The Weakest Link, Solved One Careful Step at a Time
A factory assembling products from several parts can only finish as many products as its scarcest part allows. A production line moves only as fast as its slowest stage. A decision judged by its worst outcome is only as good as that worst case. In all these situations, the smallest number in the set decides everything — mathematicians call this the minimum function, and it is the weakest link made precise.
The minimum function has an awkward feature: when the identity of the smallest input changes, it develops a sharp corner, and the optimisation methods that engineers rely on — which need smooth curves they can differentiate — struggle with corners. The usual workaround is a smooth approximation of the minimum. But an approximation that overshoots is dangerous: it would promise more products than the parts allow, or more capacity than the line has. What engineers want is a smooth estimate that always stays safely below the true minimum, and gets closer step by step.
A new step that never overshoots
In a paper published in the Journal of Inequalities and Applications, Prof. Thomas A. Weber of the Chair of Operations, Economics and Strategy presents a method that does exactly that. The harmonic-increment transformation starts from any smooth estimate that lies strictly below the minimum, measures the gaps between the inputs and that estimate, and adds a carefully chosen increment: the harmonic mean of those gaps, divided by the number of inputs. The result is still smooth, still strictly below the minimum — and closer to it. Repeating the step produces an increasing sequence that converges to the minimum, with explicit error guarantees. With two inputs, each update removes at least half of the remaining error, and starting from zero with positive inputs, the relative error falls below 0.1% after 10 updates.
The method also works alongside existing approximations, including the widely used LogSumExp soft minimum, and a companion construction based on the geometric mean gives smooth upper bounds. Together, the lower and upper estimates trap the minimum inside a shrinking interval.
Two ladders in an alley
The idea is beautifully illustrated by a classic puzzle, the crossed-ladders problem. Imagine two ladders crossing an alley, each reaching from the foot of one wall to a fixed point on the opposite wall. If those upper contact points are three metres and six metres above the ground, the ladders cross at two metres. That crossing height is the first harmonic increment.
Now replace the ladders with shorter ones whose lower ends are raised to that crossing height, keeping the upper points fixed. The new crossing is at 2.8 metres. Repeat again and it comes to approximately 2.9882 metres. Each crossing creeps closer to three metres — the lower contact height, the true minimum — from below, never passing it.
That is the whole method in miniature: a safe estimate, improved by a guaranteed fraction each time, until it is as close as you need. The same logic applies wherever a weakest link governs the outcome, from selective assembly — where output in each class depends on the less abundant compatible part — to decisions judged by their lowest payoff across a set of scenarios.
For students, this is maths doing what it does best: turning a practical headache (a corner you cannot differentiate, an estimate you cannot trust) into a procedure with a guarantee. You do not just get closer; you know exactly how much closer each step gets you.
If your child wants to see how maths like this powers real decisions — and to build the confidence to tackle problems where the answer must be safe, not just close — a first lesson in maths can start with puzzles like the crossed ladders and build from there. Book your first lesson
Sources:
- phys.org, "New mathematical method improves calculations for systems limited by their smallest input", 29 October 2026: https://phys.org/news/2026-10-mathematical-method-limited-smallest.html
- Thomas A. Weber, "Smooth approximation of the minimum function from below (and above)", Journal of Inequalities and Applications (2026), DOI: 10.1186/s13660-026-03519-3
London Tutors — post your tutoring request and we will match you with a verified tutor.