Traceability is the requirement that shapes everything. You need to know which batch of which ingredient went into which production run, which finished lot that produced, and where that lot went. If something is wrong with an ingredient, that chain is what lets you identify the affected products rather than withdrawing everything you have ever made. For a small producer this can begin as a disciplined spreadsheet with lot codes, and it needs to exist from the first production run rather than being introduced later.
Labelling is a regulatory matter rather than a design one. Ingredient declarations, allergen statements, net quantity, nutrition information where required, and your business details all have specified formats, and the rules differ depending on where you sell and how much you produce. Getting a label reviewed by somebody who knows the requirements before printing thousands is the cheapest version of this. Registered product identifiers are also needed if you intend to sell through retailers, and retrofitting them onto printed packaging is expensive.
Inventory in this category has a dimension others lack, which is that stock expires. Any system you use needs to handle dates and rotation rather than only counts, because a quantity on hand that includes product past its date is a number that misleads you twice. This is the main argument for using something built for food rather than general inventory software.
Production planning matters once you are making more than one thing, because ingredients are shared across products and lead times differ. Knowing what you can make from what you hold, and what to order when, is the difference between a production run that completes and one that stops halfway. A spreadsheet handles this at small scale provided somebody maintains it.
Selling through both your own channel and retailers means two different sets of requirements, and the wholesale side needs price lists, case quantities, and lead times you can actually hold. Businesses that grow through retail frequently discover their production capacity was the constraint rather than demand, and that is a planning problem worth modelling before accepting a large order.