Hello Reader!

Who profits when nobody checks? I've been asking myself that since an envelope landed in my mailbox, and the answer applies to far more than my mail.

The Solver Always Does What You Pay It For

When a client's model returns a strange plan, people often suspect a bug. Sometimes it is one. More often, the solver did exactly what the objective told it to do. If the objective rewards low cost and nothing else, the solver will happily run the inventory down to zero, keep one machine busy around the clock, or send a truck on a long detour to save a single cent.

The solver isn't lazy or greedy. It follows the reward, precisely and without judgment. So when a plan looks odd, my first question is: what does this objective actually reward, and what does it forget to punish?

Checking details like this is what I do every day. And recently, I caught myself asking the same question about something far away from any model: my utility bill.

The Letter That Always Says Sorry

The letter was polite. It apologized for an error in my annual utility bill and announced a bank transfer to refund the difference.

I had received this kind of letter before. Twice, in fact.

I have lived in my flat for three years. Every year, the administration sends a utility bill. Every year, I read it line by line, the same way I check a model. And every year, I have found an error worth a triple-digit amount.

The errors varied. One bill left the entire state heating subsidy with the landlord, although landlords have to pass it on to their tenants. Another billed one part of the heating costs twice. Another billed the building's electricity twice.

To be fair, the administration always corrects its mistakes. When I point one out, they refund me and every other tenant in my building, without anyone else having to ask.

But when the third letter arrived, a different thought kept me busy. Three bills, three errors. That is not bad luck anymore. That is a pattern. And patterns have causes.

My building gets its money back because one tenant checks. What about the buildings where nobody does?

What If Every Cent Cost a Euro?

Let's look at this the way we would look at any optimization problem: who makes the decision, and what does it reward?

The administration decides how much effort goes into checking a bill before it goes out. Careful checking costs time. Mistakes cost them nothing. If a tenant spots an error, they refund the amount, and that's it. If nobody spots it, the money stays on their side of the table.

So the incentive doesn't just fail to push toward correct bills. It points the other way. Every error that slips through unnoticed is a small reward.

I don't think anyone in that office sits down and plans to overcharge. They don't have to. The incentive does the work for them, one skipped double-check at a time.

Here is what I would change. For every cent an administration overcharges, and a tenant catches, the tenant gets the cent back. On top of that, the administration pays one euro, 100 times the amount, to a tenant protection organization.

The striking part: this works even if almost nobody checks. With a 100x penalty, overcharging becomes a losing game once roughly one tenant in 100 catches the error.

Admittedly, 100 times is a massive penalty. Set it that high, and administrations would likely start charging a risk fee: insurance against an over-tuned parameter. The exact number matters less than the direction. With a well-tuned penalty, a rigorous check before every bill goes out becomes the cheapest option. And that check isn't hard anymore. We automate unit tests for software every day. We can automate a plausibility check for a utility bill too.

The lesson reaches far beyond utility bills. Your solver follows its objective. So do people. When you model a supplier, a colleague, or a partner, don't ask what they promise. Ask what their objective rewards. That tells you what they will do.

Be the One Tenant in 100

My neighbors got their money back because one person checked. Ideas spread the same way. If you know someone who would enjoy seeing everyday life through an optimization lens, forward this email to them.