← journal·2026-09-20
The 100ms Budget: Where Milliseconds Go
Not financial advice. Verify claims independently.
"Sub-100ms" is a slogan until you write the ledger. Here is ours, measured on the public RT/SOFT dashboard and mirrored in the Stock Picks paper feed.
The ledger
| Stage | Budget | Notes |
|---|---|---|
| Exchange → ingest | 0–8ms | Vendor colo; not ours to shave |
| Normalize + shard | 4ms | Fixed; hash + schema |
| Pub/sub fan-out | 8ms | Per-connection schedule |
| Edge PoP egress | 18ms | Region dependent |
| Client decode + paint | 6–40ms | Device & main-thread |
Cross-region hops dominate. That is why we publish per-region p50/p95/p99 instead of a single vanity number.
What "transparent" means
Competitors often show a synthetic ping to their nearest edge. We show:
- Ingest → fan-out → edge breakdown
- Live channel table with p50/p99
- The same numbers the Stock Picks mobile client sees
If the dashboard disagrees with your SDK telemetry, file a ticket — the dashboard is wrong, or your clock is.
Rehearse before you buy
Open the Stock Picks live demo, watch the quote stream, and compare your stopwatch to our public SLA strip. Marketing claims are cheap. Millisecond ledgers are not.
Put it on a live wire
Stock Picks runs on RT/SOFT — open the paper app and watch the same fan-out path that powers this latency dashboard.
$ open Stock Picks →