Competitive data collection and briefs
Structure dimensions (features, pricing, channels) with sources and timestamps; deliver comparison tables and takeaways, noting staleness and compliance boundaries.
Case category · Experiments & insight
5 cases Category 6 of 20
This band targets analytics science and strategy research: competitive intel, feature documentation for models, data-quality playbooks, A/B design through readout, and company-wide metric definitions. Emphasis on reproducibility and auditability; agents help with templates and first drafts, while statistical and causal claims stay with humans.
In the case hub it is Experiments & insight (#cat-insight), paired with Data & analytics: this band stresses methods and experiment assets; the data band stresses routine querying and reporting.
Collection dimensions, price/feature comparison, conclusions.
Feature sources, transforms, missingness, model inputs.
Rules, outliers, cleansing logs, replay validation.
Hypotheses, sample size, significance, business readout.
Numerator/denominator, windows, exclusion rules, versions.
Structure dimensions (features, pricing, channels) with sources and timestamps; deliver comparison tables and takeaways, noting staleness and compliance boundaries.
Record lineage, transforms, missing/outlier handling, and train/serve parity to support model review and reduce silent drift.
Prioritize rules, outlier strategies, and cleansing log formats; enable replay and reconciliation for audit questions about “what changed in the data.”
Move from business hypothesis to metrics and power; cover stratification and multiple-testing cautions; connect results to confidence intervals and decisions, not only p-values.
For each core KPI document numerator, denominator, time window, dedupe rules, and version history—your single source of truth for dashboards and OKRs.