In the not too distant past, Carolyn Mooney, Nextmv CEO, indulged my request to perform what we termed the Neapolitan ice cream sandwich of demos: three different decision models, three different experiments, one presentation.
What resulted were a series of repeatable experiments using Nextmv that helped us understand the impact of making changes to decision models for vehicle routing, order fulfillment, and shift scheduling.
This post breaks down those three experiments in the context of a fictional Nextmv Farm Share Company. Check out the short video clips below to watch the magic unfold.
Routing: To add cold chain vehicles or not?
The Farm Share Company has added refrigerated dairy items into its product lineup. The company has a hybrid fleet of vehicles: some are cold chain ready (or have refrigeration) and some are not. How does investing in upgrading the entire fleet to be cold chain ready impact business KPIs?
In this video, Carolyn walks through the process of configuring a batch experiment, running it, and exploring the changes in model output. (She also gives a walkthrough on the anatomy of a decision app — IDs, descriptions, run history, changelog, versions, instances, etc.)
— Discovery! —
When the capability of the entire vehicle fleet increased to handle refrigerated items, we saw that the total time on road decreases and vehicle utilization increases. This fleet upgrade is looking like a positive change for the daily KPIs. Now it's possible to compare the upgrade cost and see how long it will take to recoup the investment.
Order fulfillment: To add an inventory cap or not?
Moving on from cold chain logistics! The Farm Share Company recently expanded its delivery zone beyond what their own fleet can service and now needs to account for third-party carriers and distribution centers (DCs). The order fulfillment decision model selects carriers and DCs to minimize total delivery and handling costs, but they discover that they’re sending too much produce to the DCs — their inventory capacity is being exceeded. This isn’t good because The Farm Share Company cares about the environment and minimizing food waste.
— Discovery! —
Because each DC can only handle a certain amount of inventory, it's necessary to push more orders to further DCs, meaning delivery cost per carrier likely increases. But the handling costs have stayed roughly the same. We can see how much the delivery costs increase and weigh that against the desire to be food waste conscious.
Shift assignment: To increase shift break time or not?
Our final optimization problem centers around the Farm Share Company's workforce: assigning workers to shifts. In this decision model, we’re looking to minimize the amount of overage hours while respecting worker availability and preferred times. In other words, as the decision app assigns workers to shifts, it’s looking to minimize the chances of having an on-shift worker with nothing to do. In this example, we’re looking to account for new labor laws that increase the break time between shifts while keeping the shift lengths the same. So what happens to our overages if we increase the time between shifts from 8 hours (which is currently the required break reflected in the production model) to 24 hours?
— Discovery! —
Overage went up for the staging model where we allow for 24 hours between shifts instead of the 8 hours currently modeled in production. This means we have more workers assigned to shifts than the demand for work to be completed. This serves as a good starting point for thinking through how to adjust our operations to better utilize our workforce.
To try out testing or not?
But of course! You can try out our testing features in Nextmv with a free trial. Make sure to check out the latest with acceptance tests and shadow tests. (More curated testing experiences are on the way!) Try it out, let us know what you think or what we can improve in our community forum, and contact us if you’d like to chat live.
