When finance leaders see their first Fabric invoice, the question is almost always the same: why does this number move when our SKU didn't? The answer lives in three mechanics that rarely make it into the sales deck.
Smoothing spreads your spikes
Fabric smooths background workloads over 24 hours and interactive workloads over shorter windows. A heavy overnight job doesn't hit your capacity all at once; it's spread across the day. That's good news, until daytime interactive use lands on top of yesterday's smoothed batch load.
Bursting borrows from the future
Capacities can burst above their nominal size to finish work faster. The bill for that burst is repaid by future idle time. If there's no idle time, you get throttled, and throttling usually shows up at month-end, exactly when finance needs reports.
Throttling is a symptom, not the disease
The fix for throttling is rarely a bigger SKU. In our benchmark of 85 estates, the top ten workloads accounted for 61 percent of capacity consumption, and most had obvious tuning opportunities: duplicated refreshes, unpartitioned Spark jobs, and dev workloads running on production capacity.
Before you buy a bigger SKU, find out who is using the one you have.
What good looks like
Attribute every capacity unit to a workload and an owner. Separate dev and prod capacities, and pause dev at night. Schedule heavy batch work inside smoothing windows. Then, and only then, decide whether you need to size up. Clients who follow this sequence save an average of 27 percent in the first quarter.