There is an implicit assumption that the input data will always be 1-minute interval data starting at 00:00 of some day and ends in 01:40, which then have to be processed into 5-minutes intervals.

This information completely affects how the kata should be approached, so either it should be mentioned or the generated data should be made more generic.

It's impossible to verify failed results. All data seems to match. Is this a thing with numeric type conversion that is required?

expected: [{:date=>2016-05-01 00:00:00.000000000 +0000, :open=>9.75275391573086 (#BigDecimal:55d56ebe2dc0,'0.9...142E1',27(27)), :low=>9.89882377209142 (#<BigDecimal:55d56e963310,'0.9898823772 09142E1',27(27)>)}]
got: [{:date=>2016-05-01 00:00:00.000000000 +0000, :open=>9.75275391573086 (#BigDecimal:55d56ec04ab0,'0.9...142E1',27(27)), :low=>9.89882377209142 (#<BigDecimal:55d56ebf0100,'0.9898823772 09142E1',27(27)>)}]

(compared using ==)

There's no explanation what candlestick chart is, and no specifications about the expected output.

with an interval of 5 minute

Relative to what?

How do you "calculate" the balance?

All users have operations. This is not apparent from the description.

• Returning user_id twice using different column names is pointless
• This is no different from this kata