BigQuery gives you two fundamentally different ways to pay for running queries, and picking the wrong one can multiply your bill or leave performance on the table. The choice between on-demand and Editions comes down to how predictable and heavy your workload is. Here is how they compare.

Two ways to buy compute

The core difference is what you are actually paying for. On-demand charges by the data your queries scan, so you pay per query with no commitment. Editions charges for compute capacity, measured in slots, that you reserve or autoscale, so you pay for processing power over time rather than per byte. Storage is billed separately in both.

On-demand, in short

On-demand is the default and the simplest model. You are billed at $6.25 per TB of data scanned, with the first 1 TB each month free, and you manage nothing. It shines for spiky, unpredictable, or low-to-moderate usage, because you pay only when you query and never for idle capacity. Its weakness is that a few careless full-history scans, or steadily growing usage, can make the per-TB bill climb faster than you expect.

Editions, in short

Editions is the capacity model that replaced the old flat-rate pricing. You choose a tier – Standard, Enterprise, or Enterprise Plus – and pay per slot-hour, starting around $0.04 per slot-hour on Standard and $0.06 on Enterprise, with autoscaling and optional committed-use discounts of up to roughly 40%. It suits heavy, steady, predictable workloads, because a fixed pool of slots gives you a predictable bill and consistent performance regardless of how much data each query scans.

Point by point

The trade-offs line up like this:

  • Billing basis – on-demand: bytes scanned; Editions: slot-hours of compute.
  • Predictability – on-demand: varies with each query; Editions: steady, capacity-based.
  • Best workload – on-demand: spiky or light; Editions: heavy and continuous.
  • Management – on-demand: none; Editions: choose tiers, reservations, and autoscaling.
  • Discounts – on-demand: none beyond the free TB; Editions: commitment discounts up to about 40%.

How to choose

Start on on-demand, because most teams begin with usage that is too light and too spiky to justify reserving capacity, and the free monthly terabyte covers a lot. Watch your monthly bytes scanned as you grow. When on-demand spend becomes both large and steady, price the same workload in slots, since at high, predictable volume Editions with a commitment is usually cheaper and gives more consistent speed. Some organizations run both: Editions for scheduled production pipelines and on-demand for ad-hoc exploration.

A worked example

A quick example makes the crossover concrete. Suppose your team scans about 50 TB a month. On-demand, minus the free terabyte, that is roughly 49 TB at $6.25, around $306 a month, with zero management. If that load is steady and continuous, a modest Editions reservation running the same work can cost less per month and deliver more predictable speed, especially with a one-year commitment discount. If instead that 50 TB is really a few heavy days and a quiet rest of the month, on-demand often stays cheaper, because you pay nothing on the idle days.

Storage is separate either way

Neither model changes storage pricing, which is billed per gigabyte and drops for data untouched for 90 days under long-term storage rates. It is easy to fixate on query pricing and forget that a growing GA4 export accrues storage cost regardless of how you run queries, so prune or archive old partitions you no longer analyze.

Avoid the common mistakes

A few errors show up repeatedly:

  • Reserving slots too early, paying for idle capacity a few queries a day never use.
  • Staying on on-demand through explosive growth, where uncontrolled full scans quietly outrun a reservation's fixed cost.
  • Comparing sticker prices instead of your own measured bytes scanned and slot needs. The decision is rarely permanent, so revisit it as usage grows, because the right model at 5 TB a month is often the wrong one at 100 TB.

The pricing choice is really a question about your workload's shape, not a hunt for the single cheapest label. On-demand rewards spiky, modest, hands-off usage; Editions rewards heavy, steady, predictable usage with capacity you control. Measure your real bytes scanned and how variable they are, match the model to that pattern, and you pay for the way you actually use BigQuery instead of the way a pricing page assumed you would.

Want a stronger data analyst role or a raise? Grab the FREE Product Analyst Playbook and get the exact roadmap to your next offer.