Questions, answered straight

On customer data there is no answer key, so historical fit alone cannot determine which estimator recovered the true effects. SmartInfer Research compares methods on synthetic datasets with known economic ground truth, but no method was universally superior across those experiments. Galileo uses the product estimator configured for the analysis and reports its diagnostics and limitations; a research estimator is not silently substituted because it performed better on one benchmark.

We test Galileo on specified synthetic regimes where the data-generating effects are known, then measure recovery, uncertainty behavior, and decision quality against that answer key. Those tests show how the method behaves under the conditions represented by each benchmark. They do not guarantee accuracy on customer data, where the true effects are not observable. The How we validate page scopes each public result to the benchmark it evaluates.

Galileo’s current published product artifacts use Hill saturation and geometric adstock, with nonlinear response fitting, regularized channel coefficients, and classical uncertainty estimates. The product surrounds that estimation with identifiability and credibility checks.

The current published product artifacts are the basis for this description. Galileo 0.2 is separately identified as a Bayesian research estimator; it is not presented here as the deployed product estimator.

Galileo models baseline and calendar patterns before estimating channel response. This reduces the risk of assigning ordinary seasonal movement to channels that happened to spend during the same period. Known events and business changes can be included as context, but observational controls do not guarantee that every confounder has been removed.

Because synthetic data has known ground truth. We plant real effects and measure whether the engine recovers them — that's the only way to show you not just a report, but that the report is checked. A demo on real client data would show a plausible answer with no way to prove it right.

Galileo estimates the aggregate channel response associated with how the channel was operated during the modeled period. Creative, targeting, placement, and content choices are therefore reflected jointly in the observed channel history; Galileo does not identify which individual choice caused the result.

Isolating a creative, placement, audience, or personalization effect requires independently varying data at that level or a controlled experiment. If channel strategy changes materially during the analysis period, a refit or narrower analysis may be required.

Galileo can model how traffic varied with channel spend and where the fitted response saturates. It labels the result as a traffic-response analysis and does not convert sessions or clicks into revenue claims or financial budget recommendations.

It doesn't break it, and — importantly — we won't tell you that adding a precise promo calendar improves accuracy, because right now that isn't true. Modeling promotions as a separate effect is an active correction we treat as a known limitation, not a selling point, until that path is recalibrated. We'd rather disclose it than sell it.

Galileo is designed and tested to keep quantitative answers bound to computed engine output. If the requested quantity is absent from the analysis, the agent should refuse to estimate it and direct the user to the available report evidence.

It won't estimate numbers its engine didn't compute, it won't update a model when new data contradicts it (it asks for a refit instead), and it won't speculate about analyses it doesn't have. Individual-user, campaign-level, and product-segment attribution are out of scope — this works at the whole-channel, week-by-week level, and it isn’t a substitute for controlled experiments (lift tests, or geo holdouts that pause spend in some regions to read the true effect).

A weekly CSV: date, spend per channel, and revenue. Six months is the minimum, twelve is better. The one thing that matters as much as length is variation — a channel whose spend never changes can’t be told apart from everything else moving alongside it (it can’t be “identified,” in the jargon), and the report will tell you so rather than guess. Most teams export this in under an hour.

Galileo watches for exactly that. If the data shifts enough that a single model would be stale — a new channel, a big step-change in spend, a fundamental shift in how the business works (a “structural break”) — it refuses to quietly update on top of the old model and asks for a refit instead. A stale model that keeps answering is more dangerous than one that admits it's out of date.

CSV upload — one weekly file. Extra context is optional and enters two ways: two event-marker columns in the same file — promo_depth (weekly promo depth, or 0/1) and stockout (0/1) — which enter as linear event controls with a modest, honestly-limited effect; or the optional interview, where known seasonality, past lift studies, and channel pauses are captured as priors and checked against the data before they can affect the model. Your data stays in your environment: stored on disk within the Galileo instance, no external database, no third-party analytics, nothing leaving your network unless you choose to share it.

One column per sub-channel — but only split one out if its spend moves independently and is large enough to matter. If two lines always rise and fall together (the jargon is “collinear”), Galileo can’t tell their effects apart, and it says so plainly rather than inventing a split. When in doubt, start coarse — a single “YouTube” column — and separate later if the spend genuinely diverges. The full data guide has naming examples and the split rule.

Offline and owned activity can be represented when it has a meaningful weekly measure and enough independent variation to estimate separately. Galileo then estimates its response under the same model assumptions and credibility checks used for other channels. A flat contract or nearly constant activity may be absorbed into baseline rather than separately identified.

Galileo estimates channel response at the business or product scope represented by the input outcome. A product should be modeled separately only when it has its own spend and outcome history. Creative or personalization effects remain inside the aggregate channel estimate unless separately varied or evaluated through an experiment. See the full data guide for scope.