Categorizing and Prioritizing Feedback in a Start Stop Continue Exercise
This skill teaches you how to sort, cluster, and dot-vote on collected Start, Stop, and Continue feedback items so your team commits to the highest-impact actions instead of drowning in a long, undifferentiated list.
After collecting feedback, read all items aloud, then group duplicates and near-duplicates into themed clusters within each Start, Stop, and Continue column. Use dot-voting (each person gets 3–5 votes) to surface the highest-impact items. Select the top 2–3 voted items across all categories and convert them into specific, assigned action items with owners and deadlines.
Outcome: Your team consistently exits retrospectives with a short, prioritized list of concrete commitments rather than an overwhelming wall of sticky notes that never get acted on.
Prerequisites
- Familiarity with the Start Stop Continue framework
- A completed collection phase with raw feedback items from team members
- Basic facilitation skills for group discussions
Overview
Collecting Start, Stop, and Continue feedback is the easy part. The hard part — and where most teams fail — is turning 30+ sticky notes into 2–3 commitments the team will actually follow through on. Without a structured categorization and prioritization step, your start stop continue exercise produces a feel-good discussion that changes nothing.
This skill covers the full workflow from raw items to ranked priorities: reading and clarifying items, grouping duplicates into themes, using dot-voting or similar techniques to surface what the team collectively cares about most, and selecting a manageable number of actions to commit to. It's the bridge between 'we talked about it' and 'we're doing something about it.'
Whether you're running a sprint retrospective, a quarterly team review, or a project post-mortem using the Start Stop Continue framework, mastering this prioritization step is what separates teams that continuously improve from teams that just continuously meet.
How It Works
The categorization and prioritization process works in three distinct phases, each solving a specific problem.
Phase 1: Clarification removes ambiguity. When someone writes 'better communication,' that could mean anything. By reading each item aloud and letting the author add 10 seconds of context, the whole team builds shared understanding of what each piece of feedback actually means.
Phase 2: Affinity Clustering reduces cognitive load. A board with 40 individual items is overwhelming. By grouping related items into themed clusters (e.g., three separate notes about standup meetings become one cluster called 'standup format'), you compress the decision space from dozens of items to 8–12 themes. This makes voting meaningful rather than scattered.
Phase 3: Democratic Prioritization via dot-voting ensures the team's collective wisdom — not the loudest voice — determines what gets acted on. Each person distributes a limited number of votes across the clusters they believe would have the highest impact. The constraint (limited votes) forces genuine prioritization rather than 'everything is important.' The result is a clear signal of where the team's energy and attention should go.
This three-phase approach works because it respects both divergent thinking (everyone's input matters) and convergent thinking (we must narrow down to act). It's the same pattern used in design thinking affinity mapping, adapted specifically for the start stop continue exercise context.
Step-by-Step
Step 1: Read Every Item Aloud for Shared Understanding
Before any grouping or voting, the facilitator reads each sticky note or digital card aloud, one at a time. After reading, the author has 10–15 seconds to add brief context if the wording is unclear. No debate or discussion happens at this stage — just clarification.
Work through one category at a time: all Start items first, then Stop, then Continue. This keeps the team focused and prevents cross-category confusion. If an item is truly incomprehensible even after the author explains, rewrite it on a fresh card with the author's approval.
Tip: Set a visible timer for the clarification window. Without it, this phase balloons as people start debating items instead of just clarifying them.
Step 2: Identify and Merge Exact Duplicates
Before doing any thematic grouping, pull out items that say essentially the same thing. If three people wrote 'stop having meetings without agendas,' stack those cards together. Count the duplicates — the number itself is useful data (three people independently flagged the same issue).
Only merge items that are genuinely saying the same thing. 'Stop having meetings without agendas' and 'Start sending agendas before meetings' are related but different — one is a Stop, one is a Start. Don't merge across categories at this stage.
Tip: Write the duplicate count on the merged card (e.g., '×3'). During voting, seeing that three people independently raised the same concern carries weight.
Step 3: Create Affinity Clusters Within Each Category
Now group related (but not identical) items into themed clusters. Physically move sticky notes near each other on the board, or use your digital tool's grouping feature. Work through one category at a time.
For example, within the Stop column, 'stop Slack messages after 6pm,' 'stop weekend emails,' and 'stop scheduling 5pm meetings' might cluster under a theme like 'after-hours boundaries.' Let the team suggest groupings — the facilitator proposes, the team confirms or adjusts.
Give each cluster a short, descriptive label that the whole team agrees on. The label should capture the theme without losing nuance. Aim for 3–5 clusters per category. If you have more than 6 clusters in a single category, some of your clusters may be too granular.
Tip: If an item could fit in two clusters, ask the author which theme feels more accurate. Don't put items in multiple clusters — it dilutes the voting signal.
Step 4: Conduct Dot-Voting Across All Clusters
Give each team member a fixed number of votes — typically 3 to 5 dots total (not per category, total across all categories). The constraint is critical: it forces people to make real tradeoffs.
Participants place their dots on the clusters they believe would have the highest positive impact on the team if acted on. They can spread votes across multiple clusters or stack multiple votes on a single cluster they feel strongly about.
For remote teams, use the built-in voting features in tools like Miro, FigJam, or Retrium. For in-person teams, use physical dot stickers or marker dots. Set a 2-minute timebox — voting should be quick and intuitive, not agonized over.
Tip: Have everyone vote simultaneously (or in private for remote teams) to avoid anchoring bias. If the team lead votes first and puts three dots on one cluster, others tend to follow.
Step 5: Tally Votes and Rank Clusters
Count the dots on each cluster and write the totals visibly. Rank all clusters across all three categories from most votes to fewest. This cross-category ranking is important — maybe the most urgent item is a Stop, not a Start.
Read back the top 5 clusters with their vote counts. At this point, it's often clear which 2–3 items the team most wants to address. If there's a tie or a near-tie, a brief 2-minute discussion can break it — but avoid relitigating the entire list.
Tip: Don't ignore low-vote items entirely. A cluster with only 1 vote from the team lead might still signal something important. Note it on a 'parking lot' list for future retrospectives.
Step 6: Select 2–3 Commitments and Define Action Items
From the top-ranked clusters, select no more than 2–3 items to commit to for the next sprint, month, or review period. Fewer is better — teams that commit to 6 actions typically complete zero.
For each selected commitment, define: (1) a specific action — not 'improve communication' but 'add a written agenda to every recurring meeting by Thursday before the meeting'; (2) an owner — one named person accountable, even if the whole team participates; (3) a check-in date — when the team will assess whether this action was taken and whether it helped.
Write these commitments in a shared, visible place: the team wiki, the sprint board, or a pinned Slack message. They should be visible every day, not buried in meeting notes.
Tip: End the session by reading the commitments aloud and asking each owner to confirm. This micro-commitment in front of peers dramatically increases follow-through.
Examples
Example: Sprint Retrospective for a 7-Person Engineering Team
A software team has just finished a two-week sprint. During their Start Stop Continue retrospective, they collected 34 sticky notes: 12 Start items, 11 Stop items, and 11 Continue items. The board feels overwhelming and several items look similar.
The facilitator reads each item aloud. Two Start items ('start writing acceptance criteria before sprint planning' and 'start defining done criteria upfront') are nearly identical — they're merged into one card marked ×2. After clarification, the team clusters the remaining items:
Start clusters: (A) 'Upfront requirements clarity' (4 items), (B) 'Automated testing' (3 items), (C) 'Cross-team syncs' (2 items), (D) 'Documentation' (2 items). Stop clusters: (E) 'Scope creep mid-sprint' (4 items), (F) 'Unnecessary meetings' (3 items), (G) 'Manual deployments' (3 items). Continue clusters: (H) 'Pair programming' (5 items), (I) 'Daily standups format' (4 items), (J) 'Friday demos' (2 items).
Each person gets 4 votes across all 10 clusters (28 total votes in the room). Results: Scope creep mid-sprint (E) gets 9 votes, Upfront requirements clarity (A) gets 7, Pair programming (H) gets 5, everything else gets 2 or fewer.
The team selects two commitments: (1) 'Product owner will lock the sprint scope after planning — any new requests go to backlog. Owner: Maria. Check-in: next retro.' (2) 'Dev leads will draft acceptance criteria in the ticket template before refinement. Owner: James. Check-in: next Wednesday standup.' The Continue item (pair programming) is acknowledged and celebrated but doesn't need an action — it's already working.
Example: Quarterly Marketing Team Review (Remote)
A 12-person marketing team runs a quarterly start stop continue exercise using Miro. They've collected 52 items across three columns. The remote format makes clustering harder because people can't physically move sticky notes as a group.
The facilitator shares their screen and reads items column by column, asking for brief clarifications via chat (not voice, to save time). For clustering, the facilitator proposes groups and the team reacts with thumbs up/down emojis. This keeps 12 people from talking over each other.
After 15 minutes, they have 14 clusters. The facilitator recognizes this is too many and asks: 'Can we combine any of these?' The team merges 'social media calendar' and 'content planning' into one cluster, and 'vendor management' and 'agency communication' into another. Down to 12 clusters.
Each person gets 4 anonymous votes via Miro's voting feature. The top result is clear: 'Stop saying yes to every stakeholder request' with 15 votes — nearly double the runner-up. The team creates one strong commitment: 'All campaign requests must use the intake form and be prioritized by the marketing ops lead weekly. Requests without the form are returned. Owner: Lisa. Starts Monday. Check-in: next month's team meeting.'
Best Practices
Limit commitments to 2–3 items maximum per retrospective cycle. Research on implementation intentions shows that fewer, specific commitments outperform long wishlists. If the team pushes for more, ask: 'Which of these would you cut to make room?'
Always vote across all three Start, Stop, and Continue categories simultaneously rather than voting within each category separately. Cross-category voting reveals the team's true priorities — sometimes all 3 commitments come from the Stop column, and that's valid.
Use silent, simultaneous voting rather than sequential or verbal voting. This prevents the HiPPO effect (Highest Paid Person's Opinion) from skewing results and gives introverted team members equal influence.
Preserve the original sticky notes or cards even after clustering. The specific language people used often reveals nuance that the cluster label obscures. You may need to revisit the originals when defining action items.
Track the vote distribution, not just the winner. If a cluster gets 12 votes and the runner-up gets 11, that's essentially a tie. If the winner gets 12 and the runner-up gets 3, there's a clear mandate. Adjust your discussion time accordingly.
Revisit the previous retrospective's commitments before starting any new categorization. If last session's actions weren't completed, the team needs to understand why before adding new ones. This builds accountability into the start stop continue exercise rhythm.
Common Mistakes
Trying to discuss and debate every item before voting, turning a 30-minute exercise into a 90-minute ordeal.
Correction
Use the read-and-clarify phase only for comprehension, not evaluation. Save discussion for after voting, when you only need to discuss the top 2–3 clusters. The vote is the team's discussion — it aggregates everyone's judgment efficiently.
Creating too many clusters (10+ per category) because the facilitator is afraid to group anything together.
Correction
Aim for 3–5 clusters per category. Ask: 'If we acted on this theme, would it address most of these individual items?' If yes, they belong together. Overly granular clusters scatter votes and produce no clear winner.
Letting the facilitator or team lead decide which clusters to act on instead of using the vote results.
Correction
The whole point of dot-voting is democratic prioritization. If the leader overrides the results, team members will stop engaging honestly in future retrospectives. If the leader has critical context the team lacks, they should share it before voting, not after.
Committing to vague themes ('improve our deployment process') instead of specific actions ('add a pre-deployment checklist to the CI pipeline by next Friday — owned by Sarah').
Correction
After selecting the top-voted clusters, spend 3–5 minutes per cluster converting the theme into a SMART action item: specific behavior, named owner, and concrete deadline. If you can't make it specific, it's not ready to be a commitment.
Giving unlimited votes so every cluster gets at least one dot, defeating the purpose of prioritization.
Correction
Use a strict vote budget: 3–5 total votes per person across all categories. The scarcity forces genuine tradeoffs. A good heuristic is total votes = (number of clusters) ÷ 3, rounded down.
Other Skills in This Method
Facilitating Start Stop Continue Retrospectives
How to plan, run, and timebox an effective Start Stop Continue retrospective session with your agile team, from setting ground rules to closing with action items.
Using Start Stop Continue in Performance Reviews and 1-on-1 Meetings
How to adapt the Start Stop Continue framework for individual performance conversations, manager check-ins, and self-reflection outside of team retrospectives.
Running Start Stop Continue as a Team Icebreaker Activity
How to use a lightweight Start Stop Continue exercise as a warm-up or icebreaker to build psychological safety and normalize giving feedback in new or forming teams.
Building Start Stop Continue Templates and Worksheets
How to design reusable templates, worksheets, and digital boards (Miro, Google Docs, Notion) that structure the feedback collection and make sessions efficient.
Writing Effective Start Stop Continue Feedback
How to write clear, specific, and constructive feedback items in each category—avoiding blame, staying behavior-focused, and making each item immediately actionable.
Crafting Actionable Start Stop Continue Questions and Prompts
How to write and select targeted questions for each Start, Stop, and Continue category that elicit specific, constructive, and actionable feedback rather than vague responses.
Frequently Asked Questions
How many votes should each person get in a start stop continue exercise?
A good rule of thumb is to give each person 3–5 total votes, or roughly one-third of the total number of clusters. The constraint is essential — if people can vote on everything, you learn nothing about relative priority. For a typical retrospective with 8–12 clusters, 3–4 votes per person works well.
Should we prioritize within each category separately or across all three?
Vote across all three Start, Stop, and Continue categories simultaneously. This reveals the team's true priorities without artificial balance. Sometimes the most impactful actions are all Stops, and forcing one commitment per category wastes a slot on a lower-priority item.
What if the team can't agree on how to cluster feedback items?
If two people disagree on whether an item belongs in Cluster A or Cluster B, ask the original author which theme feels more accurate. If the item genuinely spans two themes, let the author choose one placement — don't duplicate it. Clustering doesn't need to be perfect; it just needs to be good enough for meaningful voting.
How many action items should a team commit to after a start stop continue exercise?
Commit to no more than 2–3 action items per cycle. Teams that commit to 5+ actions typically complete none of them. It's better to fully execute 2 improvements than to half-attempt 6. You can always tackle the next-highest-voted items in the following retrospective.
Can I use this prioritization method for a start stop continue exercise with just 2 people?
Yes, but skip dot-voting since there aren't enough voters for meaningful aggregation. Instead, after clustering, have each person rank their top 3 clusters, then discuss where your rankings overlap and differ. Focus your commitments on areas where both people agree on high impact.
What do I do with low-priority items that didn't get votes?
Don't discard them. Add unvoted items to a 'parking lot' document that's reviewed at the start of the next retrospective. Sometimes items that seemed minor become urgent over time. The parking lot also signals to contributors that their feedback was heard even if it wasn't prioritized this cycle.