Factorio Bottleneck Guide: Fix the Factory Before You Make It Bigger
Find what is slowing your Factorio factory: trace power, ore, belts, circuits, oil, trains, and science before adding more machines.
Start with the machine that is waiting
Find what is slowing your Factorio factory: trace power, ore, belts, circuits, oil, trains, and science before adding more machines.
A stalled factory can look extraordinarily busy. Belts move, furnaces glow, trains arrive, and the one item you need still never reaches its destination. The useful question is not how many machines you own. It is which missing input, blocked output, or shared service prevents the next useful operation. This guide works backward from that question through a conventional Nauvis factory, rather than asking you to replace your entire save with somebody else's blueprint.
The distinction matters for returning players in 2026. Wube announced the 2.1 experimental release on June 26, 2026, warning that designs and mods could need adjustment and that upgraded saves could not be downgraded to 2.0. That announcement is not evidence that every player now runs the same release. Check your installed version and enabled mods before copying a ratio. The examples below assume normal-quality equipment, ordinary base recipes, and no productivity or speed bonuses unless stated otherwise.
This is an analysis of the developer's documentation and official wiki, not a timed playtest. The ten official press-kit screenshots are archival gameplay illustrations, with files dated 2020 in the archive; they show systems and layouts, not the current interface or a newly benchmarked save. A screenshot can make a factory legible, but it cannot establish its sustained output.
Read a rate, not a mountain of stored plates
Open the production statistics before changing the line. On the default PC controls, P opens the production window, which tracks production and consumption over selectable periods. Filter to the item that is actually short. A graph dominated by iron plates can hide a small but important shortage of circuits. Use a longer interval to separate sustained behavior from the temporary burst that follows emptying a chest; the wiki specifically warns that short sampling windows can distort rates.
Then walk backward from the stalled consumer. If its output is full, it does not need more ingredients. If one ingredient is absent, follow that ingredient's route. If several unrelated machines slow together, inspect their shared power supply before rebuilding each recipe. An idle assembler is a symptom, not a diagnosis. Write down the missing item and the nearest upstream machine that should supply it; this gives the next inspection a precise destination.
A storage buffer is useful, but its contents answer a different question from throughput. As an illustrative calculation, a chest holding 600 items could cover a deficit of 10 items per second for only 60 seconds. That is arithmetic, not a measured result from the pictured factory. A seemingly recovered line may simply be spending inventory accumulated while its consumers were stopped.
Mining: separate output speed from remaining reserves
Mining drills connect the map to the factory. The official wiki gives an unmodified electric mining drill a rate of 0.5 iron ore per second. Under those conditions, 30 continuously operating drills provide 15 ore per second, matching a normal yellow belt's total unstacked capacity. This is a planning baseline, not a promise: depleted coverage, blocked outputs, inadequate power, productivity research, and modules change the result you observe.
Inspect the drill's covered resource area, not just the size of the ore patch on the map. A drill can stop because its own reachable tiles are exhausted while ore remains elsewhere. Conversely, working drills may pause because the receiving belt cannot accept more. Adding drills helps the first kind of shortage only when they reach available ore; it does not solve the second kind of transport constraint.
Treat resource longevity and delivery speed separately. Productivity can increase output per resource removed, while reduced resource drain concerns how quickly the deposit is consumed. For a basic starter mine, the practical habit is simpler: leave an accessible route for another output belt and plan the next deposit before the existing supply becomes erratic. Expanding an outpost also creates new power, transport, and defense obligations. The ore count alone is not the cost of expansion.
Smelting: count the conversion, not the rows
A normal stone furnace produces iron or copper plates at 0.3125 per second under continuous operation. Forty-eight such furnaces therefore produce 15 plates per second. The corresponding normal steel or electric furnace rate is 0.625, so 24 reach the same nominal output. Those calculations describe a fully supplied line without modifiers; they do not mean that a row with that many buildings is necessarily filling its belt.
Separate the four checks: ore arrives, fuel arrives where required, plates can leave, and the equipment has its necessary energy. Stone and steel furnaces use fuel; electric furnaces move that dependency onto the electrical network. Replacing a fuel-fed line with electric machinery can remove one belt problem while exposing a power problem elsewhere. A full output lane is also different from a full belt, because each belt has two lanes.
Steel illustrates why visual symmetry misleads. The wiki lists normal stone-furnace steel output at 0.0625 per second, one fifth of its iron-plate output rate. Giving steel the same footprint as iron does not give the same number of finished items. Decide which product rate the downstream recipe needs, then build the conversion capacity for that rate. Keep the arithmetic close to the specific recipe instead of assuming every furnace row is interchangeable.
Belts and inserters: repair the narrowest handoff
For ordinary unstacked items, yellow, red, and blue belts move up to 15, 30, and 45 items per second across both lanes. A red belt does not create twice as much ore; it can carry twice as much if the source provides it and the destination accepts it. Before upgrading an entire route, look for the point where a dense queue becomes a thin stream. A short lower-tier segment or a constrained transfer can limit the rest of an expensive line.
Inserters add another boundary. They have a pickup side and a drop side, can be limited by available power, and will not place an item into a full or incompatible destination. Many production buildings are intentionally not filled to their maximum inventory by inserters. An arm waiting beside a machine with some ingredients inside can therefore be behaving normally, rather than proving that its control logic has failed.
Diagnose one connection at a time: does the correct item reach the pickup tile, can the arm reach it, and can the target accept it? Only after those checks should a faster arm be the presumed answer. The objective is continuous useful work at the destination, not maximum movement everywhere. A belt that stops because all required products are stocked can represent a healthy factory, not a fault.
Circuits: the small recipe that magnifies a copper shortage
Electronic circuits are an early intermediate that feeds many later systems, from inserters and production buildings to advanced circuits and modules. Their standard recipe uses three copper cables and one iron plate per circuit, while the ordinary cable recipe makes two cables from one copper plate. That makes cable production a particularly useful place to distinguish recipe counts from physical items moving along a belt.
With equal-tier assemblers and no modifiers, the familiar ratio is three cable assemblers for two electronic-circuit assemblers. It is a matched-recipe relationship, not a universal rule for mixed machines. Changing crafting speeds, introducing productivity, or using an expansion-specific recipe changes the calculation. The assembling machine 2, for example, has normal crafting speed 0.75, so a recipe's displayed base time is not automatically the time that machine takes to complete it.
Direct insertion from a cable assembler into a circuit assembler can avoid carrying the expanded cable volume along a long belt. It also makes a shortage easier to inspect: you can see the plate supply, cable production, and circuit consumer together. If the cable machine waits for copper, another circuit assembler adds demand without fixing supply. If circuits back up, look downstream before assuming the circuit block needs to grow.
A main bus buys clarity, not unlimited capacity
A main bus places frequently used ingredients along a shared corridor, with production branching away from it. The official tutorial presents it as an organizational choice, not a required winning layout. It uses more space and more belts than a tightly packed arrangement. Its strongest benefit for a learning player is that the route from source to consumer stays recognizable after another production block is added.
A wide corridor can nevertheless conceal insufficient supply. Splitting one incoming stream across four belts does not multiply the production feeding it. When those belts stop and gradually fill, they may look impressive while consumers are idle. Start the consumers again and the shortage returns. Reserve room for future lanes if useful, but distinguish that reserved capacity from output you already have.
The same tutorial notes why copper cable is often made locally: it takes more belt space than the plates from which it is made. That does not make a cable bus forbidden; it makes the transport cost a conscious decision. For a repair, trace the branch that starves and identify what earlier branches consume. Reorganizing every line at once can erase the evidence you needed. Extend a working pattern only after checking its actual input budget.
Oil: a missing product can begin with a full tank
Oil processing breaks the intuitive rule that more input always means more output. Advanced processing produces heavy oil, light oil, and petroleum gas together. If one output cannot leave the refinery because its destination is full, the refinery can stop producing all of them. A petroleum shortage can therefore coexist with crude oil waiting at the inlet and excess heavy oil at the other end.
Read the selected recipe and the fluid connections in alt-mode. Follow every output, including the one you currently consider unimportant. A tank provides room while you investigate, but storage is not a permanent balancing strategy if production continuously exceeds use. Cracking converts heavier fractions toward lighter ones; it should be considered alongside the factory's need to retain heavy oil for other recipes, rather than treated as permission to consume every last unit.
This is a good place for a small, understandable control system instead of a complete imported refinery blueprint. A threshold-based circuit can make a cracking operation conditional on available stock, with the threshold chosen for your factory rather than copied as a magic number. First establish that the pipes, recipe, water, and power are correct. Logic attached to a physically disconnected system cannot rescue it. Keep the first fix visible enough that you can explain why it worked.
Power: fix the service that every line shares
The electrical network distributes available energy to connected consumers. If demand exceeds supply, many unrelated parts of the factory can lose effective speed together. That is why electricity deserves an early check, not a final glance after replacing belts. The network view lets you compare demand and supply; generators running below their potential can also be normal when consumers do not need more power.
At normal quality, a boiler produces up to 60 steam units per second and a steam engine consumes up to 30 per second for 900 kilowatts. One fully supplied boiler can therefore feed two such engines, with 1.8 megawatts of combined generating capacity. Those are equipment conditions, not a guarantee of delivered power. Fuel, water, steam connections, and connection to the intended electrical network all still matter.
Follow the power plant's own supply chain. A coal belt shared with an expanding production line can become a weak point; an impressive row of engines without steam is only installed capacity. Solar power introduces a day-night issue, and accumulators store surplus energy rather than generating it. Look at the troublesome period, not only the moment you opened the panel. The repair should cover the factory's operating cycle, including new consumers that came online after the previous power expansion.
Trains: the delivery is a round trip, not a locomotive
Rail moves resources between distant parts of a factory, but the locomotive is only one component. A useful route includes a loading process, a reachable destination, unloading capacity, and a return path. A train arriving full does not establish that it can sustain the receiving factory. Inspect whether its cargo is actually emptying and whether its wait conditions match the intended operation.
Signals divide the network into blocks. Chain signals are useful before a crossing because they can prevent entry when the train cannot leave through its selected route, reducing the chance that a waiting train blocks other traffic. This does not make every junction safe by itself. Leave an exit section long enough for the train that will wait there, and verify the direction and placement of the signals. A copied intersection designed for shorter trains can behave badly when your longer train extends back into it.
Train stops can also impose limits on how many trains reserve them as a destination. If the interface says the destination is full, inspect those limits and the available stops rather than immediately adding another locomotive. For an illustrative delivery budget, a 2,000-item shipment every 100 seconds averages 20 items per second before other constraints. The gaps between deliveries still need buffering. Average capacity and smooth delivery are related, but not identical.
Robots and modules change the question, not the need to diagnose
Logistic robots provide flexible item movement between compatible chests inside connected logistic coverage. They still need charging. If bots queue at roboports, adding more bots can add more competition for the same service. Check whether the provider and requester are in the same network and whether the route has adequate charging capacity before treating the robot count as the only measure of throughput.
Construction coverage and logistic coverage are different areas. A chest that looks close to a roboport is not necessarily inside its item-delivery network. Start with a small provider-to-requester route that you can inspect, then expand. For sustained high-volume transport, compare the entire system's behavior rather than assuming a flying item is inherently faster or better than a belt.
Modules introduce another set of trade-offs. Speed modules improve crafting speed while increasing energy demand; productivity changes output per input; efficiency focuses on energy use. A faster machine can simply reveal that its inserter or ingredient supply was already near its limit. Recalculate the neighboring stages after an upgrade, and inspect the electrical consequences. The best module placement is the one that solves your current constraint at an acceptable cost, not the one that makes a single tooltip display the largest number.
Defense belongs in the production plan
On Nauvis, pollution connects production to attacks. The pollution cloud can reach enemy nests and trigger pressure on the factory, while evolution is affected by pollution produced rather than only the cloud absorbed at nests. These are different mechanisms. Making the visible cloud smaller is not the same as resetting enemy progression, and a quieter moment is not proof that a new outpost can be left unprotected.
The practical implication is to inspect the expansion route before the emergency. A mine that produces ore but cannot keep its defenses supplied is not a dependable input. Ammunition, repair materials, electricity where required, and a way for the player to reach the problem belong to the same logistical decision. The production statistics include pollution and losses, which can help distinguish ordinary growth from a recurring interruption that is consuming replacement resources.
This is not an argument that every player must choose the same combat settings. It is an argument for matching the factory plan to the map actually being played. In a cooperative save, give the outpost's supply and repair responsibilities a clear owner or a clear shared routine. When everyone assumes somebody else is maintaining the perimeter, the first evidence of a missing delivery may be the loss of the delivery route itself.
Science is where the whole chain becomes visible
Labs consume the science-pack types required by the selected research. If one required type is absent, having abundant stock of the others does not substitute for it. That makes the laboratory area a useful final diagnostic point: choose the missing pack, inspect its production, then follow its scarce ingredient upstream. More laboratories help only when the factory can support the additional consumption.
Research costs are expressed in units with specified packs per unit and a time requirement. A technology requiring 100 units of one red and one green pack needs 100 of each, not 100 packs split between the two colors. Likewise, a production target should name the pack types and conditions instead of treating a large total item count as a meaningful science rate.
Finish a repair by observing the destination under demand. Did the missing pack arrive, did research resume, and did the fix hold after stored material was used? Then save a small blueprint of the working block if it will help later. Blueprints preserve layouts and place building ghosts; they do not prove that another location has the same supplies. Factorio becomes easier to reason about when expansion follows an understood result. Build the next factory section because the current one works, not because a larger one might conceal why it stopped.
Comments
Post a Comment