Start-Stop-Continue

Most leadership feedback is too vague to act on. A manager reads "communicate better" on Monday and does nothing different on Tuesday, because the feedback doesn't name a behavior. Start-Stop-Continue fixes this. It asks three questions: what should this leader start doing, stop doing, and continue doing. Each one forces a specific, behavioral answer. The result is feedback a leader can apply the next morning.

Three questions, zero ambiguity

After respondents rate a leader on eight dimensions, we ask three open-ended questions: What should this leader start doing? What should they stop doing? What should they continue doing?

The framework was created by Phil Daniels, a psychology professor at Brigham Young University, and popularized through Harvard Business School by Thomas DeLong (HBR, 2011). It has since become the default feedback structure in Agile retrospectives, executive coaching, and peer review programs worldwide. We chose it over more complex alternatives (SBI, STAR, free-form essays) for one reason: it forces specificity.

What most people get wrong

The most common mistake is treating all three prompts as variations of "what should change." They are not. "Start" and "Stop" are about change. "Continue" is about reinforcement. Skipping "continue" or treating it as an afterthought turns the exercise into a list of complaints.

The second mistake is vagueness. "Start being more transparent" is not useful. "Start sharing the reasoning behind decisions in the team chat" is. The framework only works when respondents name observable behaviors, not traits.

Why three categories, not one

Most feedback frameworks ask some version of "what could this person do better?" Without structure, the answers collapse into vague evaluations: "communicate more," "be more strategic," "show more empathy." These read like horoscopes. They describe a feeling without naming what to do about it.

The three-category structure forces a different kind of thinking for each prompt. "Start" asks for a new behavior. "Stop" asks the respondent to name something specific they want less of. "Continue" asks them to identify what is already working. Together they produce feedback that is specific, balanced, and actionable.

One thing per category

We ask for one answer per prompt, not a list of five. When people can list everything they can think of, the signal drowns in noise. One item forces them to choose the one that matters most, and it removes the decision fatigue of an open-ended text box.

Numbers tell you where. Words tell you what.

A leader's dimension scores show them where their gaps are. Start-Stop-Continue tells them what to do about it. A low score on Openness tells you something is broken. "Stop interrupting people when they disagree with you in meetings" tells you what. One without the other is incomplete. Together, they give a leader everything they need to start closing the gap.

How the analysis reads the open text

Start-Stop-Continue responses from a full team do not reach the leader as a stack of individual voices. They reach the leader as patterns.

The system scans all open-text responses and identifies recurring behaviors, weighted by how many independent respondents named the same theme. When 4 of 7 people describe some version of the same request, the report surfaces that as a single insight with its prevalence count. Not a verbatim quote. A pattern with weight behind it. The leader knows how broadly shared the feedback is without being able to connect it to any individual source. "4 of 7 respondents named some version of this: schedule more regular one-on-one conversations." That sentence is more useful than any single comment, and more honest about what the team actually thinks.

The analysis goes one level further. Start-Stop-Continue responses are cross-cut against the eight dimension scores. If multiple people name some version of "explain your reasoning" in Start, and the leader's Openness score sits in the lower half of the scale, those two signals are read together. The report surfaces two to four recurring themes that connect qualitative feedback to quantitative gaps, so the leader does not have to make the connection manually.

The structural principle underneath all of this is frequency before identity. The question the analysis answers first is: how many people are saying some version of this? That question matters more than who said it, and answering it first is what makes the feedback safe to surface at all. A leader cannot reverse-engineer which of their 8 respondents said something when the response is reported as a team-wide pattern with a count. The weight is real. The trail is gone.

References

1. DeLong, T.J. (2011). "Three Questions for Effective Feedback." Harvard Business Review.
2. Sheridan, S. & Burt, J. (2014). "Stop, Start, Continue: A structured approach to feedback quality." Assessment & Evaluation in Higher Education, 40(5).