An Experiment to Improve Team-wide Feedback

March 3, 2018

I’ve been working at Pivotal Cloud Foundry for the last 16 months. During that time, I’ve mostly been an engineer, but am currently putting on a Product Management hat for a little while! More on that in a future blog post.

Pivotal is a company that spends a lot of time thinking about what good feedback looks like. Not as in feedback from customers or end users, although that is also extremely important; I’m talking specifically about peer-to-peer feedback that is intended to help individual contributors grow as engineers, designers, product managers, technical writers, and so on.

We have lots of rituals and tools for keeping feedback at the forefront of our regular work cadence – way more than I’ve seen elsewhere so far. For example, during the first week that someone joins, odds are they’ll learn what “TASK” (Timely, Actionable, Specific, Kind) feedback looks like, either during HR on-boarding or team on-boarding. We have an internal web application, called “Feedback” (very creative, I know) that basically provides a nicer interface than Gmail for giving someone else feedback. Some Labs teams also schedule “speed dating” feedback sessions, especially with larger client teams. Throughout offices, volunteers who have longer-term context about the company culture will also conduct “Core Practice” sessions, with common topics being delivering and receiving feedback.

Those things vary a bit from one office to the next. One constant is that everyone has a manager, and a common activity that managers will conduct is to have 1-to-1s with other members on the team to collect feedback. This feedback is delivered on a regular cadence, often fortnightly, but it depends on the individuals’ preferences.

I recently worked on a Cloud Foundry Foundation team. Foundation teams often have engineers from different companies working together on Open Source codebases. My specific team was working on the Cloud Foundry Services API, which is the point of integration between the platform and service brokers. We had three engineers from Pivotal, and one engineer from SUSE named Jen who was based out of Nuremberg. Jen brought up at one point that she had a feedback review session with her manager coming up, but she hadn’t received much formalized feedback from our team. It seems obvious in retrospect that as Jen didn’t work at Pivotal, she couldn’t have a Pivotal manager constantly gathering feedback on her behalf!

This got me thinking. Over one weekend I came up with an experiment to try to address this deficit. I asked myself, “Why is it the case that only managers should gather feedback on behalf of their reports?”, and “What if we spent a week playing the role of managers and did this exact exercise, for someone else on the team?” In a team that has a healthy degree of psychological safety and shared understanding of empathetic behavior, I could come up with no compelling reasons not to at least try.

When we finally found a week when no one was traveling, this was the email I sent to the team:

As mentioned at standup this morning, we have a retro item from a long time ago that we’d create time to give each other feedback. This week, we’ll be giving each other the gift of feedback by actively collecting feedback on behalf of another team member.

I wrote a sophisticated algorithm to randomly generate who will be collecting for whom:

irb(main):001:0> [‘denise’, ‘matt’, ‘jen’, ‘sam’, ‘alex’].shuffle => [“sam”, “alex”, “matt”, “jen”, “denise”]

To prevent inner loops, we should each collect feedback for the person after us. So,

  • Sam is getting feedback for Alex
  • Alex for Matt
  • Matt for Jen
  • Jen for Denise
  • and Denise for Sam.

Does that make sense?

We have all week to gather insights from the rest of the team, and on Friday right after lunch (13:30 - 15:00) we can all deliver the feedback we’ve collected and synthesized.

As you’re soliciting, offering, and synthesizing feedback, try to keep in mind the TASK model: Feedback should be Timely, Actionable, Specific, and Kind.

It took a day or so for people to get into the swing of things. Some people preferred collecting feedback over email or Slack; others booked in 15-minute chats with others on the team. I personally found the in-person exchanges more efficient.

The team decided to block out 30 minutes on Friday after standup for everyone to synthesize the feedback they gathered, and 1.5 hours in the afternoon to deliver it. Feedback delivery sessions were private 1-to-1s.

What did we learn?

Unsurprisingly, that it is really hard to give good feedback to everyone on the team. The fact that Feedback Week was coming up made us all think harder about our interactions with each other in the days leading up to it. Some of us decided to start making time each week to jot down feedback for teammates.

For people who collected on behalf of a Product Manager or anchor, it was also a good opportunity to read up on what those roles actually entail. I took some time to research the anchor role by digging through our internal Google Drive.

The light social pressure created by watching everyone else gather feedback, combined with the constrained timeframe of one week, motivated (at least) me to think ahead about making sure every feedback-gathering conversation was as productive as possible. We learnt a lot about how to improve our own feedback delivery methods from watching others do it. It’s uncommon for non-managers at Pivotal to collect feedback for others, and I gained more empathy for the managers!

Would we do this again?

I would recommend this exercise, but because it’s relatively heavy on time investment, perhaps only once every six months or every quarter. The feedback we each received was valuable, actionable, and thoughtful.

It’s worth pointing out that the bigger the team is, the more time this will take. 4-6 people is a good number because you can deliver everyone’s feedback in 1.5 hours or less, assuming each delivery session takes about 30 minutes. If you have a bigger team, it may be worth splitting the team into subgroups to assign collector/recipient pairs. (The reason I wanted to avoid internal loops during the assignment was to maximize the potential to parallelize the feedback delivery sessions on the last day.)

I had originally believed that this experiment required a high degree of social familiarity among the team to work. After I rotated off this team, and two new team members from yet another company joined, the team ran Feedback Week again – with great success. I still believe that psychological safety matters, but I’m now more confident recommending this experiment for teams that are still storming and norming. If you try this on your team, please tell me how it goes via email or Twitter!