The platform your teamsand your agents run decisions through.

One working model of how your business behaves, built from the data your systems already produce. Simulate a move against it and you get the outcome, the knock-on and the cost before you commit to anything.

One example, a manufacturing and distribution business.

The recording below follows a single change through that operation, step by step. It is one illustration of the three capabilities below rather than a definition of the product: the same questions are asked in pricing, portfolio and revenue work.

01

Ask what a change would do, in plain words

Your operation is already on screen. Point at the site under pressure and type the move you are weighing up. The simulation starts from there.

02

It gathers what it needs to simulate

It pulls the moves open to you and the plan you already committed to, so the prediction runs on your operation rather than an industry average.

03

It checks the move is even possible

How busy the line already is, how much of the plan is actually being made, and whether anything on the ground would stop this move working at all.

04

Every prediction quantified

What the move does to your service level, how likely it is to help at all, what it costs and what it earns back, over a period you set.

05

It predicts the knock-on, day by day

Where the site you are protecting ends up against plan, what the move does to everyone downstream of it, and where you land at the end.

06

The cost of indecision

The same move simulated against no action at all: the day it pays for itself, what it earns, what it costs and what you keep.

Ergodic platform demo: a working model of a manufacturing and distribution business. An analyst types a change in plain language, adding overtime at a plant that is struggling, against the live picture of the operation. The model finds what it needs on its own, the options available and the plan already committed to, checks how busy the line is and how much of that plan is being made, then plays the change forward: what it does to service, how likely it is to help, what it costs and earns back, what happens to every site downstream, and the return against doing nothing, including the day it pays for itself.

What is true, why it changed, and what might happen if you act.

One model of how your business behaves, and three questions you can put to it. Atlas holds what is true right now. Root Cause Analysis works out why something changed. Equinox plays a proposed action forward to show what might happen if you take it. Nothing here is specific to one industry.

Your businessas a modelATLASWhat is true nowRCAWhy it movedEQUINOXWhat an action would doTHE FEEDBACK LOOPWhat actually happenedTHREE QUESTIONS · ONE MODEL · EVERY OUTCOME FED BACK

The three capabilities, on one incident.

A supplier is running late and a major customer is about to miss their orders. Here is the same problem seen three ways: what is true now, why it is happening, and what to do about it. Open any marker to see what the platform is doing at that point.

Nothing here is drawn by hand. Your orders, shipments and stock positions build the picture, and it updates itself as new records arrive.

What you have

Live values, not last night's extract

Every supplier, route and site carries its real numbers. Supplier B takes 18 days in practice, against the 14 days the contract promises.

Supplier B · observed lead time 18d · 62 open POs
Marker 1 of 3

Illustrative figures. For a decision worked end to end, with the console and the scenario comparison, see how EWMs work.

Pointed at the decisions that repeat and commit.

Four decision types run through everything below: a forecast, a price, a launch, and the report on why a number moved. The Use Cases page shows the same four grouped by the team that owns them.

Forecasting

What is coming, and how wrong could we be

NOW

Demand, cover and cash rolled forward with the spread around them, so a plan can be stress-tested rather than argued.

Asks the model: what happens next, and how sure are we?
Pricing

What a price move does to everything else

+5%+7%+10%MARGIN · CHURN

Each candidate price played forward through demand, capacity and the accounts it touches, with the trade-off shown rather than assumed.

Asks the model: which move wins on balance, and what does it cost?
Launch

What a launch takes from the rest of the book

OVERLAP · WRITE-OFF

Cannibalisation, pre-build and the stock you're left holding, modelled before the commitment date rather than counted after it.

Asks the model: what does this new thing do to the old one?
Reporting

Why the number moved, with the evidence

KPIBINDING DRIVER

A monthly review where each movement traces to a cause and its evidence, instead of a pack assembled by hand the week before.

Asks the model: why did this happen, and what is still binding?

Starting cut of four decision types. The final set will be agreed with the wider team.

Evaluated on four public benchmarks.

Ergodic evaluated its approach on four public benchmarks that anyone can inspect. Three run a decision out over days or months rather than a single turn; the fourth measures how accurately a fault is traced to its cause. They measure agent performance, not prediction accuracy. Here are the headline numbers, with the method and caveats on each.

Incident response
+53pts

Live incidents repaired by smaller models, 47% to 100%

Root cause
+11pts

Right service named first, 79.4% to 90.6%

A trading month
3.3×

Net assets at the end of the month, 11 of 13 paired runs

Running a business
+28%

Cash after 91 simulated days, across 180 runs

Read the benchmarks → How agents use the platform →

See it run on your own operation.

Bring one decision that's expensive to get wrong. We'll build the model around it and show you the options, the knock-on effects and the workings in a session with your team.

Talk To An Expert