Finding Bottlenecks Before They Find You

Forecasting tells you where you're headed. It doesn't tell you why your current pace is what it is. For that, PtahCast's flow metrics dashboard exists to answer a different question: not "when will this be done," but "how has delivery actually been going, and where is it getting stuck?"

Four charts, one shared window

Everything on the metrics dashboard reads from the same underlying history and shares the same sample window — 30, 60, or 90 days, same trade-off as forecasting: wider is smoother, narrower is more reactive.

The throughput run chart is the same weekly completed-ticket data that feeds forecasting, just shown directly as a bar chart with a dashed mean line. Weeks with zero completions still show up as empty bars — a run of them is worth investigating before it quietly drags your next forecast down with it.

The cycle time scatter plots every completed ticket by the date it finished, with height showing how many days it took from first pull to Done. Three dashed lines mark the P50, P85, and P95 cycle times. A tight cluster near P50 means predictable delivery; dots scattered above P85 mean something is dragging — check the slowest-tickets table beneath the chart for a pattern worth naming.

The two charts that actually show you where work is stuck

The run chart and scatter tell you about pace. Two other charts tell you about location.

Work in progress over time is a stacked area chart of how many tickets currently sit in each column, day by day (Done is excluded — this is about work still moving, not work that's finished). If one band keeps growing wider than the rest and stays that way, that's a column accumulating tickets faster than they're leaving it — the visual signature of a bottleneck, and a strong candidate for a tighter WIP limit.

The cumulative flow diagram answers a related but different question: not "what's sitting where right now" but "how has each stage's total intake grown over time." Bands here only ever grow, stacked from Done at the base up through earlier stages. Read it by slope, not height — a band with a steady, steep slope is a stage moving at a consistent pace; a band that flattens means work has stopped arriving there, or stopped leaving. A band that keeps widening over the window is the classic CFD bottleneck signature: tickets entering that stage faster than they exit it. The vertical gap between the Done line and the top of the whole stack is, roughly, your total work-in-flight — watch that gap over time, and if it's growing, more is coming in than is going out, regardless of which single column looks worst.

Why this pairs with forecasting rather than replacing it

Rising WIP alongside flat or falling throughput is usually the earliest warning sign available — it shows up in the flow charts well before it shows up as a worse forecast date. Flow metrics are the diagnostic; forecasting is the projection built on the same evidence. Look at both, in that order: understand why your pace is what it is before you trust where the simulation says it's heading.

A forecast built on top of an unexamined bottleneck is still an honest forecast — it's just honestly projecting a problem forward instead of catching it early.