Partial Optimization and Overall Optimization Considered from Examples in Everyday Life in Indonesia

The criteria for optimization vary depending on the perspective: in the case of traffic congestion, it’s throughput (total movement per unit of time); for corporate activities as a player in market principles, it’s profit maximization; and in manufacturing production activities, it’s productivity improvement or cost reduction.

The "Optimal" Way to Alleviate Traffic Congestion in Indonesia

Recently, I’ve been commuting to the MM2100 service office in Cibitung two or three days a week, often leaving precisely at 5 p.m. to head home. However, both on the way there and back, I get stuck in terrible traffic jams near the TOL (toll road) gates. Most drivers, including myself, are filled with the desire to reach their destinations as quickly as possible, leading to a tense, cutthroat struggle with dangerously narrow gaps between cars as everyone vies to get ahead.
When I’m completely stopped in the overtaking lane and glance over to see the left lane suddenly flowing smoothly, I feel a surge of frustration. At the same time, I exhaust myself with thoughts like, “What if I switch lanes only to end up behind a slow-moving large trailer that acts as a bottleneck, making me regret not staying in my original lane?” If I had boundless energy and flawless driving skills to avoid accidents, constantly switching lanes might be the fastest way to arrive.
Much like the "Prisoner’s Dilemma," where two prisoners could both receive light sentences if they stay silent but end up confessing to avoid a harsher penalty—resulting in both getting the heaviest punishment—the best solution to ease congestion would be for everyone to slow down without changing lanes. However, in an environment where people operate on the assumption of bad faith and can’t trust others, they think they’ll be the only fool left behind unless they outmaneuver everyone else, ultimately worsening the traffic jam.
In Japan, there’s an unspoken rule when riding escalators: stand on the left and leave the right side open for people walking (though it seems to be the opposite in Kansai). However, the most efficient way to move on an escalator is for everyone to stand calmly on both sides. Despite efforts by railway companies to promote this, it hasn’t caught on because people think, “I’m in a hurry, so if those who aren’t in a rush stand to one side, neither of us should have any complaints.”

The Concept of Partial Optimization vs. Overall Optimization

Overall optimization only works in situations where stakeholders share the same interests, such as a factory aiming to “increase productivity,” “ship without delays,” or “reduce manufacturing costs” to maximize collective benefits. However, in scenarios like highway traffic jams or rush-hour escalators at stations—where everyone is worried about being late to their own company—it’s difficult to achieve because individual interests are misaligned.
Even within organizations with shared interests, reality often involves over-ordering materials out of fear of production line stoppages due to shortages, overstocking intermediate inventory to avoid halting downstream processes, or piling up product inventory to prevent delays to customers. This leads to increased inventory costs. Conversely, insufficient materials, work-in-progress, or product inventory can result in opportunity losses. These outcomes of partial optimization ultimately undermine overall optimization by increasing both inventory costs and opportunity losses.

In the production processes of Indonesia’s two- and four-wheeled vehicle parts factories, the front-end process is often molding or pressing, while the back-end process is assembly or welding. The takt time differs between the automated front-end and the manual back-end. If production plans are made with a forward-looking approach from the front-end, inventory accumulates there. Thus, aligning the front-end plan with the slower takt time of the bottleneck back-end process achieves overall optimization.
In capitalist economies like Japan or Indonesia, business players in the market operate on competitive principles, aiming for partial optimization by outsmarting rivals to maximize their own profits. However, in situations where public welfare—like traffic jams or escalators—takes precedence, consideration for others becomes necessary. This involves limiting one’s own desires to respect others’ benefits, leaning toward a more socialist mindset.
Just as centipede races or three-legged races aim for team victory, organizations internally pursue overall optimization while externally adopting a partial optimization approach in the market. However, real organizations are made up of people who want to finish quickly and go home or achieve results in the shortest time for their own career advancement. The ability to overcome this tension and align with the organization’s collective pace is a key difference that sets humans apart from animals (with exceptions like wolves or lions that hunt in packs).

Clarifying the Criteria for Optimization

So far, I’ve discussed optimization criteria purely in terms of throughput (in this case, total movement per unit of time). However, in reality, not everyone stuck in traffic or on an escalator is necessarily in a hurry. In fact, those with time to spare might gain a supreme sense of satisfaction—unmatched by anything else—by yielding to those in a rush, contributing to others (not as self-sacrifice). This could be seen as a bit of a stretch, but it holds water.
A common pattern in optimization discussions is the hellish loop where someone counters with, “Overall optimization doesn’t necessarily equate to the optimization of human happiness,” derailing the conversation. This happens because the criteria for measuring optimization aren’t clearly defined. For instance, when our company discusses optimization in manufacturing, the fundamental axis is “reducing inventory by shortening procurement and manufacturing lead times,” while factoring in constraints like equipment and personnel as production resources.
In comedy routines or debates, people might deliberately shift the topic to get a laugh or steer the discussion in their favor. However, most people do this unconsciously and without a plan, causing the conversation to go in circles, wasting time and ending in a conclusion that benefits no one.
During discussions about optimizing factory productivity through equipment, someone might argue, “Optimizing production resources doesn’t necessarily maximize customer satisfaction,” derailing the focus. This stems from unclear criteria for measuring optimization.
Our company focuses on IT-driven process improvements, and clients often say, “Just optimize it automatically.” This is like asking, “What’s the most delicious dish in the world?” Without knowing what form of optimization the person considers ideal, we’re left with no way to proceed.

An Optimization Approach for Manufacturing Production Planning

In manufacturing, a baseline production plan is created based on customer orders received by the sales department, anticipating shipments. Using the bill of materials, the net requirements for parts to be manufactured in each process and the raw materials to be used are calculated. Production lots are scheduled by staggering them according to the manufacturing lead time, assigning tasks to production resources. If workload exceeds capacity, tasks are shifted forward or reassigned to alternative resources for leveling, forming the production schedule.
In this process of leveling overloaded tasks to fit within the daily capacity of production resources, moving excess tasks causes a chain reaction where previously assigned tasks get pushed out recursively. To automate this adjustment—while considering pre-set parameters like dispatching rules (task assignment order) and resource evaluation (task assignment priority)—a production scheduler creates an optimized schedule.
However, relying solely on parameter settings in a production scheduler has limits in deriving an optimal solution that satisfies constraints like resource leveling, minimizing setup times, reducing delays, or batching furnace loads. For better results, an optimization approach is needed: repeatedly running simulations, sampling the outcomes, and evaluating the one where the objective function (the function to maximize or minimize in an optimization problem) is closest to the optimal value.
In the era of big data, analyzing data to optimize and predict factors that degrade factory productivity and profitability—such as potential delays, raw material shortages, productivity drops from frequent setups, or excess inventory from lot consolidation—is essential for faster decision-making and improved operational efficiency.
This optimization approach isn’t limited to manufacturing. It’s used in retail for exploring effective floor space utilization, in marketing for offering trend-aligned products, in transportation for securing the safest and lowest-cost delivery routes, and in finance for determining asset portfolios with the highest investment returns. Leveraging data analysis and optimization in business management has already helped many companies achieve sales growth and cost reductions.