Metrics - You are what you measure (DevOps Perth)

DevOps is no longer just the concern of cutting edge start-ups in Silicon Valley and is gaining wide scale adoption within established industries. This session focuses on the Metrics pillar of DevOps and explores how we can leverage metrics to drive the software delivery process based on data rather than gut feel and opinions.

Transcript

1. Metrics You are what you measure @robdcrowley robdcrowley
2. Empower you with new ideas to bring your team together! Metrics. What are they? What to measure? How to visualise? Tooling. Some secret sauce to make our lives easier Metrics as a catalyst for cultural change Goals for Session
3. Metrics 101 What metrics are and why we measure
4. To inform To compare To motivate To understand To improve To protect Not: Just because successful companies do! Why do we measure
5. A quantifiable measure of any component or process used to gauge the performance of your business – Application – Business – People – Process – System What is a metric
6. Basis of measurement Assumptions Level and usage Desired trend When to use When to stop How to game Warnings Facets of a metric
7. Red is good, Green is worthless
8. Metrics 201 What should you measure
9. “Lack of direction, not lack of time, is the problem. We all have twenty-four hour days.” - Zig Zigler
10. Pivot: Base conversations on metrics instead of non actionable opinions Use stakeholder input as the basis for your initial set of metrics – “What do you want out of this?” – “How quickly do you want this?” Metrics can help guide us
11. Awesome book, poor title Core Principles – Minimum Viable Product (MVP) – Continuous Deployment – Split Testing (aka A/B Testing) – Actionable Metrics – Pivot – Innovation Accounting – Build-Measure-Learn Loop
13. Acquisition – finding new users Activation – getting users to give your product a try Retention – making sure those users stick around Referral – have your loyal users invite others Revenue – making some money from all this Note: These steps are not strictly sequential Pirate Funnel (AARRR!)
14. Do not focus solely on metrics that focus on reducing errors Focus on achieving excellence and positive outcomes will ensue
15. DevOps State of the Union focuses on these four core metrics – Deployment Frequency – Lead Time for Changes – Mean Time to Recover (MTTR) – Failed Change Rate Example Metrics
16. Fact: We only realise business value from the work we do when it’s released into production The bottleneck or constraint is the limiting factor on our ability to deliver more quickly Strengthening any link other than the weakest is a waste of time and effort Use metrics to identify current constraint Theory of Constraints
17. Metrics 301 How to visualise and share metrics
18. Display all metrics on a dashboard – Business (i.e. Application dropouts) – Dev (i.e. Performance metrics) – Ops (i.e. Web Server CPU Usage) Single URL for all data Make it easy for colleagues to access dashboard Visualisation
19. Use a tool that can handle different kinds of metrics A few standouts are: – StatsD/Graphite/Grafana – Heka – Logstash/Elastic Seach/Kibana Tools
20. StatsD Whisper Carbon Graphite Dashboard Apps UDP TCP HTTP Graphite and StatsD
21. Metrics 401 Potential pitfalls when using metrics
22. Case Study: US Health Service In the 1990s the US Health Service decided to make hospital mortality rates public. What better metric for hospital performance? Observed Behaviour: Best way to improve rating was not to admit critically ill patients in the first place You are what you measure
23. “The actual company values, as opposed to the nice-sounding values, are shown by who gets rewarded, promoted or let go” - Patty McCord
24. Case Study: Enron Enron which went bankrupt from fraud and whose leaders went to jail, had the following company values: Integrity, Communication, Respect, Excellence Where these really the core values at Enron? You are what you measure
25. “In the absence of clarity around an objective, any measure will do” - Bob Paladino Too Many Metrics
26. Depict rosiest picture possible but do not accurately reflect the key drivers of your business – Inaccessible – Not auditable – Not actionable Vanity Metrics
27. Metrics based on output rather than outcomes Aim to maximise outcomes and minimise output – Total lines of code – Number of bugs fixed – Total hours worked – Lines of code per developer Useless Metrics
28. “We’re not doing this data analysis thing just because. We’re gathering data points because we’ve seen it work to great effect in other organisations and industries.” - Anonymous Cargo Cult Metrics
29. Metrics being used for evil rather than for good Confronting this organisational anti-pattern can be a good first step Information sharing is the best defense Break down knowledge silos in the organisation Weaponized Metrics
30. Questions