Atherya and the Bottleneck

How bottlenecks are managed today

Today, bottlenecks are managed in three ways. All three are wrong.

With machine-by-machine OEE. Dashboards full of percentages, one per machine, all waiting to improve. But raising the efficiency of a machine that is not the constraint does not produce one more part by the end of the day. It only creates more work in progress on the shop floor.

With nameplate capacity. Every planner, whether an ERP module or the shift supervisor's spreadsheet, multiplies catalogue cycle time by available hours. But a press's real capacity is not its nameplate. It is its observed cycle time on that compound, with its health today, its typical stops and its ongoing degradation. A plan built on nameplate capacity is a plan built around a machine that does not exist.

With instinct. People ‘know’ the bottleneck until the mix changes, a different order arrives, or a machine begins to slow down and the constraint moves unnoticed. The factory keeps protecting the wrong machine.

Then comes the most serious problem: maintenance and production do not speak. Planners do not know which machine is deteriorating. Maintenance does not know which machine matters most. The loudest machine gets attention, not the one whose stop costs the most.

Software holds the plan. Machines decide what is possible.

Conventional schedulers cannot plan against reality for one simple reason: they do not know the machines. Maintenance systems cannot reason about the bottleneck because they do not know the orders.

Atherya knows both. Everything starts here.

Finding the real production bottleneck in data

Atherya does not ask where your bottleneck is. It measures it.

It reconstructs the chain as it truly works. The same item code moves from mixer to presses, presses to ovens, ovens to finishing. Atherya reads when a campaign starts and ends upstream and downstream, deriving actual lead time and the real sequence of the chain. Not the one drawn on the layout: the one confirmed by data.

It reconstructs lines as they really operate. Two presses belong to the same line when they make the same products on the same shifts. Atherya also finds machines that alternate: when one runs, the other stops, often because one operator covers both stations. No organisation chart tells you that.

It measures who saturates their hours. For every machine family, it calculates how much of open plant time is actually occupied. The bottleneck is the resource with no free hours left, not the one that merely looks slow.

It is not fooled by codes. In ovens, for example, one ERP value can mean a different physical oven or merely a different program on the same oven. That detail separates capacity you truly have from capacity that exists only on paper. Atherya distinguishes them by seeing which run at the same time and which always alternate.

It measures the time the constraint wastes. The factory's most expensive hour is when the bottleneck is stopped or starved while the rest of the plant runs. Atherya counts it and measures the material buffer ahead of the constraint: whether it risks starvation or has room to breathe.

Schedule on effective capacity

At the centre is one concept: effective capacity.

For each machine and product, Atherya calculates what it can really produce: observed rate, multiplied by the availability that machine achieves in reality, multiplied by its health at that moment.

It separates two conditions every other system treats as identical:

If a number cannot be measured, Atherya does not invent it. A machine without measured cycles does not receive a ‘default’ capacity. It receives a written reason why the number is missing.

• 01A machine with a stable chronic defect that has always run a little differently. That is its character: capacity is not discounted, because doing so would punish it twice.
• 02A machine with ongoing degradation that is getting worse. That capacity is discounted, because tomorrow it will be lower than today.

Bottleneck prevention: protect the constraint before it stops

Once you know the constraint, maintenance changes meaning. You do not intervene on the machine making the most noise. You intervene where one stopped hour costs the whole factory.

Constraint degradation becomes the first news of the day. Each evening the shift supervisor receives a short list ordered by value: which machine is leaving its normal, which is running slower than usual on the same item, which had an abnormal mould change, which started late. Every item carries evidence, why it matters and what to do. The list itself is tested against history: if flagged machines do not stop more often than others, it is noise, and we know it.

Shift opening matters. Every cold restart wastes time on the machines least able to afford it. Atherya first measures whether cold truly matters, then proposes when each machine should start so it reaches operating conditions before the shift begins.

Separate press fault from mould fault. The same mould runs on different presses. If an item slows on every press that runs it, the mould is wearing. If one press slows across all its items, it is the press. A fault that looks like a machine problem can be the mould, and fixing the wrong thing releases no capacity.

Look upstream. When the mixer has an unusual day, Atherya checks whether presses leave their normal more often in the following days. It compares days in the same month, so summer heat moving everything together does not create false links.

Manage it: move work where capacity exists, with evidence

Who makes what, and where. Atherya recommends moving an item to another press only when that press has made it before, is genuinely faster with statistically solid evidence rather than a lucky average, has free hours while the plant is open, and is robust. A faster but fragile press is flagged, not loaded. When an important item runs on an unreliable machine while a dependable alternative has already made it, Atherya says so.

A shadow plan beside yours. Atherya calculates its production plan against effective capacity alongside the real plan, touching nothing. It compares both with the same speeds and queues, recording where it would have acted differently and why. It is the most honest way to learn whether a system deserves trust before giving it any.

Proposals, not orders. When the difference matters, Atherya proposes a move with all evidence attached. A person approves or rejects it. A rejected move is not proposed endlessly. A manually locked order stays where the person put it, and the plan works around it.

Autonomy only inside guardrails. Where the company chooses, the system may approve its safest proposals within explicit limits. Never on critical-priority orders. Never towards a machine at risk or under maintenance. Only after the same proposal has proved itself repeatedly, with your sign-off. Every decision stays signed, so it is always clear what a person decided and what the system decided. Atherya remains read-only on your systems: execution in the ERP stays with people.

Why Atherya is different

Others sell one piece of the problem.

A planner that does not know machine health. A maintenance system that does not know what each machine is worth. An OEE dashboard telling you to improve everything — meaning nothing.

Atherya combines what none can see alone: which machine is the constraint, what it can truly produce today, whether it is deteriorating, what reaches it from upstream, and where work can move without risk. Every number and proposal carries a written reason.

The truth in one line

You cannot manage a bottleneck you cannot locate, on capacity that is not real, with maintenance that does not know which machine matters most.

The bottleneck is not a planning problem or a maintenance problem. It is where the two meet. That is exactly where Atherya works.

Protect the constraint, and you protect the factory.