OVERVIEW
What the Analytics Lab is
The Analytics Lab is an analysis environment, not an automated opinion generator. You choose the question, data and method. The platform calculates, checks and explains the result.
Begin with what you want to learn, not whichever test looks interesting.
Statistics are produced by a versioned analysis kernel, not invented by a language model.
Inputs, transformations, parameters and limitations remain inspectable.
AI can organise and explain evidence. It does not make investment decisions.
FIVE-MINUTE START
Your first analysis
- 1Open Analysis Studio
The Studio is the default Analytics view and contains every app in one workspace.
- 2Choose an analysis app
Select the app whose question matches yours. Avoid choosing a method only because you recognise its name.
- 3Load or select data
Upload CSV/XLSX, paste a table, or load a public series. Text and option apps use their own specialised inputs.
- 4Describe your question
Plain English is enough. Mention the variables, comparison, time horizon or evidence set you care about.
- 5Review and run the method
Edit steps and parameters before approval. Then inspect results, robustness and limitations.
- 6Pin or export
Save useful findings to the Notebook or export a plan and analysis record.
INSIDE THE STUDIO
A simple research IDE
Ask → edit → run
Your question becomes an editable methodology. Column choices and parameters stay visible.
- Analysis Studio
- The guided, question-first environment that can launch every app.
- Apps
- Focused workspaces for running one family of methods in depth.
- Playground
- A chat and workflow surface for chaining supported methods across datasets.
- Notebook
- Persistent browser-local research memory for findings and your own notes.
DATA AND CLEANING
Get the data trustworthy first
The same dataset workspace is shared across quantitative apps. Original data is preserved; approved cleaning actions form a reversible pipeline.
Before running a method
- Confirm units, frequency and time ordering.
- Inspect missing values and duplicate rows.
- Check whether numbers were imported as text.
- Decide whether an outlier is an error, an event or a valid observation.
- For joined series, verify keys and frequency alignment.
ANALYSIS APPS
Choose by question, not by complexity
Summary & Distribution
Understand one variable before testing it.
- Input
- One numeric column
- Output
- Centre, spread, shape, normality and missingness
Regression & Correlation
Measure how two variables move together.
- Input
- Two numeric columns
- Output
- Pearson, Spearman, OLS, uncertainty and residual diagnostics
Group Comparison
Compare two independent groups or paired measurements.
- Input
- Outcome + group, or two paired columns
- Output
- Effect size, parametric and rank-based tests
Causality & Time Series
Study ordered observations and predictive timing.
- Input
- One or two ordered series
- Output
- Stationarity, autocorrelation, Granger and cointegration
Risk & Portfolio
Measure historical risk and benchmark exposure.
- Input
- Prices or returns; optional benchmark
- Output
- Volatility, drawdown, VaR, CVaR, beta, alpha and rolling exposure
Technical Analysis
Describe historical price behaviour.
- Input
- One ordered price series
- Output
- RSI, MACD and Bollinger bands
Qualitative & Research
Organise evidence from documents and text.
- Input
- News, filings, calls or interview text
- Output
- Codes, themes, tone screens, quotations and evidence records
Derivatives & Options
Inspect theoretical option values and payoff structures.
- Input
- Contract and model assumptions
- Output
- Black–Scholes value, Greeks and payoff diagrams
HOW AI HELPS
AI is the interface, not the calculator
- Translate a question into editable analysis steps
- Explain method choices and terminology
- Identify metadata and data-quality concerns
- Summarise computed results and caveats
- Organise text evidence and themes
- Invent or privately recompute statistics
- Change data without approval
- Present association as real-world causation
- Recommend buying, selling or holding
- Provide targets, allocations or rebalancing instructions
When a question is clear, structured rules select a compatible tool without an AI call. AI is reserved for ambiguity, planning and explanation. Calculations always run through the common analysis kernel or a specialised deterministic canvas.
READING RESULTS
Read the evidence in the right order
How many observations were usable? Are ordering, grouping and missingness credible?
How large is the measured relationship, difference, exposure or risk?
What do the interval, p-value or confidence label say about sampling uncertainty?
Do residuals, stationarity, robustness or sensitivity checks weaken the result?
What can this method and dataset not establish?
REPRODUCIBILITY
Make every result inspectable later
A defensible analysis record includes the research question, source data, dataset fingerprint, approved cleaning steps, method, selected columns, parameters, engine version, metrics and limitations.
Machine-readable methodology, app, steps and parameters.
Human-readable methods and results for reports or review.
Saved findings with your own interpretation and provenance.
The transformed table used by the analysis.
PRIVACY AND BOUNDARIES
Private by default
Uploaded datasets, cleaning pipelines, autosaves and Notebook entries stay in the browser. Quantitative calculations run locally. Export a project when you need a portable backup.
FAQ
Common questions
Do I need to know which statistical test to use?
No. Start with the research question and review the proposed methodology. The descriptions and assumptions remain visible so you can make an informed choice.
Does AI calculate the statistics?
No. Registered quantitative tools execute through the deterministic analysis kernel. AI may explain the results it receives.
Can I use the result as a buy or sell signal?
No. The platform measures data and evidence. It does not recommend transactions, allocations, target prices or investment decisions.
Why did a method refuse to run?
Common causes are insufficient observations, incompatible column types, identical variables, missing time ordering or assumptions that the selected data cannot satisfy.
How do I continue an analysis later?
Use Save project to export the data, cleaning pipeline, lineage and Notebook. Restore that project in a future session.