Book a demo
Menu

Planning optimization for chemicals

Know what every planning decision is worth.

Mizahn models your chemical network as it runs: feeds, units, co-product splits, tanks, logistics, and customers. It plans across realistic ranges of demand, prices, yields, and outages, and tells you what to run, where, when, and for whom, with the margin behind each call.

Book a demoGet the technical brief

Fig. 01One cracker, five products

Illustrative

One steam cracker, five linked co-products (illustrative)Three candidate feeds, ethane, propane, and naphtha, each with a cost, enter one steam cracker task. Its output splits into five co-products: ethylene, propylene, crude C4, pygas, and hydrogen with fuel gas. Each co-product has a yield range rather than a single value, and a price. Decision markers sit on the feed slate, the cracker severity, and the route for each product. The objective underneath is margin: price times yield times rate, minus feed, energy, and freight.FEEDSTASKCO-PRODUCTSYIELDSPRICEFEED SLATEETHANE$PROPANE$NAPHTHA$STEAMCRACKERSEVERITY−+ROUTE EACH PRODUCTETHYLENE$PROPYLENE$CRUDE C4$PYGAS$H2 + FUEL GAS$OBJECTIVEmaxmargin= Σ price × yield × rate − feed − energy − freight
  • Decisions Mizahn makes: feed slate, rate and severity, where each product goes
  • Priced in dollars: feed costs and the value of every co-product
  • Planned across ranges: yields, prices, and demand are not single numbers
A single task with linked outputs. Feed choice and severity move every co-product yield at once, so Mizahn plans and prices the whole slate together.
  • Priced in dollars
  • Planned under uncertainty
  • MILP and MINLP
  • Co-products built in
  • Your actual assets

01The problem

Why chemical plans end up back in a spreadsheet.

Most planning software was built for discrete manufacturing. Ford doesn't make half a Mustang when it produces an F-150. Chemical plants don't work that way. Crack ethane and you make ethylene, but you also make propylene, butadiene, hydrogen, and fuel gas, whether you want them or not, and each one has its own market.

  1. 01

    Built for one product per recipe.

    MRP-based planning tools assume a recipe makes one product. A cracker makes five, in ratios set by feed and severity. Co-products, recycles, and blends become manual adjustments, and the plan drifts away from the plant.

  2. 02

    The real plan lives in a spreadsheet.

    When the system cannot represent the plant, planners rebuild the plant in a spreadsheet. The spreadsheet becomes the real plan, and the expensive system becomes a reporting layer.

  3. 03

    Every team optimizes its own number.

    Supply chain is measured on service, manufacturing on rates and run lengths, finance on working capital. Nobody can say what one target costs the others, so trade-offs get settled by negotiation instead of margin.

  4. 04

    Averages hide the risk.

    A plan built on average demand, average yields, and average prices looks fine on paper. The first week reality moves, the gap shows up as a missed order, a distressed sale, or an expensive buy-in.

  5. 05

    Reports, not decisions.

    Dashboards show what happened. Rules-based planners find a plan that works. Feasible is not the same as optimal, and planners are still left to work out what to do.

  6. The technical brief goes deeper: how to represent a co-product network, price its trade-offs, and plan it under uncertainty.

    Get the technical brief

02How Mizahn plans

Three ideas behind every plan.

01Economics

Every decision priced in dollars.

Mizahn values each plan by margin, not by volume or service-level proxies. Feed, energy, freight, inventory carrying, contract penalties, and product prices sit in one objective, so a trade-off between teams becomes one comparison instead of a negotiation. For the constraints that bind, Mizahn shows what relaxing each one is worth: one more rail car, a debottleneck, a turnaround moved by a week.

  • Margin, not proxies
  • The cost of every target
  • The value of each constraint

02De-averaging

Plans that hold when the forecast doesn't.

Averages hide risk. Mizahn plans across realistic ranges of demand, feedstock prices, yields, and outages, built from your existing forecasts, your history, and your planners' judgment. You see which decisions hold across the range, which ones depend on a single forecast, and what it costs to protect against the bad cases.

  • Ranges, not point estimates
  • Demand, prices, yields, outages
  • Protection priced in dollars

03Prescriptive

Decisions, not dashboards.

Mizahn does not stop at reporting what happened. It uses mathematical programming (MILP, and MINLP where the chemistry is nonlinear) to produce the decisions themselves: what to run, where, when, and for whom. Each one comes with what was traded off and why.

  • What to run, where, when, for whom
  • Mixed-integer optimization
  • Trade-offs spelled out

What a decision looks like.

Not a heat map to interpret. A specific instruction, the margin behind it, the constraints that bind, and the point at which the answer would change.

Planners can see why the plan says what it says, and what it would take to change it.

How Mizahn optimizes

Decision

Illustrative

run
cracker on ethane-propane mix, high severity, weeks n to n+3
allocate
propylene to contract customers first, spot capped at q
binding
rail cars at site A, week n+1
value of relaxing
λ, $ per added rail car per week
holds across
the demand and price ranges in the scenario set
changes if
naphtha falls below its break-even price p*

03Example decisions

The decisions Mizahn makes.

Each one is optimized for margin and planned across the uncertainty that matters for it.

See every use case

04Built for process manufacturing

Your plant, modeled as it runs.

Co-products are native.
A task can consume several feeds and make several products at once, in linked ratios. Mizahn plans all of them, including where each one goes and what it is worth. Recipes, recycles, and blends are part of the model, not workarounds.
Your assets, not a black box.
Units, reactors, tanks, rail cars, and pipelines are modeled as themselves, with their real limits and connections. Planners and engineers can see why the model recommends what it does.
Modern software engineering.
Mizahn is built with modern software practices, so the model grows with your business. Add a unit, a product, or a site by extending the model, not by waiting for a vendor release or starting a re-platforming project.

05How it works

From your plant to a priced decision.

  1. Step 1

    Model your network.

    We build a model of your actual assets with your team: feeds, units and their co-product yields, tanks, logistics, and customers. It starts from data you already maintain, including your existing demand forecast.

  2. Step 2

    Describe what is uncertain.

    Ranges for demand, feedstock and product prices, yields, and unit availability, drawn from your history and your planners' judgment.

  3. Step 3

    Optimize for margin.

    Mizahn solves for the plan that earns the most across those ranges, inside every constraint in the model.

  4. Step 4

    Act on the decisions.

    What to run, where, when, and for whom, with the margin behind each decision and the value of relaxing each constraint.

How it works in detail

06The team

Built by people who have run these plans.

Mizahn is built by process engineers and operations researchers who have planned, scheduled, and optimized production inside major chemical producers. [TODO: approve naming Dow and DuPont]

It comes from years of hands-on planning work across process-industry clients. The platform is in production today as the planning system for a major pharma company's advanced therapies business, where it replaced a legacy planning system.

About the team
Background
Process manufacturing and operations research
In production
Planning system for a major pharma company's advanced therapies business
Built from
Years of hands-on planning work across process-industry clients

Next step

See Mizahn on a network like yours.

Book a working session with the team. Bring a decision you are working on now: a feed slate, a campaign plan, a turnaround date. Or start with the technical brief.

Book a demoGet the technical brief