Step 1: Recall what stratification actually involves. Stratified sampling means partitioning the population of \(N\) units into \(L\) non-overlapping strata, and then drawing samples separately from each stratum.
Step 2: Every stage of building the strata brings its own decision: which variable to stratify on (criterion), how many groups to cut the population into (number of strata), and exactly where the cut points fall along that variable (points of demarcation). All three of these vanish if there is no stratification at all.
Step 3: The overall sample size \(n\), however, is decided independently of whether the population is stratified or not. Even a plain simple random sample needs its size fixed using cost and variance considerations. So this decision is not born out of stratification.
Therefore the option that does not belong to the stratification-specific problem list is $\text{Fixing the sample size}$.
\[\boxed{\text{Fixing the sample size}}\]