SysPad › Lessons › analytics lessons › Amazon QuickSight
How Amazon QuickSight works
Serverless BI dashboards on top of Athena/Redshift/S3. Priced per author/reader seat plus SPICE in-memory capacity - a people cost, not a traffic cost.
Every analytics pipeline diagram trails off at the right edge: Glue jobs, Athena queries, Redshift tables... and then what? Someone has to LOOK at the numbers. QuickSight is that last hop, and it breaks the mental model of everything before it: it is priced per person, not per request.
What you will learn
- Where does QuickSight sit in the analytics pipeline?
- What is SPICE and why does it exist?
- How does per-seat pricing change the cost conversation?
- When do embedded/capacity sessions beat seats?
The pipeline terminus
QuickSight is serverless BI: connect it to Athena, Redshift, RDS or S3, build dashboards, share them with the team. It is a human-facing terminus, so in a SysPad simulation traffic ENDS there; if thousands of RPS are flowing into a BI tool, the diagram usually wanted an API instead.
SPICE: the in-memory trick
Live-querying Athena on every dashboard view would be slow and expensive (Athena bills per TB scanned). SPICE is QuickSight's in-memory engine: import the dataset on a schedule, and dashboard interactions hit RAM instead of re-running warehouse queries.
People pricing
Roughly $24/month per author (builds dashboards) and $3/month per reader (views them), plus SPICE capacity beyond the included allowance. At hundreds of readers, capacity pricing (per session) becomes cheaper than seats, and it is the model used for embedding dashboards in your own app.
⚓ On the SysPad node, cost moves with the authors/readers/SPICE sliders and ignores RPS entirely. That inversion IS the lesson.
What about embedding a dashboard in our product for customers?
Embedded QuickSight with session-capacity pricing is exactly that: your app requests an embed URL, customers view dashboards without QuickSight accounts, and you pay per session bundle. It competes with building charts yourself; the break-even is developer time vs session fees.
Flashcards
- An architect draws API Gateway → QuickSight to "serve analytics to the mobile app" at 500 RPS. What is the category error?
- QuickSight serves human dashboard sessions, not API traffic. The mobile app wants an API over pre-aggregated tables (or embedded dashboards via session pricing), never per-request BI.
- Support escalates: "the ops dashboard says zero orders today but orders are flowing." The pipeline is healthy. Likeliest cause?
- The SPICE dataset has not refreshed (failed or scheduled nightly). SPICE trades freshness for speed and cost; check the refresh schedule before blaming the pipeline.
- Rolling dashboards out from 10 analysts to all 800 employees: what happens to the bill, and what is the standard move?
- Per-reader seats jump toward $2,400/mo; at that scale session-capacity pricing usually wins, since most of the 800 will peek monthly, not daily. Price both before rolling out.
Sources
Open the SysPad canvas · Official Amazon QuickSight documentation