liftlens

Confirm the lift before you launch using liftlens, an enterprise-grade A/B testing platform with:

• CUPED, SRM detection, sequential testing (OBF/Pocock)
• Heterogeneous treatment effects (Causal Forest, S/X/T-Learner)
• FastAPI + JWT auth, Streamlit dashboard
• Parallel execution (Joblib/Dask/Ray)
• Interactive HTML/PDF reports with Plotly
• SQLite experiment registry
• 74%+ test coverage, CI/CD ready

Quick Start

# 1. Clone & install
git clone https://github.com/dare-afolabi/liftlens.git
cd liftlens
poetry install --with dev,docs

# 2. Generate synthetic data
poetry run generate-synthetic --n_users 10000 --output data/synthetic.csv

# 3. Run experiment
poetry run liftlens run \
  --config examples/revenue_test.yaml \
  --output output/run_1

# 4. Launch API + Dashboard
docker-compose up

• API: http://localhost:8000/docs
• Dashboard: http://localhost:8501
• Report: output/run_1/report.html

Configuration (YAML)

# examples/revenue_test.yaml
name: Revenue Lift Test
data:
  type: csv
  path: data/synthetic.csv
baseline_col: baseline
outcome_col: outcome
group_col: group
metrics:
  - name: revenue_lift
    func: mean_diff
  - name: arpu
    func: ratio_metric
    params:
      denominator_col: active_users
transform:
  winsorize: [0.01, 0.01]
  cuped: true
  log_transform: true
stats:
  sequential: true
  method: obf

Features

• CUPED Variance Reduction
• SRM Detection (Chi²)
• Sequential Testing (OBF/Pocock)
• Bootstrap + Permutation Tests
• Multiple Comparison Corrections
• Heterogeneous Treatment Effects
• ANCOVA, Mixed Effects, GAM
• FastAPI + API Key Auth
• Streamlit Dashboard
• HTML/PDF Reports (Jinja2 + WeasyPrint)
• Parallel Processing
• SQLite Registry
• Docker Compose

Project Structure

liftlens/
├── core/           # Session, registry
├── config/         # Pydantic schemas
├── data/           # IO, validation, transforms
├── metrics/        # Primary, robust, composite
├── stats/          # Inference, power, sequential, HTE
├── viz/            # Plotly: dist, effects, diagnostics
├── report/         # Jinja2 templates + builder
├── workflows/      # Pipeline, presets
├── engine/         # Parallel execution
├── api/            # FastAPI, Streamlit, auth
├── utils/          # Logging, decorators
└── cli.py          # Typer CLI

Development

# Install pre-commit
poetry run pre-commit install

# Run tests
poetry run pytest -m "not slow"

# Lint & format
poetry run ruff check .
poetry run black .

# Build docs
poetry run mkdocs serve

License

Apache 2.0