Forty specific things worth being told about, with the threshold, the person who should receive it, and the case for not setting it at all.
Most store teams set alerts the way they set phone notifications: everything on, until it is unbearable, then everything off. Both states cost the same thing, which is that nobody finds out about the one that mattered.
Each recipe below is written to be argued with. It gives a threshold with the reasoning attached, says who should receive it and who should not, and names the action in the imperative. Where an alert is usually a bad idea, the page says so and points at the digest instead.
The test running through all of them is the one in the five-minute rule: if a competent person cannot take a useful action within five minutes of seeing it, it is not an alert.
A fixed number for every product is why this alert gets ignored. Set it from days of cover instead.
The expensive version of a stockout is the one where you are still paying to send traffic to it.
The order is placed, the stock is not there, and the clock on the customer's expectation has already started.
A negative number means you sold something you did not have. Correcting the number does not fix that.
A supplier who is going to be late usually knows a week beforehand and will tell you if asked.
A flagged order is a decision, not an emergency. The alert only helps if it reaches someone allowed to make it.
Nothing is wrong, nobody is blocked, and the order simply sits there until the customer asks.
A parcel that moved yesterday is fine even if it shipped a week ago. One that has not moved since Tuesday is not.
There is a deadline, the evidence is easier to gather now than later, and nobody sees the notification email.
The point is not to catch anyone. It is to make written decision limits mean something.
An overdue alert on a board where half the items are overdue is a description of the board, not a signal.
Grants get celebrated and removals get forgotten, which is how a store accumulates access nobody remembers giving.
Absences are the hardest thing to alert on, because nothing arrives to trigger anything.
Demoting things to a daily summary is what makes the remaining alerts worth reading.