At Nextmv, I’ve engaged with many organizations looking to automate and optimize their decisions for routing and other operations. Organizations live or die by making these decision workflows efficient. Margins are the name of the game and the biggest challenge to a growing operation.
Depending on where they are in their journey, we see the most successful companies scale by 1) automating and optimizing manual workflows, 2) leveraging off-the-shelf technology, and 3) using a single stack for decision automation.
When they do this, it’s amazing to see how quickly they can grow. Whether that growth comes from accomplishing more with your existing staff or having the ability to launch new delivery regions with confidence, decision science provides tremendous value to an organization. To illustrate this, I’ll walk through some examples. Let's dive in.
Automating manual workflows
There is a Y Combinator saying: “Do things that don’t scale.” We’ve certainly experienced this in our early operations at Nextmv and we see it all the time from our users.
In the early stages of a typical small delivery company, their operations probably look something like this: a single person or a small team manually assigns vehicles to routes using spreadsheets or a simple UI. It can take several hours to plan a day’s routes and could even require real-time adjustments when exceptions arise (as they often do).
Manual workflows make sense while you sort out the kinks. They even help you learn about and understand your users. But it only works for so long. New challenges appear. How do you add more delivery zones without scaling the team linearly? How can you make sense of the map when your volume grows by 3x, 5x, or 10x? What development is delayed because of the time spent on manual routing (and maintaining the UI needed to do so)?
We see this story play out all the time. We’re currently working with a local food delivery service based in the US who was originally handling hundreds to more than 1,000 orders a day using a manual workflow. Over the next year-plus, they have a goal to deliver several million orders. With Nextmv, they have been able to eliminate excess hours dedicated to manual planning and exception handling and continue driving toward their future goal.
Leveraging off-the-shelf technology
But what if you already hired a data scientist or operations research (OR) expert? (They’re rare, don't let go!! 🦄) It’s likely you’re looking to build and maintain in-house optimization tooling.
Our founding team lived this life. It can take several months of wading through academic papers, data translation cartwheels, and a heavy dose of integration love from your software team to ship that new dispatch algorithm to production. It usually means re-implementing a generic heuristic applied to your use case or translating business data into matrix math to pass to a traditional optimization solver (that's usually running on-prem).
This process works until something changes: quarterly business goals shift, real-time hyperlocal trends emerge, or SLAs get updated. What happens when your volume increases and causes your decisions to take 1 minute instead of the 30 seconds you planned for? Adapting or scaling in-house tooling to account for these changes often takes months and can’t benefit from new features in decision science. It’s far from ideal.
Your data scientists shouldn’t need wild integration help to ship updates and your developers should be able to leverage common optimization patterns to make critical decisions at scale. What matters is that their decision tooling easily accounts for new business logic while also being fast and nimble for real-time operational changes.
Take, for example, the engineering team at Milk Moovement, a Canadian dairy supply chain software company. They don’t have (or need) hours to find the optimal decision for how to reroute drivers if dairy volumes at one farm are higher than expected and exceed truck capacity. They need the best routing decision in the tight time frames they have to keep their operations going.
This is the power of Nextmv. We help customers make the best decisions in the time they have to solve their problem. And we make this possible by providing a decision stack that’s easy to use as is or customize as needed.
Using a single decision stack
A guiding principle for developing Nextmv is to create a product experience where all operational decisions look and feel the same. Whether that's a routing problem for vehicles, a scheduling problem for staff, or allocation problem for cloud resources — all decisions are codified and built, tested, and managed just like any other software. Your devs and data scientists can now cover a range of problems without context switching between tools.
As a result, Nextmv has broad applicability across organizations. We have a customer who provides a nation-wide delivery service in the US. They initially adopted Nextmv to solve batching problems. As they implemented and learned the stack, they quickly saw opportunities to apply Nextmv to several other use cases. They don’t see a need to procure another software tool — they can just use a single stack to solve and simulate their operational decisions. This saves time across the board, from procurement to adoption to integration.
We’ve reached the end of our run time
These are just a few ways we’ve seen customers scale and create value with Nextmv. We get really excited about the opportunities to improve organizational efficiency with decision automation and optimization. If you’d like to learn more, create a free Nextmv Cloud to get started or contact us to chat directly with our team.