CTO Craft Con Nov. 2023: Building a solid foundation for scalable structures, processes and values

- Drawing on his extensive experience of running engineering groups at unicorns like Twilio and Checkout.com, as well as a range of successful startups, Serge will share his lessons learned on how to best architect your tech and engineering function, and how he is looking to implement these lessons at carbon intelligence startup, Sylvera
- Going beyond ‘best practices’, Serge will share his approach to moving from surviving to thriving as a startup and establishing critical foundations to enable future scalable growth
- Also exploring how scaleups and larger organisations in the midst of growth can use this approach to identify potential scale issues ahead of time and how to get back on a positive trajectory after hitting roadblocks
- Discussing the benefits of a ‘build and own’ approach to minimise friction, increase efficiency and develop trust, and the power of creating a data-driven, transparent and accountable engineering culture from Day 1
- While there’s no such thing as a one-size-fits-all roadmap, and the final vision may differ from organisation-to-organisation, Serge will share the key jigsaw pieces you should consider implementing, realigning or re-implementing to develop a harmonious and high performing tech and engineering function

Transcript

1. BUILDING A SOLID FOUNDATION FOR SCALABLE STRUCTURES, PROCESSES AND VALUES… Serge Kruppa CTO, Sylvera … and succeeding as a startup!
2. One of our products CPaaS ~450 → ~5000 employees IPO Sr. Director Eng. Fintech ~2000 employees Series D VP Eng. Climate Tech ~160 employees Series B CTO Me SERGE WHO?
3. WHAT ARE THE ODDS OF FAILURE ? A. 99% B. 90% C. 78%
4. 90% OF ALL STARTUPS 7.5 OUT OF 10 VENTURE BACKED STARTUPS FAIL -Startup Genome -Shikhar Ghosh 1% OF ALL STARTUPS BECOME UNICORNS -CB Insights FAIL
5. HOW DO WE THE ODDS? BEAT … and achieve startup success!
6. “You can’t looking forward; you can only connect them looking backwards.” CONNECT THE DOTS -Steve Jobs VUCA Volatility Uncertainty Complexity Ambiguity
7. “ is the philosophical belief that the way to truth is through experiments and empiricism.” -Wikipedia EXPERIMENTALISM Mad Agile scientist
8. WE HAVE BEEN CONNECTING THE DOTS SINCE MAMMOTH HUNTING TIMES! BETTER SCENARIO PLANNING = BETTER ODDS OF EATING TONIGHT DAG? CAUSE → EFFECT
9. CYNEFIN Image: IT Revolution, Cynefin: Four Frameworks of Portfolio Management Are there any we can apply to the startup domain? BEST GOOD PRACTICES
10. 150K VC-BACKED STARTUPS $643B INVESTED IN 2021 -Crunchbase -Pitchbook -Startups built on AWS DATA THERE IS
11. A Deep Dive Into The Anatomy Of Premature Scaling Digital Scaleups The Organization Life Cycle: Integrating Content and Process Steven H. Hanks … and a gratuitous cat picture :) TENS OF THOUSANDS OF PROFESSIONALS Twitter/X Instagram Intercom THOUSANDS OF STARTUPS SURVEYED + +
12. EARLY STAGE STARTUP LATE STAGE STARTUP DISCOVER VALIDATE GROW OPTIMIZE
13. EARLY STAGE STARTUP LATE STAGE STARTUP SEARCH FOR A REPEATABLE AND SCALABLE BUSINESS MODEL SEARCH FOR PRODUCT / MARKET FIT UNDER CONDITIONS OF EXTREME UNCERTAINTY Dominant Problem
14. EARLY STAGE STARTUP • Simple org structure • Small team (< 50) • Founder/s supervising work • General job assignments • Few formal systems • Ad hoc planning • Take on pragmatic tech debt • Build PoCs and MVPs • Initial engineering foundations (CI/CD, etc.) • Highly collaborative CHARACTERISTICS Fix dominant problem Jeopardise scaling EARLY LATE
15. SCALING PAINS EARLY SIGNALS • There are too few good managers • People are stretched too thin • People spend too much time "putting out fires” • There’s too little follow up to plans and often things don't get done • People lack understanding about where the firm is headed • The founder becomes a bottleneck • People are not aware of what other people are doing
16. DELAYED SCALING IS RISKY
17. Fix dominant problem Danger of premature scaling LATE STAGE STARTUP • Fast traffic growth • Feature release rate increase • More production capacity • Rapidly expanding headcount • Emerging formal systems and org structure • Formation of functional teams • New management layers • Quest for more efficiency • Start paying down tech debt • Hardening to handle scale CHARACTERISTICS EARLY LATE
18. PREMATURE SCALING KILLS OF ALL EARLY STAGE STARTUPS 70% -Startup Genome Image: Architectural Digest, Burj Al Babas
19. Zhang Ruimin,CEO of Haier,world #1 home appliance maker composed of 4000 internal startups “THE LARGE COMPANY DISEASE” As companies grow, they tend to lose efficiency. They become less innovative, less responsive to external stimuli. They lose their customer-centricity, find that they have fewer people close to the buyer, and risk becoming bloated.
20. 1 2 3 4 5 6 7 8 9 10 Weeks < 25 25-75 75-200 > 200 Developers 1-2 2-4 3-6 6-10 LEAD T ME FOR CHANGES Exclusive data courtesy of
21. Instagram Linear Whatsapp 37 Signals 6 15 50 80 SMALL IS BEAUTIFUL NUMBER OF ENGINEERS 100-150k customers 1B users 1000 customers 30M users Acquisition Now
22. HOW TO MITIGATE SCALING RISKS ? EARLY STAGE STARTUPS
23. EARLY STAGE STARTUP Building a product without problem / solution fit Systematically experiment to validate assumptions Tracking metrics that fit the business model Lean product development Prioritise delivering customer value Developers far removed from the customers and their problems Building the wrong thing, too many "nice to have" unused features FAILURE FACTOR CAPABILITY CONTRIBUTING FACTOR CAUSES MITIGATES Failure to fix the dominant problem
24. Hiring too many people too early Hire a generalist engineering team, made of passionate and curious people Keeping a flat org structure with minimal management overhead Investing into scalability of the product before product / market fit Executing with a regular feedback loop Tracking metrics that fit the business model Keep financial discipline (constraints) to stay focused Over-optimising something that simply shouldn’t exist EARLY STAGE STARTUP Building for millions of users before having thousands of users Premature scaling
25. HOW TO MITIGATE SCALING RISKS ? LATE STAGE STARTUPS
26. Accumulation of tech debt, often due to experiments and shortcuts Automate toilsome development and operational tasks LATE STAGE STARTUP Feature factories with no balanced Product / Tech attention Subtract what doesn’t work anymore Building strong and stable teams Developer productivity has dropped as organization grows Failure to fix the dominant problem Slow cycle time, long release cycles and lead time for change Handovers cause waste and break virtuous feedback loops Continuous software delivery practices Westrum "generative" culture Lean management End-to-end ownership and DevOps culture
27. LATE STAGE STARTUP Current organization structure doesn’t support current scale Failure to fix the dominant problem Ineffective local optimisations Reorg introducing value-stream teams Remove dependencies between teams (loose coupling) Hit the brakes in order to scale faster later Complicated internal team and partner dependencies Lack of systems thinking Efficiently executing the irrelevant Executing with a regular feedback loop Lack of constructive feedback within the team Tracking metrics that fit the business model Lean product development Absence of trust
28. TECH TEAM How is it structured?
29. Platform value-stream teams that own data products in a data mesh organisation Cross- functional Data Product 1 Data Product 2 Data Product 3 Governance Agile Coaching Quality Ops
30. Data Platform value-stream Web App teams enabled by platform teams Web App | 3 Governance DevOps DevEx API Web App | 2 Web App | 1 Agile Coaching
31. CULTURE Software delivery performance Organisational performance to higher levels of satisfaction PREDICTS LEADS -Accelerate
32. WESTRUM CULTURE Image: IT Revolution, Westrum’s Organizational Model in Technology Organizations The good one!
33. • Wear the customer’s shoes • Ruthlessly prioritise • Seek progress over perfection • Draw the owl • Be frugal • Organise into small teams • Trust is the #1 thing we sell • Think long term • Be humble • Share problems, not solutions • Own it • Stay curious • Do what’s right • Collaborate with kindness SELECTED VALUES For Early Stage For Late Stage You Build It, You Own It!
34. VALUES CAPABILITIES Wear the customer’s shoes FAILURE FACTORS POSITIVE OUTCOMES Draw the owl Organise in small teams Ruthlessly prioritise Trust is the #1 thing we sell Think long term Be frugal Seek progress over perfection Be humble Share problems, not solutions Own it Stay curious Do what’s right, even when it’s hard Collaborate with kindness INFLUENCE POSITIVELY + MITIGATE MITIGATE Lean product development Continuous software delivery practices Westrum "generative" culture Building strong and stable teams Remove team dependencies (loose coupling) Outsource undifferentiated heavy lifting Automate toilsome development and operational tasks = - - Accumulation of tech debt Complicated internal and partner dependencies Overcomplicated distributed architecture Developer productivity has dropped Building a product without problem / solution fit Efficiently executing the irrelevant Hiring too many people too early
35. THE ROAD TO HELL IS PAVED WITH GOOD INTENTIONS -Proverb
36. CANNOT EFFECTIVELY ENFORCE THEMSELVES. THAT’S THE JOB OF RITUALS. SOMETHING THAT AMAZON CALLS VALUES MECHANISMS was heavily influenced by Amazon
37. TEAM 1:1 Skip Level 1:1 All Hands Meeting BUSINESS Quarterly Business Review Bi-Weekly Business Review PRODUCT … and planning! PR/FAQ TECH Incident Post-Mortem Bi-Weekly Metrics Review Architecture Decision Record SELECTED MECHANISMS
38. IF YOU CAN’T MEASURE IT, YOU CAN’T IMPROVE IT. -Lord Kelvin
39. DORA leading metrics measure software delivery performance and can be controlled and changed • Deployment Frequency • Lead Time for Changes • Mean-Time-To-Restore • Change Fail Rate • Reliability POSITIVE OUTCOMES POSITIVE OUTCOMES
40. THE How to succeed as a startup? VALUES MECHANISMS INPUT METRICS CAPABILITIES BIG PICTURE
41. VALUES CAPABILITIES FAILURE FACTORS POSITIVE OUTCOMES INFLUENCE POSITIVELY + MITIGATE MITIGATE = - - MECHANISMS APPLY INPUT METRICS FOCUS + THE BIG PICTURE
42. https://nakedmanager.substack.com/ THANK YOU! FIND ME ON SUBSTACK AS THE NAKED MANAGER