Junk Drawer
For all those little papers scattered across your desk
Junk DrawerFor all those little papers scattered across your desk
This is my shortlist of “Day 1 questions” for new teams to help me get up to speed. In no particular order:
Examples drawn from real work.
What do we do? How do we accomplish it?
This could be things like “manage case lifecycles via tenant-configured platform services,” “keep data flowing with automatically managed Kafka instances,” or “provide cost recommendations to clients by analyzing data ingested from their systems.”
Ideally this comes with a domain breakdown of key terms. Examples: tokenized plastic/PAN; drain/kill/rehydrate; Snowflake contract and connection details.
What is the budget? I want to know the total expectations and by subsystem (we should have those, right?). How do we know if we are (not) meeting those budgets? What characteristics do we measure (mean, stddev, IQR, p90, shape of distribution1, etc.) and what are they? What are our target specs end-to-end (our hardware and our clients’ hardware, with a focus on the clients).
See Performance inequality on the web for what happens when we get this right (or wrong).
Scale factors are the things whose increase will be felt across the system. Examples might be number of customers, objects, transactions, queries, etc.. How have we designed to meet changing needs with scale? How do we know if we are (not)?
We may be planning to reimplement and rearchitect if we can’t meet new scale requirements, and that’s an acceptable answer as long as we understand the cost and know when (preferably before) we hit those points.
How do we know what our customers need? How do we know if those needs are (not) met? How do we know when something goes wrong and what it is? How do we get side-by-side feedback? How is customer engagement or success incorporated back into the product?
How do we communicate (meetings, chat, email)? How do we track work-to-do, work done, priorities, and roadblocks?
How do we develop and deploy code or land it in main? This includes
We must not commit crimes of statistics, after all. Don’t assume unimodal, Normal distributions. Please provide a measure of spread with a measure of central tendency. ↩