ApplyingML

  • Papers
  • Resources
  • Mentors

Writing Weekly 15-5s

A 15-5 is a weekly update that takes 15 minutes to write and 5 minutes to read. I learned about it from Gergely Orosz and Will Larson. (They call it the 5-15 but I’ve remembered it as 15-5 because the 15-min writing comes before the 5-min reading.) I’ve been writing weekly 15-5s over the past few months and found it helpful for myself and the team.

First, it provides visibility in a scalable way. My intended reader is my manager and team, though other stakeholders have found it useful enough and subscribed. As a result, we spend less time on status updates during 1-on-1s and more time on strategic issues.

Second, it tracks how productive (or unproductive) I’ve been. Each week, I log the main chunks of work and outcomes. This includes blockers, bottlenecks, and any friction in general. Over time, patterns emerge on what’s slowing down the team most.

Third, it documents all artifacts I’ve created or reviewed. In my work, this is mostly code (e.g., repos, pull requests), documents, and experiments and metrics. I also include documents I’ve reviewed and provided feedback on. A handful of docs each week leads to 200+ documents a year! Having a bullet point on the topic, author, and a link to the document makes it easy to retrieve.

Forth, it earns trust. Providing visibility helps the team know what I’ve shipped and am blocked by, and allows them to provide feedback or course correct. This helps prevent data scientists from going down rabbit holes that don’t provide value to customers or the business, something we’re occasionally guilty of.

To get started, just set up a personal page on your org’s internal wiki or confluence. At the end of each week, summarize your work as a new 15-5 entry. It’s also useful to document key decisions, such as agreement that a feature is out of scope or on the tech stack.

While reflecting on the past week, my key tasks for next week naturally come to mind, so I add them too. They’re usually no more than three bullet points.

An optional step is to copy the 15-5 and send it via email to your manager and whoever else is subscribed. I build on the email chain so the most recent email has all the past 15-5s. They don’t necessarily have to read it, but if they want an update, it only takes 5 minutes.

Remember, it should take only 15 minutes (or less) to write. The less effort it takes, the easier it is to make it a habit. Making it 5 minutes to read also increases the likelihood that someone will read it. As a reference point, my 15-5s are mostly bullet points and images. Here are three weeks of hypothetical 15-5s from someone in an IC role:

Week 1
- Wrote first draft of design doc for Project Alpha
- Ran initial experiments using proposed methodology
- Started as onboarding buddy for Xander, a senior AS in X Org
- Reviewed the following docs:
- Project Bravo (Babbage): Link to doc
- Project Charlie (Carmack): Link to doc
- Project Delta (Dijkstra): Link to doc
Key tasks for next week
- Complete experiments and add results to design doc
- Finalize draft and review with leadership
Week 2
- Reviewed Project Alpha design doc with leadership team
- Positive feedback and aligned on methodology & system design
- Outstanding discussion on serving via Yankee or Zulu
- Wrapped up experiments on proposed methodology
- Metrics and charts available in this cloud doc
- Onboarded to data pipelines package
- Reviewed the following docs
- Project Echo (Edelman): Link to doc
- Project Foxtrot (Fanning): Link to doc
Key tasks for next week
- Start developing data pipelines for Project Alpha
- Decide on Project Alpha serving via Yankee or Zulu
Week 3
- Developed pipelines and published a pull request
- Discussed with tech team and agreed on serving via Zulu
- It met our needs with lower operational and $$$ cost
- Reviewed the following docs
- Project Golf (Gosling): Link to doc
- Project Hotel (Hopper): Link to doc
Key tasks for next week
- Respond to comments and merge data pipelines PR
- Onboard to Zulu and run simple loadtest with latency metrics

Read more guides?


© Eugene Yan 2022AboutSuggest edits.