Supply and demand optimization
Tributary
Tributary combines the forecasting abilities of frontier models with the possibilities that Manifold models, so a manufacturer is more resilient.
It runs on our foundation model of the supply network: the state of components and materials across it, what parts are made from, what depends on them, what can substitute for them, and where the risk sits.
What it lets a manufacturer see and decide
Tributary shows what a stop anywhere in a manufacturer’s network does to its own deliveries: when the shortage arrives, how deep it goes, and when output is back at plan. It shows this as a range across many futures, with the bad case marked.
Then the manufacturer decides: which planning rules to run for the cash they tie up, where added capacity shortens a recovery, and the latest week each response can start.
From a whisper market to an aggregator
Today the market runs on people passing emails to each other: requesting that things be made, requesting prices for them, requesting suppliers to make them, and, on the other side, people deciding whether they want to meet any of those requests. We call it a whisper market.
The market has to move from a whisper market to an aggregator: a system that brings supply, demand, and capacity together. We’ve seen it happen in financial markets and in software markets. It happens anywhere all the highways are cleared and numbers are allowed to meet each other in the middle, and decisions are made on the numbers.
Predictive and adaptive
This problem takes machine learning, whole factory optimization, and outcome data. We can now predict outcomes we don’t want, and outcomes we didn’t even know we needed to predict. We can adapt to those outcomes in simulation, ahead of time. That creates a loop of improvement in how we make supply and demand decisions.
It takes whole factory optimization because a factory that isn’t optimized doesn’t have a clear picture of what it needs, what it’s creating, or what it’s supplying. Combining these data sets, whole factory and outcome, and then connecting them to a marketplace, is supply and demand optimization.
The gap between a demand and the capacity to meet it
When a demand emerges in the marketplace, it usually starts very small. The factories that recognize a small demand as one that’s growing exponentially usually don’t have the capacity or the adaptability ready.
That creates a reactive delay between the marketplace deciding it needs something and manufacturers deciding it’s worth producing. The companies that build capacity for this demand ahead of time take on a level of risk that’s not sustainable. Preparing decisions in advance shortens that delay.
Where a manufacturer starts
The marketplace is fragmented today. A manufacturer can start with its own optimization inside it, by connecting the data it already has on its factory, its suppliers, and the people who buy from it. That’s the initial solution, and that’s Tributary.
Why we built it this way
A stop at one site reaches the customer weeks later and several tiers away, so we model the network: components, materials and sites, and what each one needs from the others. How bad a stop gets depends on things nobody knows on the day, so we play each disruption across many futures and look at the spread.
A plan that’s good on average can still miss the quarter, so we judge every result by its bad case. Planning rules are usually set once and by hand, so we search them against those futures and keep the ones with the smallest bad case for the cash they tie up.
A plan nobody can check doesn’t get run, so every action the chosen policy takes is listed with its cost. Some recoveries wait on a material or a qualification, and no planning rule shortens them. Tributary also ranks where a little more capacity pays. And events that are on their way go onto the same network, so each response has a deadline.