Plain-language guide

This page explains tscfbench without assuming causal-inference jargon.

The core question

A before/after question sounds simple:

• Did the launch really increase signups?
• Did the heatwave really increase ER visits?
• Did the outage really change grid demand?

The hard part is deciding what would have happened otherwise.

tscfbench helps you estimate that missing path and package the result into a chart, a report, and a shareable handoff.

Translations from research language to everyday language

• Counterfactual → the path you think would have happened without the event.
• Treated series → the one series you care about.
• Controls → other series that help anchor the missing baseline.
• Donor pool → the comparison units used to build a synthetic baseline.
• Placebo test → a sanity check asking whether the same method would claim an effect where there should be none.

Three beginner patterns

One product launch

Use one outcome series and a few control series.

One treated region

Use one treated city/state/region and several comparison units.

One public event

Use a public attention or market series with peer-series controls.

What makes tscfbench different from just installing a model package?

A model package gives you an estimate. tscfbench gives you the estimate plus the workflow output: charts, a report, a share package, and agent-ready structured files.