Flow metrics tell you how work moves, not how busy people looked. Cycle time, lead time, and throughput are different questions. A cumulative flow diagram makes the queues visible. Little's Law ties them together so you cannot brag about throughput while sitting on a pile of half-done work and calling it a pipeline.
Cycle time, lead time, throughput — stop using them as synonyms
Cycle time is how long an item spends from the moment the team starts it until it meets done. Lead time is how long the customer waits from request (or from Ready) until done. Throughput is how many items finish per week or per sprint. If you report only one, say which. 'Faster' without a noun is advertising.
A short cycle time with a long lead time means work waits in a queue before anyone starts — a Ready column that is a warehouse. A healthy cycle time with miserable throughput means you are finishing rarely, perhaps because items are huge or the system is blocked. Do not celebrate a fast cycle on the only ticket that moved.
Measure from the same policies you wrote on the board. If 'start' means three different things, your cycle-time chart is a short story collection. Ageing work — items whose clocks are still running — is a leading indicator. The item that has sat in Review for three weeks will not become faster because you averaged the ones that slipped through.
The cumulative flow diagram: queues as a picture
A CFD stacks, over time, how many items are in each state. The top line is arrivals into the system; the bands are the queues; the bottom growth is done. Parallel, stable bands mean arrivals and departures roughly match. A band that widens is a queue growing — usually WIP rising in that state.
Read it left to right. A sudden vertical jump in arrivals is a push event: someone dumped a batch. A flat 'Done' band while other bands fatten is a system that is starting and not finishing. That picture will end an argument faster than your speech about focus. Bring the picture.
Do not need a perfect tool. A weekly count per column on a spreadsheet will draw the same lesson. Beauty is optional. A widening Review band is not. If the CFD is only opened at the quarterly showcase, you are using it as art.
Little's Law without the textbook voice
In a stable system, average WIP equals throughput times average cycle time. Rearranged: average cycle time is WIP divided by throughput. Ten items in progress, two finishing a week, and you should not be surprised if items spend about five weeks in the machine. You can argue with the team. You cannot argue with the relationship for long.
Unstable systems — wildly varying arrivals, items that change size, lots of cancelled work — make the averages wobble. That does not make the idea useless. It tells you to stop pretending a single average is a promise. Use a range: half our items finish inside n days; some take much longer. The tail is where sponsors get hurt.
Management levers follow the law. To cut cycle time you reduce WIP or raise throughput (or both). Raising throughput without shrinking item size or removing a bottleneck is a wish. Reducing WIP is the lever you already have a policy for. If cycle time is bad and WIP is high, you do not have a mystery. You have a choice you have not made.
Summary & next steps
Cycle time is start to done; lead time is request to done; throughput is finishes per period. The CFD shows growing queues as widening bands. Little's Law: more WIP at the same throughput means longer cycle time. Measure from written policies.
Next: continuous improvement — kaizen, PDCA, root cause, and one change at a time so you can tell what worked. That lesson is written for flow systems and for quality systems; the loop is the same.