Edge Health toggle — live data preview · 2026-05-18
DARE.CO.UK · TOGGLE DATA · 2026-05-18 (one-shot snapshot for sketch validation — not yet wired into the live dashboard)
This is what the parked Edge Health 24h / 7d / 30d toggle would render right now, fetched against today’s Cloudflare zone analytics. Source-of-truth sketch: parked_sketch_edge_health_time_toggle_2026-05-17.
The numbers — 4 metrics × 3 windows
| Metric | 24h | 7d | 30d |
|---|---|---|---|
| cache hit rate | 59.4% (+5.2pp vs prior 24h) | 56.9% (+21.8pp vs prior 7d) | 17.8% (+14.9pp vs prior 30d) |
| bandwidth offloaded | 0.19 GB (77.3% of 0.25 GB) | 0.91 GB (76.1% of 1.20 GB) | 2.42 GB (28.0% of 8.67 GB) |
| threats blocked | 626 (+54.2% vs prior 24h) | 3,878 (-61.2% vs prior 7d) | 34,159 (-30.4% vs prior 30d) |
| 5xx error share | 0.01% (2/17,784) | 0.18% (139/78,290) | 2.83% (13,793/486,734) |
Per-card narrative reading
- cache hit rate. 24h is “right now”; 7d is “this week”; 30d is “this month.” The wider the window, the smaller the delta-vs-prior swings — that’s the toggle’s value: incident-response vs trend vs baseline are three different reads of the same metric.
- bandwidth offloaded. Card title would re-label per window: “today’s GB” / “weekly GB” / “monthly GB.” Percentage of total bandwidth from cache is the more comparable cross-window read.
- threats blocked. Card body would re-label per window: “today’s count” / “weekly count + daily avg” / “monthly count + weekly avg.” The delta-vs-prior framing stays constant.
- 5xx error share. Card body would re-label per window: “today’s %” / “rolling 7d %” / “monthly %.” Watch the absolute count too — a small % over a big base can still be a real incident.
Baseline-label phrasing per window
Top-right copy on the cards row (currently vs 90-day baseline) would switch:
- 24h selected →
vs prior 24h · click any alert to view in Cloudflare - 7d selected →
vs prior 7d · click any alert to view in Cloudflare - 30d selected →
vs prior 30d · click any alert to view in Cloudflare
What this data validates / opens up
- All three windows fit inside CDN, security layer, and DNS provider sitting in front of dare.co.uk.">CF’s free retention envelope at the granularity we’d use (1d for 24h/7d/30d in this preview; the real toggle should switch 24h to 1h granularity for a truly rolling read, see refinement note below).
- The numbers above prove the adapter chain works end-to-end (
fetch(days, offset)for current + prior periods across all three windows). - Refinement when the toggle is built: extend
HOURLY_QUERYto includebytes / cachedBytes / threats / responseStatusMapso 24h can use 1h granularity (rolling last-24-hours read instead of yesterday’s-daily-aggregate). Today’s preview usesfetch(days=1)for 24h purely because the existing hourly query is requests-only. - Once the toggle is built, the same data shape pre-fetches at render time — 3× the GraphQL payload per hourly refresh, well within free tier.
- The per-card reframe text from the sketch holds up against real numbers; no surprises.
Next step (~45 min)
Wire the data into _edge_health_cards(window: str) in dare_cf_analytics.py, add the pill toggle markup, JS click handler, localStorage persistence. See the parked sketch for the cycle pre-fill.