Wednesday, November 24, 2010

Agile Retrospectives - a Rising Patton Fusion

The last session of Agile Vancouver 2010 was a unique opportunity to watch Linda Rising conduct a conference retrospective with the Agile Vancouver organizers. It was interesting to watch how she facilitated and I wrote down some of her techniques so that I could try them out. The following day during the tutorials Jeff Patton led us through a mini retrospective with his own interesting twists based on his story mapping experience. What follows is the fusion of their ideas.

Note: This retrospective works nicely using index cards that can easily be sorted and grouped around a common table. If you are using walls and stickies, you can adapt where required.

Step 1 - Set the tone:
Recite Norm Kerth's Prime Directive (Linda was able to recite this by memory):

"Regardless of what we discover, we understand and truly believe that everyone did the best job they could, given what they knew at the time, their skills and abilities, the resources available, and the situation at hand."

After reciting, ask each team member one by one if they agree to uphold this statement during the meeting and to avoid blame. A simple verbal "yes" indicates agreement. The verbal agreement is a simple influencing strategy or pattern that helps set the tone for the retrospective.

Step 2 - Framing the Retrospective:
The facilitator tells a story of a co-worker meeting you in the hallway of your company. The co-worker asks: "I know you were on project [X], how was that?". Each person responds with "It was great because...". Instead of speaking the answers out loud, give each person 3 index cards and have them write their answers down in silence. This allows input from each person regardless of personality type and keeps your team member's answers from influencing your own.

When each person has filled out their 3 cards, they read each one out loud and place them on the table. Once all cards are on the table, ask the group to silently group the answers together. Cards that are similar should be close to one another and cards that are different should be farther apart. According to Patton, the simple reason for doing this in silence is because it allows the work to happen quickly and without much discussion. In our exercise, we found this to be true.

Now that the good things are grouped together, take a different coloured index card and have the group summarize each grouping with a new card. For example, summarizing items like "team worked well together", "Bob collaborated to help me with my task", "Team rallied to complete the stories together" might be summarized with a card called "Great team work".

Step 3 - Do Differently:
Now that we have acknowledged the good things about the last period, remind the team that it is only a perfect project if they would do that project or iteration again in exactly the same way. In reality there is always something we would do differently. Ask the team to silently fill out 3 more index cards with what they would do differently. They may not write cards that associate blame, describe what wrong, or try to problem solve. In this part of the retrospective we are only identifying what we would do differently.

Once everyone has completed their cards, we again read them aloud as we place them on the table, group the cards in silence, and summarize with a different coloured index card. As the cards are read or summarized, the facilitator may need to remind the team to refrain from problem solving or directing blame.

Step 4 - Voting:
The next step is for the team to agree on which items on the "Do Differently" list are the most important. To keep the voting impartial and independent, have each team member write their top 3 items on an index card and hand the cards to the facilitator. The facilitator then tallies the votes and records the totals on the summarized cards. An alternative is to use dot voting, but I've found that dot voting can be 'gamed' too easily and that initial dots influence those who vote later (group think).

Step 5 - Experiments:
Using the top 1 or 2 voted items, ask the group to split into groups to discuss them - 1 group per item. Instruct the groups to discuss small experiments or tweaks that the team could try in the next iteration. If you practice frequent retrospectives, this discussion does not need to be about problem solving or even root cause analysis. The goal is to quickly agree on small experiments to try to resolve the issue you are discussing. For those dedicated to root cause analysis this may be a problem, but give it a try and do what works for your team while avoiding blame or delving deep into problem solving. This part should be time boxed to 10 or 15 minutes maximum.

After deciding on the experiments, have one person from each group present the idea to the group. This idea needs to be included in the backlog for the iteration and the team (not an individual) commits to completing the experiment during the iteration and examining the results.

Step 0 - Review Experiments:
At the beginning of the next retrospective, add a new step at the beginning to discuss your experiments to see how effective they were and use this information as input into future experiments.

Other notes:
If you are doing a retrospective for a larger period of time, then consider starting the retrospective by building a timeline of events. For more info, check out this blog: http://www.thekua.com/rant/2006/03/a-retrospective-timeline/

Thanks Linda and Jeff for sharing your methods and ideas.