Most small-business schedules are made on Sunday night from memory, a spreadsheet and a few text messages. The same four things go wrong: a shift nobody noticed was uncovered, someone scheduled on a day they cannot work, an avoidable overtime week, and the same two people getting all the good (or all the bad) shifts. Doing the steps in the right order fixes most of it.
For each day: start, end, how many people it takes. "Mon–Fri 6–2, two people, open" is a requirement; "Ana and Ben on opens" is already a decision. Keeping them apart is what lets you see the real total of hours you must cover.
Availability windows, the most hours they can or want to work, and the fewest you have promised. Days not on the list count as unavailable. If the weekly limits of everyone together are smaller than the hours to cover, no schedule works and it is better to learn that before the week starts.
A person must be free for the whole shift, cannot be in two places, needs enough rest since their last shift, and has a cap on days in a row. These are not preferences. The most constrained shifts (the ones only one or two people can work) go first, or the easy shifts will use up exactly the people the hard ones need.
Among the schedules that obey the rules, an overtime hour costs a premium, so keep people under the threshold when someone else is free. Giving every hour to the cheapest person is cheap on paper and loses staff, so a fair share of hours and a minimum for the people you have promised them belong in the score too.
Remove each person in turn and see which shifts go uncovered. Anyone whose absence leaves a shift open is who the week depends on: they need a backup, or a wider availability from someone else.
An invented café has seven people and 32 shifts to fill across the week, 235 hours in all, with a 10-hour rest rule, a 40-hour overtime line and everyone's own availability. Filling shifts in time order with the first free name on the list leaves 7 shifts open, 58 hours nobody is scheduled for. The roster builder, given the same people and rules, leaves 3 shifts open and tells you why for each: for example the Monday lunch has four people free, but all four are already at their weekly limit. It also reports that taking out any one of Ana, Dev or Gus leaves 40 more hours uncovered, so those three are the café's single points of failure. The fix is a decision for the owner: ask Fay or Cleo for wider availability, add an hour limit exception, or shorten a shift.
A clopening is a closing shift followed by an opening shift the next morning, leaving very little time to sleep. Some places limit it by law or policy with a minimum rest gap between shifts, often somewhere around 8 to 11 hours. Check what applies where you operate; the roster builder lets you set the gap and will not break it.
Count each person's hours across the whole pay week, not shift by shift, and give the marginal shift to someone who is still under the threshold even if they cost slightly more per hour. One overtime hour at time and a half costs more than an extra hour for anyone paid up to half as much again as the person already at the limit.
It depends on where you operate. Some cities and states have predictive-scheduling or fair-workweek rules with advance-notice periods and pay for late changes, usually aimed at larger retail, food-service and hospitality employers. This tool does not check them, so look up the rules for your location.