Choose z or t from assumptions

For independent observations with known population SD σ, z=(mean-μ0)/(σ/√n). With unknown σ, use t=(mean-μ0)/(s/√n), df=n-1. Exact small-sample t inference assumes a normal population; larger-sample approximations need a sufficiently well-behaved sampling distribution. The rule n<30 is a teaching shortcut, not the definition of a t test. Check outliers, design and distribution, not just n.