New integration: Bring your Hexaly decision model to Nextmv

Solve optimization problems with Hexaly? The Nextmv Hexaly integration provides a new way to efficiently run, test, and manage Hexaly decision models with Nextmv’s DecisionOps tools and infrastructure.

We’d like to officially welcome the Nextmv Hexaly Integration to the Nextmv platform! 🎉🎉 This new integration lets practitioners pair all the goodness of Hexaly with Nextmv in order to build more decision models, not more decision tools.  

We’re driven by freeing ops researchers, decision scientists, and data scientists to build more and better decision models that generate real business value. Optimization is a powerful tool with broad relevance. Making it easier to adopt optimization by providing the tools and infrastructure that help accelerate their operational impact is why we built Nextmv. And it’s why we’re excited to announce the Nextmv Hexaly Integration as the newest addition to our integration ecosystem. 

Hexaly is a global optimization solver of a new kind, known for its speed and scalability in solving problems such as routing, scheduling, packing, picking, clustering, matching, assignment, location, and more. Hexaly is commercial software that is free for teaching and basic research. The Nextmv Hexaly Integration allows Hexaly users to add an existing or net new Hexaly model as a decision app in Nextmv. A Nextmv decision app combines the power of deployed software with the niceties of a notebook-style experience by providing all the pieces you need to run, test, and manage your Hexaly decision model efficiently in Nextmv. Let’s take a closer look! 

Hexaly model testing, validation, and iteration with Nextmv

The Nextmv platform comes with a suite of decision model testing and experimentation tools, which give teams more confidence in safely rolling out model changes. Nextmv’s tools for scenario tests, for example, can help Hexaly modelers easily perform “what if” simulations on model scalability (e.g., what if order volume increases 10x?), solve duration (e.g., when do I see diminishing returns on solution value the longer I run?), and more.

Example scenario test setup for Hexaly decision model in Nextmv.

Acceptance tests can help Hexaly users validate model changes such as confirming that removing a vehicle compatibility constraint doesn’t negatively impact business requirements.

Example acceptance test for Hexaly decision model in Nextmv.

Plus, Nextmv makes it easy to manage test input sets for different types of volumes, regions, and more.

Example of testing input sets for Hexaly decision model in Nextmv.

Nextmv also provides tools to apply online testing to Hexaly models via the integration. These tools are useful for model experimentation using real-time, production data. Shadow testing, for example, helps you preview and understand a Hexaly model’s performance using production data under production conditions — without the production impact. (Shadow testing is also a way to have a backup model at the ready!)

When you’re ready for a new Hexaly model to start making real-world decisions, but still want to compare it to an existing production model, Nextmv makes it easy to perform model switchback or A/B tests

Automatically pick the best plan from multiple runs of your Hexaly model 

The ensemble runs feature in Nextmv automates the execution of multiple decision model runs, summarizes the output statistics, and picks the best configuration based on your criteria. (This is especially useful with non-deterministic optimization using heuristic techniques.) For example, you may want to pick the vehicle route plan with the shortest time on road that also has the lowest number of unassigned stops based on results from 15 runs of your Hexaly routing model. With ensembling, you can execute those runs in parallel and have the best solution (based on your definitions) returned to you automatically.  

Model management, observability, CI/CD integration 

A Nextmv decision app comes with a set of unique API endpoints for running, managing, and experimenting on your custom Hexaly decision model. This simplifies integrating Hexaly models with standard Git and CI/CD workflows, ensuring model observability by monitoring run history, and swapping model versions with the click of a button.

Examples of CI/CD integration via GitHub, decision model version management, and decision model run history.

Cross-team collaboration, stakeholder buy-in and optimization explainability

“Why did the model make this plan?” “Can the current model handle expected demand in the future?” “Are we sure this model update will improve our outcomes?” All valid questions that can be explored and answered through a shared platform view in Nextmv. Invite collaborators with defined permissions to review run history, experiment results, and model changes.

Example of team settings in Nextmv with different roles and permissions.

To get started, simply create a free Nextmv account to gain access to the platform and sample decision apps. To run your Hexaly models with Nextmv, send us a message and we’ll work with you to activate the right plan for your account. Check out the sample workflow below 👇👇👇

We’re excited to expand the Nextmv integration ecosystem to Hexaly. We believe in making optimization more accessible to new and existing practitioners by providing tooling, infrastructure, and practices that accelerate time to value from optimization. Let us know what you think in our community forum or reach out to us directly.

Video by:
No items found.