By: Tobias Morville & Andrea Nalupa
Part of our transportation and logistics thought leadership series
This article tackles research and expert insights around the following topics:
- Objectively understanding the potential of machine learning for fleet management and misconceptions around it
- Critical factors to consider for investments made towards machine learning integration
- Getting started with machine learning
According to Deloitte, among the automotive sector’s industries, fleet management is making its rounds as one of the more profitable ventures in the transportation market today. In manifestation, the market is currently valued at 5938.6 billion USD and is expected to grow to 7500.8 billion USD by 2023 across the globe. However, the revenues and rewards are tantamount to the impending and existing challenges within this line of work. As operations typically encompass services and processes covering the entire life cycle of up to millions of vehicles at a time (from financing and purchasing to service optimization and contract termination), firms often struggle to effectively manage their fleets – leading them into continuous pursuit of auspicious management mechanisms, with machine learning emerging as an increasingly popular option according to McKinsey.
The fleet management life cycle according to Deloitte
As enterprises turn to modern technologies in the hopes of addressing issues on efficiency, productivity, and most importantly, supply and demand, Sanchit Tiwari, a practice expert at McKinsey, discloses the benefits of integrating machine learning in fleet management that catapulted it into popularity. In his research, Tiwari identifies machine learning’s claim to fame – promising results from reducing maintenance and fuel costs to optimizing routes, driver performance, and overall productivity. However, Monstarlab’s expert insights uncover that these promises are not instant, nor delivered in each attempt. In fact, our machine learning expert and head, Tobias Morville, has identified that roughly more than half of integration attempts have ultimately failed – unveiling a timely and critical perspective on the importance of strategic compatibility and expectation alignment.
|“In a survey of more than 2500 executives, 7/10 report little or no gain from investment in AI”
MIT Sloan Management Review and Boston Consulting Group (BCG) Artificial Intelligence Global Executive Study and Research Report
In this objective undertaking towards understanding machine learning, we delve into the technology’s true potential in fleet management optimization, the determinants to the probability of it becoming either a valuable asset or added weight, and jumpstarting its sensible integration.
Machine Learning as a Vehicle for Success
Sought-after for its capability to learn complex patterns from massive amounts of data that older systems and the human mind have difficulty or are unable to comprehend and process, machine learning and its exceptional computing power transforms how companies perceive and maximize patterns and data in fleet management.
By using historical data and math-based solutions in training algorithms, machine learning empowers and improves many aspects of fleet management. From price optimization and fleet allocation during opportunities with more demands (to take full advantage of both fleet capacity and income), to route optimization and predictive maintenance (to reduce costs), here are some ways machine learning technologies can give organizations in the industry an edge in their daily operations.
|With time-series prediction, machine learning is able to provide important projections and figures on anticipated demand for different situations, on expected transport speeds, and on average maintenance gaps, that ultimately help fleet managers answer questions and make adjustments on balancing supply and demand, fleet allocation, and price optimization, among others.
|Reducing travel time and fuel costs, machine learning also provides stakeholders the efficiency and affordability that conventional methods of route planning struggle to offer. With an advanced Geocoding algorithm ensuring the most optimal clusters paired with real-time traffic analysis, and adjusted calculations able to take into account seasonal volume variations, drivers and consumers are not only able to navigate better through traffic, they are also provided travel cost and time estimations that prove more accurate.
|By analyzing records and patterns, machine learning offers businesses actionable insights on personnel performance including idling, driver behavior, job completion time, and driving speeds – allowing business owners to make necessary adjustments in operation workflows and protocols, improve customer service procedures, and safeguard regulation compliance.
|Equipped with predictive analytics and real-time data processing, businesses are also able to effectively manage resources and utilities with machine learning. Accounting patterns on risks, maintenance and repair needs, conducive working conditions, and vehicle specifications, machine learning technologies enable convenient and optimal vehicle monitoring, licensing, distribution, use, maintenance, issue management, regulation compliance, and even vehicle remarketing.
|Combined with rule-based automation and online sensory technologies, machine learning also presents stakeholders a more time-sensitive response to conflict. In this application, damage analysis, damage control, triggered emergency responses, report creation, insurance process optimization, and prompted authority alerts are all made possible at unparalleled speeds.
Factors to Consider
In hindsight, the aforementioned applications imply significantly enjoyable benefits. From cost reduction and risk minimisation to boosting productivity and increasing revenue, machine learning offers appealing possibilities to companies in the field; so much so that a number of large and successful companies such as Uber, Lyft, and Amazon have built their businesses on such premises. Rewarding as it may seem, however, such cases are not the norm. Only about 30% of machine learning integration initiatives have yielded expected positive results according to MIT Sloan Management Review and Boston Consulting Group. Before placing investments in the technology, consideration of the following factors is strongly urged to evaluate whether or not machine learning is beneficial and adequate for your business.
Outsource digital agencies may claim that machine learning is convenient and cost-effective. It can be, but cost-efficiency and convenience do not translate to its development’s affordability or accessibility. In fact, integrating or building it around a system can induce a significant strain on resources and normally requires a hefty amount of capital, time, labor, and search for qualified experts according to Morville.
|“When looking at a modern machine learning setup, the code base (the number of lines of code that runs the solution) consists of 95% code that supports the 5% machine learning code. ”
Tobias Morville, Head of Machine Learning, Monstarlab Denmark
To offer transparency, Morville explains machine learning integrations require a lot of work around the machine learning code which in fact, accounts for only 5% of the work. This 1-20 relation reveals a substantial overhead in resources that companies often completely miss when investing in ML projects. In fact, Web FX’s most recent digital marketing research reveals that the average-scaled project’s consultancy price can range anywhere from five to seven thousand USD.
Building this or integrating it into an existing platform or system also takes a significant amount of time and effort at variations dependent on the scale and urgency of each project. As Tobias also reveals the scarcity of high-level professionals, the search for practitioners qualified to implement at various scales, that have a solid background in statistics, mathematics, and machine learning, and that understand the overhead also adds workload and difficulty to this process.
All that being said, an in-depth assessment of the capacity of your resources as well as willingness to invest them must be executed before going any further. Otherwise, the risks of failure, overspending and lack of funding become more prominent.
Another important factor to take into consideration is your intention for the investment. Although machine learning is gaining traction in the industry in recent years and more organizations are attempting to deploy it, it is not necessarily a requirement, nor indicative of your business being at par with developing standards. As a matter of fact, our experts reveal that machine learning has long been in the game and is not entirely modern as it is derived from math-based solutions familiar to the market.
|“Machine Learning is not entirely a new technology. It is actually derived from existing statistical solutions ”
Tobias Morville, Head of Machine Learning, Monstarlab Denmark
Getting past what our experts believe is a hype for machine learning and evaluating or identifying intended use cases is critical in determining whether the technology fits into your strategy, budget, and overall direction. If the commitment to machine learning is baseless in terms of actual needs or is only meant for a small-scale and short-term use case, machine learning can turn out to be excessive considering the costs.
Taking into account the amount of resources exhausted and the relevance of use cases for machine learning, it is important to know and examine your options to maximize an investment. Although machine learning is exceptional at expeditious computations, scanning available alternatives will prove worthwhile.
Long-term and regular use for larger companies like Uber that employ machine learning in day-to-day operations are more compatible for machine learning investments. Morville explains that smaller enterprises, however, given the increasing availability of more affordable products with integrated machine learning capabilities, may find such ready-to-use products more cost-efficient at their scale.
Machine learning, as discussed, is useful for many industry-specific applications, but there are equally numerous practical options that offer equivalent functionality and efficacy that may be better-suited for your targets and budget.
Integrating machine learning, based on the 3/10 ROI statistic, is substantially risky. Therefore, expectations should be dealt with with caution.
In terms of returns, machine learning’s ROI is a slow build and is not guaranteed. Depending on the complexity of the use case and the amount of data processed, seeing expected results will take time to materialize. Preparedness and acknowledgement of the gap between placing an investment and getting your money’s worth is a must for staking machine learning sustainably and responsibly.
Before turning your interest into an objective or a plan, considering the aforementioned factors in exploring machine learning will ground your decisions, set your expectations, and allow your organization to understand and potentially exploit the technology more intensively. Pushing forward, informed decision-making makes all the difference.
Inspect the machinery
Bearing in mind the previously mentioned points to examine, determining your organization’s readiness for machine learning, as well as the technology’s role and importance moving forward ensures the value of the investment for your organization. A thoroughly conducted evaluation of directly involved needs and challenges, strategic pre-adoption simulation and risk analysis, as well as a comprehensive demand forecast will help set the tone of your machine learning journey.
Gear up and get your team together
Once the position of machine learning in your digital strategy and structure is figured out, reviewing and optimizing the alignment, potency, and compatibility of your supporting tech stack and talent pool to your use case and purpose for integration further increases your chances of success. If short-manned or in need of outsourced expertise, partnering with a qualified, experienced digital agency enables machine learning to do its job and come up with the best possible outcomes too.
Test drive your plans before changing lanes
Administering a well-supervised pilot program offers more concrete and actionable insights that will allow your company to get a feel of the new dynamics and allow you to make adjustments without serious repercussions and with enough time to modify implementation as needed. Piloting the utilization of machine learning on your chosen use case will help set realistic expectations on long-term gains and reveal whether or not the current approach requires adjustments in process flow, machine learning codes, and overall dynamics.
On the question of machine learning’s potential, its applications and benefits evince multitudes of possibilities. The technology’s value in fleet management, however, have been found unique to each organization, with potential and benefits varying depending on organization-intrinsic factors. In conclusion, with extensive introspection and mindful implementation, firms can more accurately measure and leverage machine learning’s capabilities within their chosen use case.
Upon consolidation of research and expert insights, this report also finds that despite recent discouraging machine learning-attributed ROI statistics, there are simple and achievable methods that can mitigate related risks and allow businesses to successfully adapt the technology – principally documented in the general recommendations made in this report.
Learn more about technology in the transportation and logistics vertical here
Head of Machine Learning, Monstarlab Denmark
- Intrado Globe Newswire, “Global Transport Market Report (2020 to 2030) – COVID-19 Impact and Recovery”, 2020
- AWS Amazon, “Predictive Maintenance Using Machine Learning”, 2020
- Deloitte, “Business impacts of machine learning”, 2017
- Deloitte, “Fleet Management in Europe: Growing Importance in a World of Changing Mobility”, 2017
- Forbes, “Key Qualities To Look For In AI And Machine Learning Experts”, 2017
- Lyft Engineering, “What is Data Science at Lyft?”, 2020
- McKinsey, “Fleet Management with Machine Learning”, 2018
- Statista, “Global fleet management market size in 2017 and 2022”, 2017
- Statista, “Challenges companies are facing when deploying and using machine learning in 2018 and 2020”, 2019
- Uber Engineering, “Science at Uber: Powering Machine Learning at Uber”, 2019
- Web FX Digital Marketing Research, “Your Guide to Machine Learning Consulting Rates”, 2020