Box plot

The box plot widget summarizes the distribution of a metric rather than a single number. Each box shows five statistics — minimum, first quartile (Q1), median, third quartile (Q3), and maximum — so you can see spread, skew, and typical range at a glance.

Use it for:

• Session-duration distribution (per platform, per version)
• Spread of in-app purchase amounts
• Any metric where the median and variance matter more than the total

---

Query requirements

Return one row per category with a label column plus five numeric columns in this order and meaning: minimum, Q1, median, Q3, maximum. Alias them clearly as min, q1, median, q3, max — the widget matches columns by these names and falls back to column order otherwise.

Use ClickHouse quantile() for the quartiles.

Session duration per platform (full SQL)

SELECT
  platform,
  min(d)              AS min,
  quantile(0.25)(d)   AS q1,
  quantile(0.5)(d)    AS median,
  quantile(0.75)(d)   AS q3,
  quantile(0.95)(d)   AS max
FROM (
  SELECT platform, session_id,
         dateDiff('minute', min(timestamp), max(timestamp)) AS d
  FROM {table}
  WHERE name = 'app_started'
  GROUP BY platform, session_id
)
GROUP BY platform

Result shape:

platform	min	q1	median	q3	max
steam	0	4	11	24	65
ios	0	3	8	18	52

Tip

Real-world maxima are often distorted by idle sessions. Using quantile(0.95) as the upper whisker (instead of max()) keeps the boxes readable by trimming extreme outliers.

---

Single distribution

For one overall box, return a constant label:

Select / full SQL: replace platform with 'Sessions' AS label and drop GROUP BY platform (keep the inner per-session computation).

---

Common examples

Session duration spread per platform — see the full SQL above.

Purchase amount distribution

SELECT
  'Purchases'                   AS label,
  min(toFloat64(value.amount))            AS min,
  quantile(0.25)(toFloat64(value.amount)) AS q1,
  quantile(0.5)(toFloat64(value.amount))  AS median,
  quantile(0.75)(toFloat64(value.amount)) AS q3,
  quantile(0.95)(toFloat64(value.amount)) AS max
FROM {table}
WHERE name = 'purchase'