Experimentation makes uncertainty testable before a team commits to an approach. In architecture, short, throwaway spikes compare candidate designs; in day-to-day work, reversible local changes can test process improvements. section 11.1 section 15.1.1
Two practical forms serve different scales of uncertainty:
- Run parallel, time-boxed spikes to compare options against shared questions before committing to an architecture. section 11.1
- Try reversible changes in your own workflow, measure the result, and share it without requiring others to adopt it. section 15.1.1
Table 11.1 Notification delivery spike comparison
|
Question
|
Spike A (Redis Pub/Sub)
|
Spike B (Kafka)
|
Spike C (PostgreSQL Outbox)
|
|
Can it deliver 5,000 notifications/sec?
|
Yes, very fast in-memory
|
Yes, with partitioning
|
Moderate, needs polling or NOTIFY
|
|
What happens when a consumer is offline?
|
Messages lost
|
Messages retained on disk
|
Messages retained in the outbox table
|
|
How complex is the operational story?
|
Simple but fragile
|
Cluster management, ZooKeeper/KRaft
|
Familiar, same DB the team already runs
|
|
What does the team already know?
|
Some experience
|
No experience
|
Deep experience
|
|
What's the delivery latency at p99?
|
< 5ms
|
10-50ms
|
50-200ms
|
Compare Redis Pub/Sub, Kafka, and a PostgreSQL outbox across throughput, offline consumers, operations, team experience, and latency. section 11.1
Test a process change
A team tried automated formatting as a pre-commit hook for two weeks. Formatting-related review comments fell to zero, helping reviewers focus on substantive issues. section 15.5.1
Listing 15.1 Eimear's team message after one week of automated deploys
Subject: Staging deploy script: 1 week of data I've been using a script to automate staging deploys this week. Here's what I measured: Before (manual): avg 17 min/deploy, 4 deploys/week = 68 min/week After (script): avg 3 min/deploy, 4 deploys/week = 12 min/week Saved: 56 min/week across my deploys alone The script handles image pull, config update, migration, restart, and Slack notification. Log tailing is still manual. It's in /tools/deploy-staging.sh if anyone wants to try it. No pressure, just sharing what worked for me.
See how a developer shared measured staging-deploy savings while leaving adoption optional. section 15.1.1



















