Joost van der Vlies: What Tech Leaders Should Know About Software-Defined Logistics

Disruptions are commonplace in the supply chain and logistics industries. Thankfully, technologies such as Software-Defined Logistics (SDL) offer game-changing solutions to obstacles in the supply chain. We speak with Joost van der Vlies, CTO and Head of Architecture at PostNL, on the benefits and challenges of implementing SDL, as well as important insights on cloud technology in the logistics space.  

SDL is an emerging approach for EDGE computing and PostNL is one of the pioneers in this area. How do you define the use of SDL for enterprises? What benefits does it bring?

SDL is all about using data and algorithms to steer the supply chain in all its aspects, from forecasting, planning, execution, monitoring, communication, and making real-time decisions automatically. For example, our network setup before determined the physical flow of a logistical item (e.g., parcel), now it is the digital twin of that item that determines the physical flow through our third-party networks. The digital twin contains not only the metadata of the parcel and the order but also customer and operator preferences which can be updated in real-time. For example, deciding on the sorting belt to change the operator from home delivery to retail delivery as the consumer updated their delivery preferences, or to change the operator from bicycle delivery to truck delivery as the item was much heavier than communicated. This means SDL is about sense, deciding, and responding, which makes logistics much more flexible and dynamic. Interestingly, this also creates a lot of new data, which can be leveraged in ways not thought of before. 

For PostNL, how are you effectively utilizing SDL as part of your cloud strategy? What is the framework and how can CTOs apply it to their organization?

Our cloud strategy is a multi-cloud strategy comprising SaaS, PaaS, and IaaS service providers, and a strong connectivity layer that also includes Edge environments. SDL is part of a more digitized business and cloud is the de facto delivery model for digital business. Within our cloud strategy, the emphasis is on cloud-native component-based application architectures, which can automatically scale depending on the logistical volume and can take part in the sense and response patterns that SDL requires. We train our machine learning models in the cloud and deploy them where decisions are made, that can be both in the cloud or on the edge. As response time and throughput are essential factors, we use global tier 1 internet service providers that provide abundant capacity, maximum uptime, and truly global coverage (for our international business), and private network partners where necessary. 

Of course, with any emerging technology, there are challenges and obstacles. Currently, what are the main challenges that tech leaders need to be aware of with SDL?

One of the challenges is that not all existing applications have been designed to operate in a real-time use case, so temporary measures might be necessary as well as a structural re-architecture. Here’s another example — when using machine learning models, it can have a more complex deployment model having an AI platform develop and re-train the model, and have it embedded in an algorithm in or near the business application it is used. And with SDL events can occur from a multitude of actors, which need to be handled in a highly scalable rule engine and using a single source of truth state machine of logistical items. Tech leaders should also be aware of the impact of the business operations on the people working thereof which their work will be impacted. Business and IT should jointly work on SDL and have a change management process from the start. 

How did PostNL overcome these pitfalls? What can other CTOs learn from your approach in tackling challenges?

Regarding algorithms, in the past years, we moved from data science hypothesis projects to the development of algorithms with learning models for use in production. That is only possible in a multi-disciplinary approach combining data scientists, data engineers, and the DevOps teams where these algorithms will run or with which it will be integrated. We were not afraid of taking a high-profile initial case and started working on this, learned from it, and eventually earned a computable award in 2020 with an algorithm that predicts when a parcel will really arrive. This is the same for the real-time data case. It requires a multi-disciplinary approach, time, and capacity for innovation, as it has a lot of consequences not all immediately known from the start. 

While cloud adoption is gaining momentum, there is still hesitancy among enterprises to fully adopt it. What should the approach be for CTOs to encourage technologies such as cloud and SDL within the organization?

Technology is not an island. Technology supports businesses to become successful. The processes of our customers and our own are getting more and more digital, and increasingly we do business with applications and machines instead of human interaction. Yes, an API is a technical way of accessing data and functionality, but in essence, it is a 100% digitized business service. Together with high volumes, the increasing number of digitized actors in our ecosystem, and the increasing flexibility our customers are asking in the e-commerce domain, cloud and SDL are essential capabilities for digital business. 

Finally, what advice can you give to other organizations that are starting to invest in cloud technology? What are the common mistakes that CTOs should avoid when making their transition into the cloud?

Firstly, cloud is not an infrastructure play. It is a full-stack play and includes, or starts with, an application strategy. Rehosting only will not provide true benefits. Understanding business drivers and the requirements for the applications in the future will be input for decisions to buy, consume or make. It also influences decisions to retire, replace, or re-architect those applications, which has direct consequences on the cloud strategy and roadmap. 

Next, skills around networking, storage, and high-performance computing are important and still very relevant when moving to cloud. You should continue having these skills onboard to avoid problems in the long run.  

In addition, the term multi-cloud is used a lot in the industry, though it is much more than using two or more public clouds. For us, any service a partner provides through the Internet is a cloud. This multi-cloud has to be managed from an overall functional, technical, and multi-supplier perspective. Lastly, when starting from a pure on-premises environment, the current IT department setup will probably not be aligned to cloud. Therefore, setting up a cloud competence centre is crucial.  

*The answers have been edited for length and clarity.