The Floor Moves Quietly

Most teams don't lose a baseline in a dramatic failure. They lose it on a Thursday, through a sequence of reasonable exceptions, each attached to something more urgent than checking whether the number everyone once agreed to watch still says what they think it says.

Companies begin fragile by physics, not necessarily by bad decisions. Process lives in people's heads. Handoffs fail. Systems drift. Most things work because someone happened to be paying attention at the right moment, and that attention is never guaranteed twice. Watch a company scale from fifty to two hundred employees in eight months and the pricing spreadsheet that three people once maintained quietly begins producing different numbers depending on who runs it, which tab they trust, and whether they know about the exception someone added for a customer six months earlier. No single moment breaks anything, but eventually nobody can explain how a deal gets priced, and the people who once could are too buried in newer problems to notice that the question is being asked.

Growth carries that fragility forward into every new surface, which is why teams eventually try to establish a floor: a number they'll check, a definition they'll keep stable, a process that doesn't get renegotiated every time the surrounding conditions become inconvenient.^[2]

We used to check a dashboard every morning. It contained two numbers that mattered for the quarter, pipeline coverage and conversion rate, updated overnight and rendered in green, amber, or red. It wasn't sophisticated, which was part of its usefulness. There were larger models, better analyses, and people capable of explaining why any given morning might be misleading, but the dashboard was the one place in the day where something didn't immediately negotiate back.

For a while, it worked exactly as intended. If conversion fell, the forecast moved. If pipeline coverage weakened, the hiring discussion changed. The dashboard didn't make decisions for us, but it set the terms under which those decisions could be made, and a number that appeared in red had enough authority to interrupt whatever story the room preferred.

Then a launch slipped. Several enterprise escalations arrived at once. A board deck was due Friday. The dashboard became the least urgent thing on anyone's plate, and urgency is a perfect alibi because the urgent things are usually real.^[1] Nobody made an explicit decision to stop looking. We missed one morning, then another, then checked only when someone remembered, which usually meant checking after the number had already become relevant to an argument. Each decision made sense in isolation. There was no obvious wrong turn, just a week of right turns that left the floor somewhere lower than anyone remembered.

The dashboard kept glowing and the numbers kept updating. Nothing about the system itself had failed. What changed was the expectation that anyone would respond to what it said.

For a while, the explanations arrived almost helpfully. Maybe we'd outgrown the metric. Maybe the real work was happening elsewhere. Maybe the segmentation changes had made the comparison less clean, or the new pricing model had altered conversion in ways the historical baseline couldn't capture. Each explanation contained some truth, and the people offering them usually understood the business well enough to make the caveat sound responsible rather than evasive.

There were also more numbers available now. I once tracked seventeen metrics in a quarter, which is a good way to feel rigorous while the one number that matters goes quiet. A baseline does not always disappear because measurement stops. Sometimes it disappears beneath the reassuring evidence that measurement is happening everywhere.

Before the number disappeared, it began requiring a preface.

In a planning session, we were debating the next quarter's hiring plan, and the model said we could support it. The historical conversion line supported the model. On paper, the baseline held.

Then someone said, "Assuming the conversion rate behaves."

Someone else added, "If the segmentation change didn't distort things."

I found myself qualifying sentences that used to land cleanly, explaining which parts of the number I still trusted and which parts might need to be reinterpreted once another month of data came in. We weren't challenging the strategy. We were circling the measurement, and because nothing collapsed in the room, we approved the plan.

That was the shift. The number had gone from something that closed arguments to something that started a smaller argument about whether it should count. It still appeared on slides and updated overnight, which allowed us to treat it as an intact baseline, but it had stopped being a constraint and become context.

There's nothing inherently wrong with revisiting a metric. Markets change, products change, and a baseline built for one version of a company can become actively misleading in another. The problem begins when the organization doesn't decide that the baseline is obsolete so much as learn, gradually, how to talk around it. The number remains visible enough to preserve the appearance of discipline and flexible enough that it no longer forces anyone to alter a plan.

That arrangement can last for a surprisingly long time. A dashboard can remain technically alive long after it has stopped governing anything, just as a process can remain documented long after everyone has built a private workaround. The artifact survives while the friction it once created gradually disappears.

The dashboard was eventually retired. There was no argument about it because by then there wasn't much left to argue over. A new set of metrics took its place, along with new definitions, new colors, and a new explanation of what the company needed to pay attention to now. They may have been better. The difficulty was that, once we'd stopped using the old baseline as a constraint, we'd also lost the ability to tell whether its replacement represented improvement or accommodation. We were simply somewhere new, and the organization had already begun writing the story of how it got there.^[3]

That's what disappears when a baseline fades: not merely the number, but the thing in the room that couldn't be smoothed over in a planning session or qualified into harmlessness in a board update. A functioning baseline doesn't remove uncertainty. It gives uncertainty somewhere to push back.

The loss rarely feels like loss while it's happening. It feels like focus, adaptation, or good judgment applied to unusual circumstances. It feels like skipping the dashboard on a morning when more important work has arrived, then doing the same thing the next morning because the work hasn't gone away.

Some mornings I still don't check.

And the forgetting feels the way it usually does, which is to say it doesn't feel like anything at all.