Wikipedia retail-interest velocity (daily)

Daily consumer-interest velocity signals for 17 shopping-relevant Wikipedia topics (shopping events, electronics, gaming, home appliances, home goods, toys, beauty, apparel, fast fashion, sporting goods), from the official keyless Wikimedia pageviews REST API (en.wikipedia, all-access, all-agents). Each day the trailing 120 UTC days of daily pageviews are fetched per topic (1.2 s pacing, polite User-Agent; API-omitted days stay explicit nulls on a full date grid, never zero-filled) and scored deterministically: views on as_of, 7-day and 30-day rolling averages, baseline_90d (median of the 90 days in [as_of-96d, as_of-7d], current week excluded) with baseline_std_90d, z_90d = (7d avg - baseline) / std (0.0 when the std is zero), wow_velocity_pct (7d avg vs the prior 7d), trend_30d_pct_per_day (OLS slope over 30d, in %/day) with trend_direction (rising >= +0.5 %/day, falling <= -0.5 %/day, else flat), spike_flag (spike: z >= 3, elevated: z >= 2, lull: z <= -2, else normal), and interest_heat = 100 * min-max-normalized z_90d within the snapshot, ranked as heat_rank (1 = hottest; ties: views desc, topic asc). Who joins this: an online shop joins per-topic daily interest heat to its category marketing calendar on (as_of, topic); a retail media planner joins on retail_category + as_of to time category campaigns around attention spikes. Primary key: (as_of, topic); join keys: as_of, topic, retail_category. Nullability: baseline_std_90d is null when fewer than 2 baseline days are observed; no country split is published (global English-Wikipedia readership). Caveats: pageview spikes can be news-driven rather than purchase intent; interest_heat is a within-snapshot relative measure, not comparable across days; the API lags ~1 day so as_of is the last complete UTC day. Sample use: filter spike_flag = 'spike' for today's surging shopping topics, or order by heat_rank for the day's hottest retail interest.