The project
Why gallop
Statistics libraries assume you already know which question you are answering. Product management skill packs know which question matters and carry no statistics. A number goes wrong in the gap: a lookup becomes a deep dive, an unvalidated metric becomes a readout, a launch with no comparison group gets an effect size anyway.
gallop puts the routing and the rigour in one place, as agent skills. One lookup, three questions and one gate before any query runs, and many leave at the gate. The rest get the method that matches how treatment was assigned and a readout that runs the checks a wrong number fails.
Who it is for
The data scientist embedded in a product team, and the analyst or product manager who runs tests without a platform doing the safety checks for them. Not for research scientists, or for anyone whose experimentation platform already fills in the detectable effect from a prior store. Those people have this problem solved.
See also
- The method map: the buckets, the layering, why measurement is a floor rather than a phase, and why theory is a ceiling rather than a report.
- The intake: the routing pass run on every request.