The Ferryman's Lost Oar: On Navigating Without a Baseline

There is a story, told in different forms along different rivers, of a ferryman who loses an oar mid-crossing. It is not a story of catastrophe. The boat does not capsize. The current is not particularly fierce. The real trouble is subtler: without that second oar, the ferryman can only paddle in a wide, laborious arc. He knows the far shore is there, but he has lost the precise, efficient line to reach it. He is adrift in a sea of relative measurement, with no absolute reference to guide him.

This is the quiet crisis of a service without a baseline. We spend immense effort building our health checks, crafting our pings, and tuning our alerts. We know when latency spikes or error rates climb. But without a deeply understood, historical baseline of what ‘normal’ truly looks like for *this specific service*, every alert becomes a frantic guess. Is this 300ms latency bad? Is a 2% error rate a cause for panic or a Tuesday afternoon? We are the ferryman with one oar, reacting to the water's movement without knowing our intended course.

The baseline is the lost oar. It is the silent partner in observability, the measure against which all noise is judged. It is not a single number, but a living pattern—the circadian rhythm of your system. It knows that latency naturally rises during the lunch hour database backup, that memory usage climbs steadily until the daily cache purge, that a handful of errors from a particular legacy API are, regrettably, part of the healthy background hum.

Without this context, we risk two equally dangerous outcomes. We either become numb, ignoring alerts that are actually cries of distress because they ‘always look like that,’ or we become hysterical, launching all-hands investigations into completely normal behavior, burning out our crew on false alarms. We are either ignoring a fire because we’re used to the smell of smoke, or calling the fire department for a candle.

Crafting a true baseline is the unglamorous work of the ferryman on a calm day. It is the deliberate, patient act of noting the river’s moods—the way the current shifts with the season, how the wind picks up at dusk, which banks are soft and which are firm. It is the log review when nothing is wrong. It is the graph stared at during peace time, not crisis. It is knowing the water so intimately that the slightest unfamiliar tug on the hull is instantly, instinctively felt. Then, and only then, when an oar goes missing, you are not lost. You know the river, and the river, in its way, knows you.

Notes & further reading

A few pages I came back to while writing this: