SQL PERCENTILE_DISC
The discrete percentile — always returns an actual value that exists in the data, never an interpolation.
What & Why
PERCENTILE_DISC selects the first observed value whose cumulative position reaches the requested percentile. It is appropriate when the result must be a value that actually appears in the data rather than an interpolated value.
See How It Works
BUSINESS QUESTION
Marketing wants each channel's observed median and 90th-percentile campaign spend.
| id | name | channel | spend | start_date | end_date | status | target_segment |
|---|---|---|---|---|---|---|---|
| 1 | Spring Launch | google_ads | 55000.00 | 2024-01-15 | 2024-03-31 | active | smb |
| 2 | Retention Webinar | 45000.00 | 2024-02-10 | 2024-04-15 | active | enterprise | |
| 3 | Finance Retargeting | 50000.00 | 2024-03-12 | 2024-05-31 | active | enterprise | |
| 4 | Enterprise Search | google_ads | 60000.00 | 2024-04-01 | 2024-06-30 | active | enterprise |
EXAMPLE QUERY
SELECT
channel,
PERCENTILE_DISC(0.5) WITHIN GROUP (ORDER BY spend) AS median_spend,
PERCENTILE_DISC(0.9) WITHIN GROUP (ORDER BY spend) AS p90_spend
FROM marketing.campaigns
GROUP BY channel
ORDER BY channel;RESULT — observed campaign percentiles
| channel | median_spend | p90_spend |
|---|---|---|
| 45000.00 | 45000.00 | |
| google_ads | 55000.00 | 60000.00 |
| 50000.00 | 50000.00 |
Discrete percentiles always return a spend value that exists in the channel.
This lesson's practice is part of Pro.
Sign up free to try it on a real business scenario