Every machine in Satisfactory runs on one honest number: items per minute. Learn to read it and plan around it, and your factories stop starving, stop backing up, and start doing exactly what you designed them to do. Ignore it and you get the classic mid-game mess: a beautiful building full of machines, half of them idle, the other half choking on a belt that cannot keep up.
This is the mental model that fixes that. No mods, no calculators required, just the three ideas that separate a factory that flows from one that fights you: rates, distribution, and clock speed.
Rates: everything is items per minute
Every producer and consumer has a fixed rate at its default 100% clock. A Smelter turns 30 iron ore per minute into 30 iron ingots per minute. A Constructor making iron plates eats 30 ingots per minute and outputs 20 plates per minute. Chain those and the ratio falls out on its own: one Smelter perfectly feeds one plate Constructor, with the plate line running slower than the ingot line by design.
The skill is not memorizing every recipe, it is thinking in the unit. Ask two questions about any machine:
- How much does it consume per minute?
- How much does it produce per minute?
Line the output rate of one stage up against the input rate of the next, and the machine count writes itself. If a downstream machine wants 45 of something per minute and each upstream machine makes 30, you need one and a half machines' worth of supply, so two producers with a little headroom, or a pair of underclocked ones hitting the number exactly. Plan the whole chain in items per minute first, place buildings second, and you will almost never have to tear a factory down.
Your starter factory probably grew by feel. The moment you start planning in rates is the moment your builds stop hitting invisible ceilings.
Belts are a hard ceiling, not a detail
A ratio on paper means nothing if the belt cannot carry it. Belt tiers cap throughput, and hitting that cap is one of the most common reasons a "correct" factory still starves downstream:
| Belt tier | Max throughput (items/min) |
|---|---|
| Mk.1 | 60 |
| Mk.2 | 120 |
| Mk.3 | 270 |
| Mk.4 | 480 |
| Mk.5 | 780 |
If a single line needs to move 900 items per minute, no single belt can do it, full stop. You either split the load across parallel belts or upgrade the belt tier. Always sanity-check your busiest line against the belt you are actually using before you blame the machines. The same logic applies to pipes for fluids, where headlift and pipe tier quietly cap what looks like a fine setup on paper.
Manifold vs balancer: the distribution decision
Once you know your rates, you have to get materials to a row of machines. There are two schools, and picking the right one per situation is most of the art.
The manifold runs one belt down a line of machines, tapping each one off a splitter in series. It is dead simple to build and it self-corrects: over time every machine's input buffer fills and the whole line reaches steady state, drawing exactly what it needs. The one catch is startup. The machines nearest the source fill first, so a fresh manifold looks unbalanced for a few minutes until the buffers along the line saturate. Give it time and it evens out.
The load balancer uses splitters and mergers to divide the input into equal shares, so every machine gets its exact fraction immediately, no warm-up. It is precise and satisfying, but the splitter-merger networks get fiddly fast, and a small miscount throws the whole thing off.
Here is the honest recommendation:
- Use a manifold for almost everything. It is simpler, it scales by just extending the belt and dropping another machine on the end, and steady state is steady state. For the vast majority of production lines, the startup delay does not matter.
- Reach for a balancer when timing matters, such as feeding a set of machines that must all run the instant the belt goes live, or when you are splitting a precious, rate-limited input and cannot afford a machine hogging more than its share.
Nine times out of ten, build the manifold, walk away, and come back to a balanced line. Save the splitter math for the cases that genuinely need it.
Clock speed: the power bill nobody reads
Power Shards let you overclock a machine up to 250%, and it is tempting to slap them everywhere for "more output." The catch is that power does not scale linearly with clock speed. Push a machine's clock up and its power draw climbs faster than its output, roughly to the power of about 1.3. So a machine running at 250% is producing 2.5x the goods but costing well more than 2.5x the power.
That one fact drives two very useful habits:
- Underclock to save power. Running two machines at 50% produces the same output as one machine at 100%, but draws noticeably less total power, because you are riding the efficient part of the curve. When power is tight and space is not, spreading the load down is a free efficiency gain.
- Overclock only where space or a scarce node forces your hand. The right time to burn shards on a 250% machine is when you physically cannot fit more machines, or when a pure resource node can feed far more than one default miner can pull. In those spots the power premium is worth it. Everywhere else, add machines instead.
The same thinking applies at the source: a miner on a pure node outproduces one on an impure node for the same footprint, so where you place extraction changes your ratios before a single belt is laid. Match your miner tier and clock to the node purity and you waste neither power nor potential.
A five-minute planning routine
Put it together and planning any new factory becomes a short, repeatable checklist:
- Start from the target. Decide how many of the final item per minute you want.
- Work backwards. For each stage, compute the input rate it demands, then the machine count that supplies it. Repeat down to raw ore.
- Check the belts. Find your heaviest line and confirm your belt tier can carry it. Upgrade or split if not.
- Choose distribution. Manifold by default, balancer only where timing or a scarce input demands precision.
- Set clocks last. Leave machines at 100% unless space or a rich node pushes you to overclock, or a power crunch rewards underclocking.
Do that on paper (or in your head) before you place a single foundation, and the build goes down clean the first time.
Where to go next
Ratios are the grammar of Satisfactory. Once they click, the whole game opens up, because every problem becomes a number you can solve instead of a mess you have to untangle. Pair this with a solid power backbone from our power and logistics guide so your overclocked lines never brown out, and skim the current update notes for any recipe or belt changes that shift the numbers above. Plan in items per minute, and the pioneer's endless to-do list finally starts to shrink.
