Cost & Usage Analytics

The Cost & Usage module is Quper's core FinOps intelligence layer. It is powered by the FOCUS v1.2 (FinOps Open Cost and Usage Specification) data model and surfaces seven distinct analytical capabilities — from unit economics to statistical anomaly detection — through a set of interconnected dashboards.

Module Structure

/cost-and-usage
├── /                         → Overview — KPI cards + spend trend
├── /unit-cost
│   ├── /summary              → Unit cost + efficiency by dimension
│   └── /timeseries           → Daily unit cost trend with lag growth
├── /period-compare           → Side-by-side window comparison
├── /optimization
│   ├── /indicators           → Idle, overprovision, concentration
│   ├── /baseline             → Z-score anomaly detection
│   ├── /stability            → Volatility (CV) classification
│   └── /advanced             → Idle days, zero-query, off-hours, underutil
└── /config                   → Threshold configuration panel

Overview Dashboard

The landing page aggregates the gold table into four KPI cards and a 30-day spend sparkline:

• Total effective cost — SUM(effective_cost) for the current billing period, displayed with period-over-period delta
• List cost savings — 1 − (effective_cost ÷ list_cost), showing the aggregate discount rate vs. public pricing
• Idle spend — SUM(idle_cost) and its percentage of total effective cost
• Tag coverage — 1 − (required_tag_missing_flag rows ÷ total rows), a FOCUS-derived tag governance signal

Unit Cost Analysis

Summary View

Computes unit cost and efficiency for any grouping dimension over two configurable time windows. The result is a ranked table showing each group with its current and previous unit cost, growth percentages, and an efficiency ratio.

The efficiency ratio answers: is cost growing proportionally to usage, faster, or slower? A ratio above 1.0 means cost grew faster than usage (spending more per unit); below 1.0 means usage grew faster (better economies of scale). The value is clamped to ±10 to suppress extreme readings from near-zero denominators.

Grouping dimensions available: service, region, resource, billing_account, workload, cost_type, or any tag:<key> value. All are direct FOCUS v1.2 columns or derived from them.

Time-Series View

Renders daily unit cost as a line chart using Recharts, with a secondary line for the day-over-day growth rate. Data is computed using a SQL LAG window function partitioned by group key — a single query pass produces all daily deltas without self-joins.

Period Comparison

Side-by-side bar chart comparing two time windows of equal length. Two modes are supported:

• Previous period — The N days immediately before the current window. Best for detecting week-over-week or sprint-over-sprint drift.
• Month-to-date (MTD) — The same calendar day span in the prior month. Best for budget-cycle comparisons.

The comparison uses a FULL OUTER JOIN so resources present in only one period are still surfaced — newly provisioned workloads appear with a 100 % increase; decommissioned ones appear with a 100 % decrease. The chart uses Recharts BarChart with negative values to show deltas inline.

Optimization Indicators

Indicators Dashboard

Three optimization signals displayed as gauge cards, each with a current value, a configurable threshold, and a status badge:

Baseline Anomaly Detection

A line chart overlaying daily spend against a 14-day rolling baseline band (mean ± 1 standard deviation). Days where the Z-score exceeds the configured threshold are highlighted with a red point marker.

The baseline window spans D-16 to D-3 — the 14 days ending two days before the current day. This prevents today's spike from influencing its own reference. A minimum of 7 data points are required before any anomaly flag is shown.

Stability (Volatility) View

Each resource or service group is classified into one of four stability buckets based on the 14-day rolling Coefficient of Variation (CV = stddev ÷ mean) of daily spend. The view renders a scatter plot of CV vs. cost magnitude, allowing teams to identify high-cost, high-volatility workloads as the primary commitment candidates.

Advanced Optimization

Four deeper waste-detection views, each addressing a pattern that the summary indicators cannot capture alone:

• Idle Days — Calendar heatmap of days whereidle_flag = TRUE per resource, plus total idle_cost attributable to those days
• Zero-Query Days — Redshift-specific: days where compute cost was charged but no queries were executed (query_count = 0 in the gold table)
• Off-Hours Spend — Donut chart of cost incurred outside business hours, computed from the silver table's hourly charge_period_start. Business hours are configurable per tenant (default 09:00 – 18:00 UTC)
• Underutilised Clusters — Table of resources where measured CPU was above zero but below 30 % (default threshold), distinguishing under-use from true idle

Threshold Configuration

The /config route renders a form backed by the qup_config_idle_thresholds table. All nine threshold values — idle CPU, idle query count, overprovisioning CPU/queue/latency, concentration top-N, business hours start/end, and underutilisation CPU — can be updated at runtime. Changes propagate to all subsequent queries without a service restart.