Planning Production Ratios in Satisfactory

Production planning in this game is one piece of arithmetic repeated at every stage: how many machines does a target rate need, and what does that demand upstream. Once you can do it for one step you can do it for a whole chain.

The one calculation

machines needed = target rate / recipe output rate

Both in items per minute. A recipe producing 20/min feeding a target of 90/min needs 90 / 20 = 4.5 machines.

The database on this site lists every recipe already converted to per-minute rates, because the game states them as per-craft quantities and converting by hand is where most planning errors start.

Fractional answers are the normal case

Ratios rarely divide evenly, and the instinct is to round up and let the extra machine back up. There is a better option.

Underclock the row to fit. Five machines at 90% produce what 4.5 at 100% produce, and because power scales by an exponent of 1.321929, five machines at 90% draw less power than 4.5 machines at full rate would.

That is the counterintuitive part worth internalising: underclocking is not a compromise, it is a power saving. A machine at 50% clock draws about 40% of its base power.

The alternative — round up and let the last machine sputter — works but wastes power and makes the line harder to reason about, because a row where one machine cycles irregularly looks like a fault.

Plan backwards, not forwards

The common mistake is starting at the miner. “This node gives 120/min, what can I make with it?” produces a factory shaped by what you happened to extract.

Start at the consumer instead. What do you actually need per minute? Work backwards through each stage, and the extraction rate falls out at the end. Then underclock the miner to match rather than running it flat out into a backed-up belt.

A worked example, targeting 30 Reinforced Iron Plate per minute:

• Reinforced Iron Plate: 5/min per Assembler, so 6 Assemblers
• Those need 180 Iron Plate and 360 Screws per minute
• Iron Plate: 20/min per Constructor, so 9 Constructors
• Screws: 40/min per Constructor, so 9 Constructors
• Iron Plate needs 270 Iron Ingot, Screws need 360 Iron Rod which needs 360 Iron Ingot
• Iron Ingot: 30/min per Smelter, so 21 Smelters, wanting 630 Iron Ore per minute

Six hundred and thirty ore per minute for thirty plates. That is the shape of this game: the visible machine is a small fraction of the build.

This is where alternate recipes earn their keep

Run that same example with Steel Screws instead of the base screw recipe, and the screw stage collapses from 9 Constructors to fewer than 2, with the iron rod chain behind it disappearing entirely.

That is why recipe choice matters more than machine count. A better recipe changes the whole tree below it, and the saving compounds at every stage.

Our comparator ranks every recipe for an item on raw ore per unit specifically so this decision can be made on evidence rather than on which one sounds better.

Belt capacity is a hard constraint

Every stage has a throughput ceiling, and exceeding it silently caps the whole line.

Belt marks carry 60, 120, 270, 480, 780 and 1200 per minute. A stage producing 630 ore per minute needs a Mk.5 belt at minimum, or parallel lines.

Check this before building the machines, because discovering it afterwards means rebuilding the routing rather than adding a machine.

When the answer is “move the factory”

If your calculation says 630 ore per minute and your nearest node cluster gives 240, the planning answer is not more miners. It is one of:

A better recipe, reducing the ore requirement rather than raising the supply.

Process at the source and ship the product, since fewer items travel than ore.

A second site, built on nodes that can actually feed the demand.

Node count is the ceiling on any factory. Machines are cheap; nodes are finite and fixed.

What about the planner sites?

Several community tools let you enter a target and get a full production tree with machine counts. They are good, and we are not trying to replace them.

We publish the pieces rather than a solver: every recipe at per-minute rates in the database, raw ore per unit in the comparator, and real power figures in the overclock calculator. That combination answers the specific questions — is this recipe better, what does this clock speed cost — that a tree diagram tends to gloss over.

For a full multi-stage tree, use a planner. For deciding which recipe should be in that tree, the comparator gives you a figure the planners generally do not.

Building in modules rather than in totals

A practical consequence of the arithmetic above: the numbers get unwieldy fast, and a factory planned as one enormous total is hard to build and harder to extend.

The alternative is to pick a module size that divides evenly and repeat it. If one Assembler making Reinforced Iron Plate needs 1.5 Constructors of plate and 1.5 of screws, then a module of two Assemblers, three plate Constructors and three screw Constructors divides cleanly with no fractions anywhere.

Build that module once, get it right, blueprint it, and stamp it as many times as your supply allows. The whole factory inherits the ratios you worked out once, and expanding means adding another module rather than recalculating.

This is also why the Blueprint Designer’s 32-metre limit is less restrictive than it looks: a clean production module usually fits, and the module is the useful unit anyway.

Sanity-checking a finished line

Two checks catch most planning errors after the fact.

Is every machine running continuously? A machine that cycles irregularly is undersupplied. Compare what the belt delivers against what the row consumes; if supply is below demand the manifold will starve its far end permanently.

Is anything backed up? A full belt behind a machine means the stage downstream cannot take what you are producing, so you built more capacity than the next stage can use. That is wasted power and wasted materials.

Between them, those two states describe every ratio mistake. Everything running steadily with nothing backed up means the arithmetic was right.

Common mistakes

Planning forwards from the miner. You get a factory shaped by your extraction rather than by what you need.

Rounding machine counts up. Underclocking the row costs less power and runs steadily.

Converting rates by hand. The game states per-craft quantities; the arithmetic is where errors enter.

Ignoring belt marks until the machines are built. Throughput ceilings cap a line silently.

Treating machine count as the constraint. Nodes are the ceiling, and better recipes raise it without building anything.