Estimate what would have happened without a campaign when geo or user-level holdouts are unavailable. Time Series Lift fits univariate StatsForecast and Prophet models to the pre-period, forecasts through the campaign and carryover windows, and compares observed outcomes with placebo-calibrated counterfactuals.
Lift after an intervention
The expected baseline is forecast from the pre-period. Coincident shocks can be mistaken for lift, so use a randomized holdout whenever one is feasible.
Time Series Lift is a strong fit when the intervention happens at a known point in time and the main question is how much behavior changed relative to a credible counterfactual.
Estimate lift for launches, sponsorships, PR bursts, and media changes when a clean geographic or user-level holdout cannot be created.
Upload a daily or weekly outcome series. The Early Access model roster uses the outcome's pre-period trend and seasonality; control-series covariates are not part of this release.
A coincident market, pricing, product, or operational shock can look like campaign lift because a univariate model cannot separate the two. Prefer a randomized geo or user-level test whenever one is feasible.
Move from a clearly defined intervention to a backtested counterfactual and a decision-ready readout.
Upload historical outcomes, mark the campaign and carryover dates, document other business changes, and align each metric to the decision you need to make.
Compare Seasonal Naive, AutoARIMA, AutoETS, AutoTheta, MSTL, and constrained Prophet where the available history supports them. Pre-period backtests determine which models are credible.
Review campaign, carryover, and full-period lift; placebo-calibrated intervals; model sensitivity; and spend economics such as iROAS before you scale, pause, or re-test.
Time Series Lift is best suited to bounded interventions that are national, bursty, or otherwise difficult to randomize.
Measure brand-search or direct-traffic response to tentpole moments like Super Bowl ads, creator drops, or national PR campaigns.
Estimate performance after a creative change when holdouts are unavailable, while accounting for forecastable baseline trend and seasonality.
Estimate the impact of pacing changes, bid-strategy shifts, offer launches, and channel-on/channel-off periods using transparent counterfactual logic.
You don't need to be a statistician. The hard math runs under the hood; you point, click, and make the call with the full picture in front of you.
Geo, user-level, and platform tests all feed one central database. That shared truth calibrates and validates your MMM, while attribution keeps everything pointed in the right direction. Run any piece on its own — or run them together and let each one make the others stronger.
Inform. Attribution gives the MMM a fast, directional read on what's working between tests.
Calibrate. Incrementality tests anchor the MMM to causal ground truth, not just correlation.
Validate. Holdout tests check that attribution and the model agree with what really happened.
Time Series Lift fills the gap when a controlled experiment is not feasible. Controlled geo and user-level tests remain the stronger causal designs when they are available.
Use geo tests when you can enforce stronger controls. Use time-series lift when the intervention is national or otherwise hard to randomize cleanly.
Learn moreUse randomized user-level holdouts when targeting and outcome capture make them possible; they separate treatment impact from coincident changes.
Learn moreStore platform lift studies alongside time-series reads so you can compare fast platform evidence with your own independent counterfactual models.
Learn moreThe models and tests matter, but the workflow around them matters too. Shako Stats is designed to become the operating system around experiment planning, metadata, documentation, and cross-test learning.
Centralize datasets, mappings, and historical records so experiments and models always start from the same source of truth.
Organize tests by audience, creative strategy, bidding logic, or business objective so learnings remain searchable and reusable.
See what tests are planned, in-flight, or completed so overlapping interventions and measurement conflicts are easier to manage.
Turn methodology, definitions, and experiment design guidance into an internal operating system instead of leaving them scattered across decks.
Connect any AI agent through MCP/API, including Claude, ChatGPT, Cursor, or any MCP client, to a read-only context layer over your real incrementality workspace: processed data, data dictionary entries, metric definitions, test designs, results, diagnostics, and model outputs.
Meet your AI Marketing ScientistOne-click OAuth for Claude & ChatGPT, read-only, anyone on the team can ask
Design tests, understand results, and connect your team to one statistics-backed measurement workspace.
Paid B2B plans built for brands & agencies.