In Splunk Infrastructure Monitoring, rollups convert raw datapoints into the chart resolution bucket before SignalFlow analytic functions run.
When you override the default rollup in data('metric', rollup='max') or in the Plot Builder, you control whether coarse time buckets preserve peaks (max), troughs (min), totals (sum), or averages (mean). Choosing average rollup on latency can hide brief 99th-percentile spikes when zooming out to a 7-day view.
SignalFlow also lets you assign multiple data() streams to variables (A, B) and combine them with standard arithmetic (+, -, *, /) or analytic functions like .percentile(95), .top(count=5), and .bottom(count=5) to surface outliers without manually inspecting hundreds of series.
errors = data('http.errors', rollup='sum').sum(by='service')
total = data('http.requests', rollup='sum').sum(by='service')
error_rate = (errors / total) * 100
error_rate.top(count=5).publish(label='Top 5 Error Rate %')