Blog / Downtime
Downtime · Guide

Blocked and Starved: Machine States Explained

OL By OEE Lab Editorial |Updated 2 November 2026 |6 min read
Short answer

A machine is starved when it is ready to run but nothing is arriving from upstream, and blocked when it is ready to run but has nowhere to send its output. In data, check the infeed and outfeed signals: empty infeed means starved, full outfeed means blocked, and if neither is true the stop starts at the machine itself.

The two states on one line

Picture a filler, a labeller and a case packer joined by conveyors. The labeller jams, and the two machines beside it stop for opposite reasons.

FillerLabellerPacker BlockedOwn faultStarved Outfeedfull Jam atthe machine Infeedempty
One fault, three machine states. Blocked sits upstream of the fault, starved sits downstream.

Blocked and starved machines have nothing wrong with them at that moment. They stopped because something else on the line forced them to.

That is why these states matter. Without them, a line report shows three machines down, and maintenance may walk to the wrong one.

How to tell them apart in data

You need two signals per machine as well as run and stop: one at the infeed and one at the outfeed.

Many machines already have these sensors, and the PLC may hold them as bits. PLC signals for OEE shows what to look for.

MachineInfeedOutfeedAlarmState
StoppedEmptyClearNoneStarved
StoppedHas productFullNoneBlocked
StoppedHas productClearActiveOwn fault
StoppedHas productClearNoneIdle: ask why
StoppedEmptyFullAnyCheck the first event
Check the alarm first: a machine in its own fault is not blocked or starved, even if a sensor agrees.

On older machines, an external sensor over each conveyor can provide the same two signals. Retrofit sensors for OEE covers the options, and machine state detection covers how run and stop get decided in the first place.

Read a whole line with one table

Here is the payoff. Lay out the minutes per state for each machine, in line order, and the constraint tends to show itself.

Example: a five-machine line over one 480-minute shift.

Machine, in line orderOwn faultStarvedBlocked
Depalletizer6 min0 min62 min
Filler14 min4 min48 min
Labeller52 min3 min9 min
Case packer9 min58 min6 min
Palletizer4 min66 min0 min
Hypothetical shift. Blocked runs high before the labeller and starved runs high after it.

The pattern points to the labeller. Machines before it are mostly blocked, machines after it are mostly starved, and it has little of either.

That machine is where an improvement project should start. Bottleneck analysis has the method for confirming it, including checks on speed that this table alone will not show.

This also tells you where a lost minute hurts. A minute lost on the depalletizer may cost little, since it spends 62 minutes blocked anyway, while a minute lost on the labeller is likely to reach the line's output.

What to do with each group

The table sorts your machines into three groups, and each one calls for different work.

  • Mostly blocked (depalletizer, filler). Leave their faults low on the list for now. Check that they have enough spare speed to refill the line, and use the V graph to see how speeds should step around the key machine.
  • Mostly own fault (labeller). Build a Pareto chart of its alarms and work down from the top.
  • Mostly starved (case packer, palletizer). Do not add speed here, since more speed only means they wait for product sooner.

Re-run the table after each fix, because the constraint can move. If the labeller improves, the blocked and starved zones will shift and may point at a new machine.

When the pattern is not clean

Real lines do not always give a neat split. Two machines with high own-fault time can create two zones, each with its own blocked and starved group.

A machine can also sit in the middle with few faults and still cause the pattern, because it runs too slowly. In that case the answer lies in its speed, and speed losses covers how to measure that.

Blocked and starved with no sensors

You can start with nothing but your eyes. Stand where you can see a machine and its two conveyors, then mark each stop with a letter.

Use S when the infeed is empty, B when the outfeed is backed up, and F when the machine itself is in fault. Add a rough time per stop, and one shift fills one row of the table above.

With a single observer, do one machine per shift. The table will be rough, yet it can still show which machine to look at first.

How blocked and starved show up in OEE

At machine level, a blocked or starved machine is not producing, so its availability falls. That is an honest number for that machine, as long as you can see the state behind it.

At line level, count the loss once, at the output or constraint. Line OEE vs machine OEE explains why adding up machine losses overstates what the line lost.

Who takes the blame when a stop travels along the line is a separate question. Upstream and downstream stops gives the rule.

Setting up the logic well

  • Add a time filter. A two-second gap between products is normal running. Start with a filter of a few seconds and tune it by watching the line.
  • Place sensors with care. An outfeed sensor right at the machine says "full" sooner than one a few metres down the conveyor.
  • Set a priority. Own fault wins over blocked, and blocked or starved wins over idle.
  • Keep the first event. When both signals are on, the state that came first is usually the true one.
  • Watch it live before you trust it. Stand at the line for 30 minutes and compare the state on screen with what you see.

Misreadings that send people to the wrong machine

  • Treating blocked time as a fault of that machine. It sends maintenance to a working machine.
  • Removing blocked and starved from machine downtime altogether. The minutes should stay visible with their own label.
  • Using one line-wide stop reason. You lose the map that shows where the trouble began.
  • Ignoring short states. Many brief starved events can mean an upstream machine with frequent micro-stops, so count them as well as adding up the time.
  • Forgetting that buffers change the picture. A long conveyor delays both states, so a blocked event may start well after the fault that caused it.

What to do first

Take one line and one shift, and fill in the table above by hand if you must. A stopwatch and a clipboard at each machine will show the pattern well enough to prove the idea.

Then ask where signals already exist and add the missing ones. Once the pattern shows the constraint, use how to reduce micro-stops if its losses are mostly short ones.

See it live in Fabrico

The table above only helps if the states are captured for each machine. Fabrico takes stops, speed and counts live from PLCs or IoT sensors, and shows OEE per line and shift with the causes operators tag at the line. See how Fabrico tracks stops.

What people ask next

What does starved mean in manufacturing?

A starved machine is ready to run, but no material is arriving at its infeed. The cause sits upstream, either in an earlier machine or in the material supply. If supply is the cause, see waiting for material downtime.

What does blocked mean in manufacturing?

A blocked machine is ready to run, but its outfeed is full and it cannot pass on what it makes. The cause sits downstream. Conveyors and buffers delay the moment when it starts.

Do blocked and starved count as downtime in OEE?

For machine-level OEE, yes, because the machine is not producing in planned time. For line-level OEE, count the loss once at the line output. Line OEE vs machine OEE shows the difference.

How do I find the bottleneck from blocked and starved data?

List machines in line order with minutes per state. The bottleneck is usually where blocked time ends and starved time begins. Confirm it with the checks in bottleneck analysis.

Can a machine be blocked and starved at the same time?

Both sensors can be on during a long upset, for example with an empty infeed and a full outfeed. Keep the state that appeared first, because later signals are often a result of it.

See OEE per line and shift on your own machines

Fabrico is a light MES for OEE. It takes stops, speed and counts from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, and operators tag the cause at the line.

Book a Fabrico demo

Related guides: Why OEE Drops After You Add a New Machine · OEE for Beverage Lines: Filler First, Then Why

← More OEE Lab guides · Upstream and downstream stops →

OEE essentials: OEE calculator · What is OEE · Six big losses · Availability · TEEP · Cost of downtime · OEE benchmark · Free OEE tracker · Free OEE software · How to track downtime · How to implement OEE · Light MES vs full MES