Executing the Process: Every Data Science Project Has a Rhythm
There is a rhythm to every data science project, whether it be primarily focused on advanced analytics or media product development. Honoring this rhythm will bring you and our clients success. The figure below outlines the milestone phases for every one of our data science projects. I ask that you understand the birds eye view of the process when estimating levels of effort, and map your process to the milestones delineated below:
For every engagement, I ask that you complete this checklist:
- Phase 1: Discovery and Spec
- Completed the pre-analysis checklist?
- Completed and shared your modeling DAG?
- Synced on the scope with the client after completing this checklist?
- Set data sharing expectations and relayed any relevant shifts to the timeline?
- Finalized data privacy and sharing requirements, along with any trainings?
- Run, examined, and shared the machine-generated codebook?
- Prepared for your phase 1 readout with a dry-run?
- Have you checked any figures or tables against the style guide (see Guru)?
- Followed up after your phase 1 readout to memorialize and set next steps?
- Phase 2: Implementation
- Scheduled your weekly sprint cadence?
- Checked each week to track level of completion and update the client?
- Set your cloud budget alerts?
- Spun up your devops, dataops, and mlops frameworks?
- Serialized model and data transformation objects for replication?
- Documented your spec changes along the way?
- Documented your modular code along the way?
- Conducted “hallway testing” on your initial results with the client?
- Prepared for your phase 2 readout with a dry-run?
- Followed up after your phase 2 readout to memorialize and set next steps?
- Phase 3: Support
- Registered and deployed your microservice (with APIM if relevant)?
- Confirmed logging and alerts are working?
- Tested your end-to-end deployment and testing routine?
- Synced with client about handoff or support agreement?
- Scheduled regular updates with client around monthly maintenance and operations?
- Scheduled opportunities to socialize your work at SDI?
- Scheduled opportunities to socialize your work on the client’s behalf?
Work Product Guidelines
To those joining the team…
Many people joining the team remark with surprise that my delivery requirements are old-school.
If you too have taken note of this, you’re not alone!
As you will find, I believe there is value in organized and thoughtful work rituals.
If that’s old-school, then I will bear the quality proudly!
I detail these work rituals on these pages. Please review them whenever you want a refresher: