CFD Testing band widens: what does it indicate and what experiments?

What's really being asked

The interviewer is evaluating whether you treat agile metrics as diagnostic instruments rather than dashboard decorations. Specifically, they want to see if you understand that a widening band on a Cumulative Flow Diagram represents growing Work In Progress in that stage, which mathematically means the arrival rate exceeds the departure rate. They also want evidence that you design empirical experiments, measure outcomes, and avoid knee-jerk structural changes like hiring.

The full answer

First, interpret the signal: the Testing band widening means inventory is accumulating there, not that testers are lazy. Second, name the system constraint: work is entering Testing faster than it exits, which could be caused by large batch handoffs, insufficient automation, flaky environments, or a policy that requires every ticket to pass through a single tester. Third, propose three specific experiments in priority order: one, reduce batch size by lowering the work in progress limit for the Development stage so smaller slices flow into Testing; two, measure touch time versus wait time to see if tests sit idle or actually take long to execute; three, run a time-boxed two-week experiment where developers spend thirty minutes daily assisting with testing while tracking cycle time and defect escape rate. Fourth, define success metrics before starting, such as Testing band width stabilizing below five items or average cycle time dropping by fifteen percent.

The mistakes people make

The biggest red flag is blaming the testing team for slow velocity without examining systemic causes. Another failure mode is proposing to hire more testers immediately, which permanently increases cost before validating whether the bottleneck is capacity or batch size. Saying you would tell developers to write better code is also a non-answer because it does not produce a measurable experiment. Finally, confusing a widening band with a flat line is dangerous; a flat line means zero throughput, while widening means delayed throughput.

What usually comes next

The interviewer may ask how you distinguish between a capacity bottleneck and a batch-size problem. They might also ask what you would do if the Testing band narrowed but the Done line flattened, indicating work is stuck in a hidden stage outside the board. Another follow-up is calculating approximate cycle time from the CFD or explaining why parallel bands represent healthy flow.

A concrete example

Suppose your CFD shows the Testing band growing from three items to twelve over four sprints while the Done line slope remains constant. You hypothesize that developers are throwing five-story batches over the wall. Your experiment: enforce a WIP limit of two stories in Development for two sprints. You predict the Testing band will plateau below six items and cycle time will drop by twenty percent. You track daily. If the band keeps widening, you pivot to measuring automated test runtime, discovering that regression takes four hours due to environment provisioning. You then experiment with containerized test environments and measure touch time reduction.