In supply chain modeling, the first big win is getting your baseline model up and running. You’ve cleaned the data, mapped the network, built out the assumptions, and validated the outputs. That alone can take weeks and sometimes even months. But once you present that first model to stakeholders, something predictable happens: the questions start pouring in.
They’re great questions, but answering them isn’t always quick or straightforward. Supply chain scenario planning becomes the bottleneck. Modelers are overwhelmed. And leadership wants results right away.
Contents:
In practice, this pattern appears across hundreds of projects. The initial model is built manually, typically using spreadsheets and semi-automated inputs. When new scenarios are needed, modelers copy the base file, tweak it manually, and rerun experiments from the original model.
This workflow is fragile, slow, and scales poorly
Often, just a few people understand the pipeline. The data isn’t reusable. There’s no version control. And every new question restarts the same painful process from data cleaning to manual Excel edits to model re-validation.
Taken together, these issues create what we call the first-model cliff. You’ve proven the value of simulation, but the workflow wasn’t designed for iteration.
Supply chain process automation: the obstacles beyond the first model
What if scenario development didn’t live in spreadsheets? What if data pipelines were modular, reusable, and validated by default? What if analysts and planners could test their own ideas without rebuilding the model every time?
These are the questions that led SimWell Consulting to the development of Pychain—an open-source framework for creating clean, repeatable, and modular anyLogistix models. Pychain, short for supply chain, transforms model creation into an engineering process with built-in support for supply chain process automation.
At its core, Pychain is a Python library that supports supply chain scenario planning and makes it scalable. It brings together:
Beyond building and validating inputs, Pychain also streamlines the processing of anyLogistix outputs and visualizing results.
In practice, the same pipeline that generates your model data can also handle post-run results, aggregate key metrics, prepare comparison datasets, and even feed them into dashboards. This closes the loop from data ingestion → modeling → results visualization without manual exports and helps improve supply chain process automation.
What is Kedro? This was the first stage of the project, presented at the previous anyLogistix Conference. We have a separate blog where we describe this powerful tool in detail. Make sure to check it out >>
Pychain helps you define structured pipelines that take in raw data from Excel, databases, Snowflake, or wherever and turn it into fully validated anyLogistix inputs.
No more guessing if your columns are named correctly. No more trial-and-error importing. Pychain uses Pandera to validate every table and column before import, so you know the input file is clean and anyLogistix-ready before you even open the model.
Example of Pandera validation rules in Pychain (click to enlarge)
At its core, Pychain uses a layered architecture adapted from Kedro’s data engineering framework. Each piece of input data flows through a clear set of stages:
Data layers of Pychain for supply chain process automation (click to enlarge)
This structure makes the pipeline both readable and modular, which helps with debugging, extension, and reuse across supply chain projects.
Pychain comes in two parts:
This separation means teams can get started quickly while still customizing for their own requirements.
For example, if your network optimization model requires 58 tables, Pychain can generate them, fill in headers, and even autofill data according to rules. This reduces repetitive work and speeds up supply chain scenario planning projects.
One of Pychain’s key innovations is the ExcelInputDataset, a custom Kedro dataset that programmatically creates the anyLogistix input file. It knows which tables are required for each experiment type and validates over 325 columns using Pandera.
Even if your pipeline hasn’t yet generated a specific table, Pychain will still write the correct sheet structure with headers, preventing anyLogistix import errors and allowing incremental model building.
Excel file created by Pychain with N of rows from Pipeline (click to enlarge)
Not every table needs to be built from scratch. anyLogistix includes several standardized tables, such as objective members, vehicle types, groups of customers, and others, that are common across projects and experiments. Pychain automates their creation, filling in headers and default values where needed.
That means less redundant work, faster project setup, and more time exploring scenarios and results instead of fixing Excel.
But SimWell Consulting didn’t stop there.
Once your baseline model is built, supply chain scenario planning often requires exploring capacity changes, cost shifts, demand growth, and factory closures. Instead of making modelers handle every request separately, SimWell built a layer on top of Pychain, a tool called Scenario Manager.
Scenario Manager data flow
Scenario Manager is exactly what it sounds like: a way for users (planners, analysts, and C-level managers) to define changes they want to simulate without editing the core model themselves.
They can fill in a simple spreadsheet or, for larger teams, use a web interface connected to anyLogistix via API. This makes it easy to crowdsource new scenarios without bottlenecking your modeling team.
For example:
You trigger the pipeline, and Pychain builds a new, validated scenario input for anyLogistix. It’s ready to import. Ready to run.
No manual rework. No rechecking field names. No broken experiments.
Need more details? Watch the full presentation by the SimWell team below – [Scenario Manager demo at 15:30].
The results show that teams cut their scenario development time by over 70%. But speed isn’t the only benefit.
It also makes onboarding new team members easier. With clear pipelines and reusable templates, newcomers can contribute without needing to reverse-engineer five versions of a model.
Tactical planning isn't just about building one good model. It’s about the ability to respond quickly, test alternatives, and update assumptions as reality shifts. That requires systems and not spreadsheets.
Ready to take the next step? Contact SimWell Consulting to learn more about Pychain and how it can transform your anyLogistix workflow into a scalable and repeatable decision-making system.
