Multi-Touch Attribution SimulationReproducible synthetic paths with caller-owned configuration and storage

Start here ​

Generation flow ​

Follow configuration loading, complete in-memory simulation, validation, table selection, storage, and manifest creation.

External configurations and writers ​

Run the package from another repository, select only the path table, or connect a trusted private database or custom-format adapter.

Dataset contract ​

Understand which tables are training inputs, which table is evaluation-only, and which invariants every adapter must preserve.

Simulator domain model ​

Review the independent Provider, Touchpoint, Product, Campaign, observation, and lineage dataclasses used by MTA-SIM without a downstream runtime dependency.

Research configuration and scale profiles ​

Configure Providers, Products, Campaigns, Ad Groups, product economics, budget levels, diminishing response, reproducibility snapshots, and the 10k CSV or 100k direct-PostgreSQL modes.

Baseline simulator ​

Review the stable simulation mechanism and the fictional public example without exposing production parameters.

Regional extension ​

See how a private caller can add regional behavior variables while retaining the same table and writer contracts.

Public release checklist ​

Audit the current tree and scrub historical commits before changing repository visibility.

Credits and provenance ​

See who contributed the maintained implementation, legacy samples, regional scenarios, and data-modeling references.

Repository analysis ​

Compare the legacy generator and peer samples with the maintained simulation pipeline.

Definitions and mathematical notation ​

Reader-friendly explanations for marketing, economics, probability, and simulation terms.

Model conjectures ​

Candidate attribution approaches that require modeling-team review. They are proposals, not implemented production claims.

Status labels ​

Documentation headings may show a short colored indicator on their right:

• Verified — directly confirmed in this repository.
• External — supported by a cited source outside the repository.
• Inference — a reasoned interpretation that is not directly measured.
• Recommendation — a proposed next action or design choice.

Public-data boundary ​

The repository retains peer-contributed scenarios, workbooks, configurations, generators, and synthetic datasets as credited research references. It excludes credentials, customer-level records, production account identifiers, private prompt transcripts, and local paths. Retained samples are not verified Amazon or brand data.

Maintainers: ZheyuanWu and GPT5.6-Sol. Peer contributors are recorded in Credits and Provenance. Repository: Trance-0/MTA-SIM-dataset.