This critical piece of business software, as the name implies, acts as a repository for all sorts of data—sales totals, customer service records, employee information—you name it. Putting all that diverse information into a super database makes it easier to analyze the information and find patterns that could boost sales or improve customer service.

An example of a fairly common reporting task is to see how many people came to a company’s web site, signed up for something, and then renewed a year later.

Traditionally, data warehouses from Teradata (TDC), Oracle (OCLCF), or SAP (SAP) ran on a company’s internal servers. But in the past few years, businesses have started to use cloud-based data warehouses like Amazon Redshift, Google BigQuery, or Snowflake. These data warehouses run on third-party cloud infrastructure—RedShift and Snowflake on Amazon Web Services, for example, and Google BigQuery on Google (GOOGL) cloud.

To help potential customers suss out which of those cloud options is fastest or cheapest, startup Fivetran came up with some benchmark tests to compare them. Fivetran, Oakland, Calif., sells technology that “cleans up” data and pumps it into a cloud data warehouse. Since it works with all three options, it can claim neutrality.

Related: Larry Ellison Says Oracle’s Next Database Will Beat Amazon

“We have no dog in this fight,” co-founder and CEO George Fraser tells Fortune. For the benchmark, outlined in a Fivetran blog post, the company used 99 standard queries that are designed for testing data warehouses.

Google BigQuery turned out to be the priciest option in this benchmark. “It was shocking that BigQuery was more expensive than the other options,” Fraser says. Although Fraser said he had heard anecdotally this might be the case, he had no hard evidence until now.

The price comparison is difficult given that Redshift and Snowflake charge by the hour while the warehouse runs. BigQuery, on the other hand, charges per query. To even things out, Fivetran estimated how much idle time would occur with Redshift and Snowflake and factored that in. Having done that, BigQuery was roughly twice as expensive as its two rivals with a median query cost of $0.041 compared to $0.023 for Redshift and $0.017 for Snowflake.

Related: This Data Crunching Startup Just Got $100 Million in Funding

Even then, Fraser maintains that all three options offer excellent price performance. The upside of BigQuery, he notes, is it is very easy to manage. “You put in a query and you get results,” he says.

In terms of query processing speed, the difference between the three warehouses was tiny, in Fraser’s view. BigQuery was fastest with a median execution time of 5.3 seconds. Snowflake was next at 5.8 seconds, and Redshift was third at 7.7 seconds. “For most users anything under 10 seconds is fine,” he says.

Get Data Sheet, Fortune’s technology newsletter

Fraser says it was a bit surprising that Redshift was slower than Snowflake. Because Snowflake is more of a managed service—meaning that the user has to do less set-up and tweaking, he expected it to show a performance hit. That was not the case.

Note: (Oct. 9, 2017 11:20 a.m. EDT) This story was updated to add detail about the benchmark’s format.