Fever charts and buffer management in Critical Chain

The horizontal axis is chain progress: the percentage of the critical chain (or a feeding chain) that has been completed. The vertical axis is buffer consumption: the percentage of the associated buffer that has been used up. Each reporting period you place a single dot at (percent complete, percent buffer consumed), and the trail of dots is your trend.

The diagonal band running from bottom-left to top-right is the zone map. Early on, consuming some buffer is fine — you have most of the chain left to recover. Late in the chain, the same level of consumption is dangerous because there is little work left over which to absorb it. That is why the acceptable region narrows as you move right: the chart grades risk against how much runway remains, not against a flat threshold.

The three zones turn a continuous number into a decision. They map cleanly to the only three responses a sponsor needs: do nothing, plan a recovery, or act now.

A deterministic Gantt implies that every task has a true due date, so a single late task reads as a problem. In reality, R&D tasks vary widely, and CCPM deliberately strips per-task padding and pools that safety into shared buffers. Once the safety lives in the buffer, the only number that protects the deadline is buffer consumption — so that is what you manage.

This is liberating for both the team and the sponsor. Task owners are not punished for a single estimate that ran long; the buffer is there precisely to absorb it. The board deck stops carrying dozens of individual task statuses and carries one honest signal instead. And because the fever chart is forward-looking — it reacts the moment buffer is burned, not when a milestone is formally missed — it gives weeks of warning that a Gantt percentage-complete bar never does.

A fever chart is only as honest as the buffer underneath it. Size the buffer too small and the chart screams red on noise; size it too large and it stays green while the program quietly slips. CCPM offers two sizing methods. The simple cut-and-paste method sets a buffer to roughly half the safety time removed from the tasks it protects — fast, but blind to how the risk is distributed.

The rigorous method derives the buffer from a Monte Carlo simulation of the chain, so the buffer reflects the actual spread between the median (P50) and a target confidence level such as P80. A buffer sized from the real distribution makes the zones mean something: yellow genuinely corresponds to the program drifting away from its P80 commitment, not to an arbitrary rule of thumb.

CritPath AI identifies the critical chain from the CPM network, sizes project and feeding buffers — including directly from its Monte Carlo engine so the buffer reflects your actual P80 rather than a guess — and tracks consumption on a live fever chart that updates as task progress is logged. The same screen shows the Drum-Buffer-Rope view, so you can see the constraint, its buffer, and the rope that paces feeding work in one place. Resource leveling is available as a separate resource-feasible view.

Because the chart is wired into the real dependency graph, the AI copilot can do more than color a dot. Grounded in your actual schedule, it explains which task is eating the buffer, why the trend turned yellow, and what a proposed re-sequence or staffing change would do to consumption downstream. When a risk fires or a decision gate is resolved, retroactive rescheduling re-derives the chain and the buffer, and the fever chart moves with it.

All of this is part of the standard product at $10 per user per month, with AI usage billed separately by metered usage — the full method stack in a modern web app rather than a 2000s desktop tool.