WW/SUPPLYCHAI
Fix the bottleneck and it moves: simulation maps where it goes next
A 1,000-node simulation shows 20% capacity gains kept the bottleneck on the same critical path for ten rounds, while 50% jumps sent it wandering across four paths.
- By
- Amara Osei
- Filed
- Length
- 888 words
- Read
- 4 min

Key points05
- Ten rounds of 20% bottleneck improvements kept the constraint on the original critical path in a 1,000-node, 50-factory simulated network.
- At 50% improvements per round, the bottleneck left the critical path at round five and crossed four paths over ten rounds.
- The longest path ran 7,660 vs 7,451 for the second-longest — a 2.8% gap that set the map's durability limit.
- The author's best real-world result was a one-off 50% lift in a plant's monthly output; typical improvements take two weeks to two months.
- Part 4 of the series publishes Tuesday, Nov. 24; the author completed a PhD on the topic in June 2026.
Ten rounds of 20% capacity improvements moved a simulated supply chain's bottleneck to ten different nodes — and not once pushed it off the original critical path. That's the core finding from the third instalment of a four-part series on multi-tier supply chain networks by Brian (Yoshiatsu) Kawabata, a former Honda, Boeing and Intel engineer who completed a PhD in computer science at Tokyo University of Agriculture and Technology in June 2026.
The question sounds academic. It isn't. Every industrial engineer knows the bottleneck shifts once you fix it. Kawabata writes that none of the IEs he asked could say how it shifts — so he built a simulation to find out.
Why only the bottleneck deserves investment?
Line up the throughput capacity of every process in a network and the lowest one governs what the customer actually receives. Only improving that node raises delivered volume. Spend the money upstream of the bottleneck and you simply grow the work-in-process queue in front of it. Spend it downstream and you buy idle time, because the parts never arrive.
Either way, the customer's order backlog stays exactly where it was. As Kawabata puts it, any improvement spent on a non-bottleneck is "a waste of time, money and effort." Widen the lens to a full supply chain and the point sharpens: when an OEM raises output, the constraint sits inside one process at one firm, and investment anywhere else is money burned.
What did the simulation actually show?
The network was built like a tree — many leaves converging on a single root. It held 1,000 nodes across 50 factories, averaging 20 nodes per factory with a standard deviation of 1. The longest path ran 7,660 units against 7,451 for the second-longest, a gap of just 2.8%. Two improvement patterns were tested: +20% per round and +50% per round, each repeated ten times.
The results split cleanly:
- At +20% per round: the bottleneck never returned to the same node twice, but stayed on the original critical path through all ten rounds.
- At +50% per round: it held the critical path through round four, jumped to another path in round five, then wandered across four different paths over the ten rounds — returning to the original critical path several times.
For context on the magnitudes: Kawabata's most successful real-world improvement raised a plant's monthly average output by 50% — once — and he calls that rare. Floor-level improvement typically takes two weeks to two months and sometimes fails to raise throughput at all.
The takeaway is that a network map has a durability limit, and that limit is set by improvement size relative to the length gap between paths.
What does this change for suppliers?
Quite a lot, if the ranking holds. A supplier sitting off the critical path is, for now, unlikely to get a sudden call from the OEM's supply chain director announcing it's the new bottleneck and demanding emergency action. Even after ten rounds of 20% improvement, the firm getting that visit was on the original critical path.
The catch is that no firm can compute this rank alone. You can measure your own lead time; you cannot know where it sits in the network without every other firm's figures and the map of how they connect. Hence Kawabata's proposal: a shared platform telling each supplier its path rank, the current bottleneck's location and capacity, the post-improvement target, and — the commercially loaded item — the OEM's production plan months and years ahead.
Privacy concessions stay narrow. A firm flagged as a bottleneck candidate discloses the spec of that one process, nothing more. Firms that have pared dwell time through hard value-stream analysis show short lead times, come up as critical-path candidates less often, and get asked to open their books less often. Good flow, it turns out, is its own confidentiality shield.
What's in it for the BN supplier and the banks?
A bottleneck firm can ignore the OEM's request — and betray every non-bottleneck firm downstream, whose equipment, staff and know-how sit idle because the constraint rate-limits the whole chain. Non-BN firms only convert spare capacity into revenue when the bottleneck improves. Cooperation also reduces the risk of being swapped out for a more pliable supplier.
Banks get arguably the sharpest idea in the piece: if non-bottleneck investment is pure waste, then the most important lending variable is whether the borrower is the constraint. Strengthening the bottleneck serves demand that already exists, so the structural risk of the investment missing is low. Kawabata even suggests lending products for SME bottleneck suppliers priced near the OEM's own rates.
The honesty is welcome, too. Kawabata flags the work as one simulation run once on one network — fixed process times, no inter-process waiting, pull ordering only, no dual sourcing, product variation or defects. Suggestive, no more. But since nobody had an answer to where bottlenecks go under improvement, a single observation beats none.
Part 4, due Tuesday, Nov. 24, takes up how far the method reaches — and where it stops. If the shared-map logic holds, the next fight in supply chain competition may be over who gets to see the ranking first.
Source: scmr.com
More from Amara Osei
Show full bio
Staff writer covering regulation, policy and their impact on Ocean and air freight, container shipping at Waybill Wire.
961 articles