How to start a snack brand

A recipe is a set of instructions for a person. A formulation is a specification for a facility: every ingredient by weight and by percentage of the total batch, a stated process with times and temperatures, a target finished weight, and an acceptable range for each. The translation is not clerical. Doubling a home batch rarely behaves like the same product, and a hundredfold scale-up almost never does, because mixing, heat transfer, and moisture loss do not scale linearly.

Write the formulation before you approach any manufacturer, because it is the document the entire conversation runs on. It also forces decisions you have been making implicitly: which ingredients are functional and which are flavour, which have a substitute if a supplier fails, and which are the reason the product is worth buying. Those three categories behave very differently when a co-packer proposes a change.

The date on the package is a claim about a product you have not made yet, which is why it has to come from testing. Two properties drive it for most snacks: water activity, which governs whether microorganisms can grow, and oxidation, which governs whether fats go rancid and texture degrades. A dry, low water activity snack can be shelf stable for a long time and still become unpleasant well before it becomes unsafe.

Shelf life testing means storing the actual finished product in the actual final packaging under defined conditions and evaluating it at intervals, for safety and for quality separately. Packaging is part of the experiment rather than a detail: barrier properties, seal integrity, and whether you are flushing with nitrogen change the answer completely. A product tested in a laboratory pouch and sold in a cheaper film is an untested product.

Get a food scientist or a testing laboratory involved here. This is the step where founders most often substitute confidence for evidence, and it is also the step where being wrong is a recall rather than a bad review.

Cottage food laws let certain low risk foods be made in a home kitchen and sold under conditions that vary substantially by state: which products qualify, what sales channels are permitted, annual revenue caps, and what the label must say. They are a genuine on-ramp for baked goods, granolas, and confections, and they are also a ceiling. Many states restrict or prohibit interstate and wholesale sales, so a cottage food operation frequently cannot grow into the business you are picturing without moving production anyway.

A licensed shared or commissary kitchen removes most of those restrictions by putting you in inspected space, while leaving the labour with you. A co-packer runs your formulation on their equipment at volume, which is where the category ends up if it works. Co-packers quote in cases and runs, ask for a formulation rather than a recipe, and expect print-ready artwork against their own film or carton specification.

Facilities producing food for the US market operate under FDA preventive controls requirements, and an allergen control programme is part of that. If your product is free of an allergen, the facility handling it still has to be able to support the claim, which is a question you ask before you print anything.

A packaged food label in the US carries a defined set of elements: a statement of identity, the net quantity of contents, an ingredient list in descending order by weight, a Nutrition Facts panel in the current format, the name and place of business of the responsible party, and a declaration of the major food allergens. Sesame was added to that allergen list, which is a recent change worth verifying rather than working from memory.

Nutrition Facts values come from either laboratory analysis of the finished product or a validated database calculation from the formulation. Both are acceptable routes and both depend on the formulation being accurate, which is another reason the weights-and-percentages document matters. Serving size is not a free choice either: it follows reference amounts customarily consumed, which are published.

Claims are where enthusiasm becomes a compliance problem. High protein, low sugar, natural, keto, clean, and gluten free each carry their own definitions and evidence requirements, and gluten free in particular is a defined claim with a threshold rather than a description. If you cannot substantiate a claim, the cheapest fix is to delete it.

Snack unit economics live at the case level because that is how every quote arrives. Build the sheet with ingredients, film or carton, secondary packaging, the co-packer run charge amortised across the run, inbound freight, storage, and outbound shipping, and then look at what a single unit really costs at the run size you can actually fund.

Two effects surprise founders. Small runs carry the setup and changeover cost across few units, so the first run is the most expensive product you will ever make, and pricing off it makes the brand look unviable when it is not. And direct-to-consumer shipping on a light, bulky snack is dominated by dimensional weight rather than actual weight, so a large bag of something airy can ship worse than a dense one twice its mass.

Decide early whether the business is direct, wholesale, or both, because the price has to work in the channel you intend to grow. A wholesale route means a distributor and a retailer both take a margin from the same shelf price, and a direct price set without room for that has to be raised later in front of existing customers.

Dough starts while the snack is still a description. You say what it is, what makes it different, and who eats it, and it returns several drafts, each with a product design, a packaging concept, and a brand. You refine drafts in plain language and nothing commits until you pick one. Drafts come in two shapes: a catalog product a manufacturer in the network already makes, which is faster and cheaper, and a custom product that needs real development work.

Building the draft publishes a storefront on its own address. You set the price and Dough shows the unit cost and what each sale leaves you before you commit, so the number on the product page and the number in the cost sheet stay attached. The storefront can gather waitlist signups or pre-orders before any production run exists, with funds held in escrow and customers refunded if the launch threshold is not met. Sampling, production with vetted manufacturers, and fulfillment follow in the same account, with ads and analytics alongside.

Design and brand lock when the product is built, so refinement happens on drafts rather than after. You own the business fully and Dough takes no equity. Pricing is one plan at $29 per month plus a share of what you sell, with no setup fee. Because Dough runs a public MCP server, the same workflow can be driven from a chat client.