Skip to main content

Cookbook

Copy-paste starting points, drawn from the shared 44-query test corpus.

Churn (binary classification)​

PREDICT COUNT(transactions.*, 0, 30, days) = 0
FOR EACH customers.customer_id
WHERE COUNT(transactions.*, -90, 0, days) > 0

Spend / LTV slice (regression)​

PREDICT SUM(transactions.price, 0, 30) FOR EACH customers.customer_id

Recommendations (ranking)​

PREDICT LIST_DISTINCT(transactions.article_id, 0, 30) RANK TOP 12
FOR EACH customers.customer_id

Daily demand, 4 weeks out (forecasting)​

PREDICT SUM(usage.count, 0, 1, days) FORECAST 28 TIMEFRAMES
FOR EACH accounts.account_id

Specific entities​

PREDICT COUNT(orders.*, 0, 90, days) = 0 FOR users.user_id IN (42, 123)

Counterfactual​

PREDICT COUNT(orders.*, 0, 90, days) = 0 FOR users.user_id = 42
ASSUMING users.plan = 'premium'

Status prediction (string predicate)​

PREDICT LAST(loan.status, 0, 30) NOT LIKE '%DENIED' FOR EACH loan.id

Missing-attribute prediction (static target)​

PREDICT articles.description IS NULL FOR EACH articles.id

Population carve-outs​

PREDICT SUM(transactions.value, 15, 45, days) > 100
FOR EACH customers.customer_id
WHERE customers.location NOT IN ('ALASKA', 'HAWAII')