Outlier

A value far from the rest, such as one huge order. It may be real, a data error or a test account, and it can move an average a lot.

Formuladecide the rule before you look: e.g. above p99, or known test accounts

How we work it through

1
Find out what the extreme rows are: real, errors or tests.
2
Apply a rule set in advance, never one chosen after seeing the result.
3
Report the result with and without them.

On the practice dataset

The query we run on Meridian, Atterna’s practice marketplace, and what it returns.

SELECT COUNT(*) FILTER (WHERE items_total_eur > 300) AS big,
       AVG(items_total_eur) FILTER (WHERE items_total_eur <= 300)
FROM meridian.orders
WHERE status = 'paid';

Result76 orders above €300 (the largest €542.94); without them the mean falls from €60.92 to €58.34

The common trap

Deleting outliers without a rule decided in advance, or keeping test accounts that inflate totals.

When an AI answers

An AI assistant

“I removed outliers and AOV is now €58.34.”

What it didn’t know: It dropped real customers’ large orders without saying why.
Ask back: “Which rule removed them, and are those orders real?”

Paste a real AI answer into the checker — it writes these questions for you →