Counting More Often

Glagalbagal had solved the problem of measuring growth over a year. But waiting twelve months to discover a bad manager seemed reckless — by the time Hrijpa came around, the damage was already done. He needed to measure more frequently.

The Monthly Plan

After replacing both managers (their successors were a pair of stegosauruses named Frojj and Mvantika, recommended by Hyjop the town gossip, which in hindsight should have been a warning sign), Glagalbagal decided to take measurements every Smujka — the monthly new moon festival, during which the town celebrated by eating fermented moss.

The process was the same as before. At each Smujka, the managers would create a pebble arrangement for their herd. Glagalbagal would compare it to the previous month's arrangement, just as he had done across years.

The First Three Months

The first Smujka measurement was encouraging. Location 1, under Frojj, had grown by a few animals. Location 2, under Mvantika, had also grown. Glagalbagal was pleased.

The second Smujka was alarming. Location 1 had shrunk. Frojj reported that a flash flood had scattered part of the herd across the river, and several animals had not yet been recovered. Glagalbagal considered firing Frojj on the spot — after all, the herd was smaller, and that was exactly the signal his system was designed to detect.

But he held off, and by the third Smujka, the missing animals had returned on their own. Location 1 was now larger than it had ever been. Meanwhile, Location 2 had shrunk — Mvantika blamed a seasonal illness that afflicted cattle every spring.

The Confusion

Glagalbagal laid out the three months of records for both locations. The pattern was baffling. Location 1: up, down, up. Location 2: up, up, down. Each individual month seemed to tell a different story. If he had fired Frojj after the second month, he would have lost a manager whose herd was actually growing. If he fired Mvantika after the third month, he might be punishing her for an illness that was nobody's fault.

He realised that his system, which worked well over a year, was producing confusing results over a month. A year of change was large enough to reflect the manager's actual work. A month of change was small enough that random events — floods, illnesses, animals wandering off and returning — could easily mask or mimic the effect of good or bad management.

The Mistake

Glagalbagal, however, arrived at this insight only after making a costly error. Between the second and third Smujka — when Location 1 had just shown a decline — he had sent Frojj a threatening message via carrier pterodactyl, warning that if the herd did not recover, Frojj would be replaced. Frojj, panicking, purchased twelve sheep from a neighbouring farm at an inflated price just to make the numbers look better for the third measurement.

Glagalbagal only discovered this when reviewing the accounts and noticing the expense. The herd had not actually grown as much as the third Smujka arrangement suggested — part of the "growth" was purchased, not bred.

This revealed a second problem with frequent measurement. When managers knew they were being evaluated each month, they optimised for the short-term measurement rather than for genuine long-term growth. Frojj had wasted resources buying animals to avoid a single bad month, when the bad month was caused by a flood that would have resolved itself.

Blortz's Observation

Blortz, who had been rehabilitated and now served as Glagalbagal's bookkeeper (the irony of a former animal-eater maintaining livestock records was not lost on anyone), made a useful suggestion.

"You have three months of records," Blortz said. "Stop looking at each month separately. Compare the arrangement from the first Smujka to the arrangement from the third."

Glagalbagal did this. Over three months, Location 1 had grown substantially. Location 2 had grown slightly. The random fluctuations in the individual months — the flood, the illness — largely cancelled out, and the overall direction was visible.

The Revised System

Glagalbagal settled on a compromise. He would continue taking monthly measurements — they were useful for spotting genuine catastrophes like a predator wiping out half the herd. But he would evaluate managers only on the cumulative change over four months. No individual month would trigger a reward or punishment.

He also realised that comparing each month to the previous month was less useful than comparing each month to a fixed starting point. The comparison against a fixed reference gave a cleaner picture of the overall direction than a chain of month-to-month changes, which could see-saw endlessly without revealing any trend.