitations - No native Python or R notebook environment; analysis is driven through the chat interface, visualization builder, and Data Agent — not arbitrary code execution - No built-in data pipeline or ETL orchestration; data prep covers transformations on already-loaded datasets, not scheduled ingestion or connector-to-connector data movement - No native mobile app; the product is browser-based only - Live database connections require network access and user-supplied credentials; there is no offline mode for warehouse queries - No white-label embedding SDK; dashboards are shared via Spaces or exported as files — they cannot be embedded in external sites via iframe - File-based datasets are limited to what can be processed in-browser; very large files (multi-GB) may require a warehouse connection instead --- ## Access Available at goldenanalytics.com. A free trial is available for individual users. Team plans unlock shared Spaces, collaborative dashboards, and additional data connectors. Enterprise plans support advanced security, SSO, and larger dataset limits. --- ## Key Concepts - **Session**: A saved analysis workspace containing one or more visualization sheets - **Sheet**: A single visualization within a session, with its own dataset, chart type, and shelf configuration - **Dataset**: A loaded data source (file or warehouse connection) available for querying - **Space**: A team-scoped container organizing sessions, datasets, dashboards, and stories - **Story**: A slide-based presentation built from data; produced by the Storyteller Agent - **VisualizationConfig**: The internal representation of a chart's shelf state, chart type, and all format settings - **Shelf**: A slot in the visualization builder (Rows, Columns, Color, Size, Trellis) onto which fields are dropped to encode data - **Field**: A single column from a dataset, classified as a dimension (categorical) or measure (numeric/aggregatable) - **Data Dictionary**: Per-dataset field metadata (name, type, description) used to give AI context about what columns mean; improves natural language query accuracy - **Recipe**: A saved, replayable sequence of data transformation steps; can be applied to any dataset with matching column structure - **Insight**: A structured AI-generated finding (headline/supporting/investigate) produced by the Analyst Agent - **Transformation**: A non-destructive data prep operation (split, clean, convert, derive) tracked in history and reversible at any point - **Monitoring Agent**: A scheduled job that watches a target (dashboard, dataset, or metric) and surfaces anomalies, trends, threshold breaches, or quality issues via configurable notifications - **ComponentSpec**: The typed definition of a single dashboard canvas element — its type, data binding, position, size, and style overrides - **D2**: The internal name for the dashboard system; refers to the freeform canvas and its component model