The Value Fly Wheel 2023 DDDesign Taiwan Conference

Slide 1

Slide 1 text

The Value Flywheel Effect 那些藏在價值飛輪下你該知道的策略 Kim, Kao Co-organizer, DDDesign Taiwan Sr. Developer Specialist Solutions Architect, AWS

Slide 2

Slide 2 text

No content

Slide 3

Slide 3 text

Slide 4

Slide 4 text

Slide 5

Slide 5 text

Slide 6

Slide 6 text

No content

Slide 7

Slide 7 text

Slide 8

Slide 8 text

7 ARR ? vs RDD ?

Slide 9

Slide 9 text

8 Steps Forward Consistent with Hedgehog Concept Flywheels Builds Momentum Accumulation of Visible Results People Line up, Energized by results Jim Collins – Good to Great

Slide 10

Slide 10 text

The Value Flywheel Effect LONG-TERM VALUE CLARITY OF PURPOSE PHASE 3 PHASE1 PHASE4 PHASE2 Join Business & Technology Strategies CHALLENGE & LANDSCAPE NEXT BEST ACTION Step 1 – CLARITY OF PURPOSE $MRR Jun, 2022 Jun, 2023 Get well plan? Revenue dropped by 60% Inactive members + 40% Termination of supplier collaboration + 30%

Slide 11

Slide 11 text

PHASE 1 – CLARITY OF PURPOSE - Raise the # of active members - Extend the contract - Net new leads/engagement - Recruit new supplier with structural campaign CEO North Star Metrics § Never embrace lagging indicator CLARITY OF PURPOSE PHASE1 Join Business & Technology Strategies

Slide 12

Slide 12 text

Lagging Metric North Star Framework Suffering in execution • Too focus on critical metrics • Difficult to choose appropriate metrics • Looking for short-term win • Lack of whole picture • None concrete execution plan 11

Slide 13

Slide 13 text

Leading Metric Impact Mapping Key Characteristics 12 • Strategy planning tool • Visualize the goals, actors • Through behaviors impact to identify potential results • Connect all stakeholders to execute tasks and deliver results

Slide 14

Slide 14 text

PHASE 2 – CHALLENGE & LANDSCAPE - I can only deal with on-premise system - Front end is the best - Agile? Are you kidding me ? - Take care the worries & concern - Enable & coach through pre-production journey - Embrace diverse stakeholders - Involve in Sociotechnical systems § Psychological safety § A sociotechnical systems view § Enable empowered engineers ENGINEERS PHASE2 Join Business & Technology Strategies CHALLENGE & LANDSCAPE

Slide 15

Slide 15 text

Sociotechnical • Define the product from scratch • Get early stage user feedback • Enhancement based on critical request • Impact the overall architecture design • Expect to gain great customer experience • Update strategies

Slide 16

Slide 16 text

PHASE 3 – CHALLENGE & LANDSCAPE - NO CI, that’s all - Lack of documents, knowledge leaking - We do what we want - RDD - Take care the worries & concern - Enable & coach through pre-production journey - Embrace diverse stakeholders - Involve in Sociotechnical systems § Code is a liability § System is the asset § Align on how you will serve customers PRODUCT LEADERS Next Best Action PHASE 3 Join Business & Technology Strategies

Slide 17

Slide 17 text

PHASE 4 – LONG-TERM Value - Reliability is not reachable - Failure is BAU - Just meet urgent requests - Problem prevention culture – Well Architected - Sustainability as top priority - Next Generation companies ahead § A problem-prevention Culture § Keep a low-carbon footprint § Map the emerging value CTO LONG-TERM VALUE PHASE4 Join Business & Technology Strategies

Slide 18

Slide 18 text

Wrap it up in transparency • Conduct the body of work in documenting the trade-off decision making • Architecture Decision Record as the long-term business growing playbook • Make sure everyone can access in your org

Slide 19

Slide 19 text

Legacy system refactoring strategy Rebuild vs Refactor ?

Slide 20

Slide 20 text

Essential capabilities behind value flywheel 19

Slide 21

Slide 21 text

20 By using a map of a business, we can anticipate change and learn to use it to our advantage ~ Simon Wardley

Slide 22

Slide 22 text

Evolution of the Value Chain Commodity Product Custom Built Genesis Just Innovated Uncertainty Semi-Product Experimented Specific Knowledge embedded Widely Adopted Turn-Key Solution High maturity Plenty of choices Standard & Low-Cost

Slide 23

Slide 23 text

Wardley Mapping Canvas • Identify Users • Verify User Needs • Start from visible to audience • Aware Momentum & Iternia of to-be pace

Slide 24

Slide 24 text

No content

Slide 25

Slide 25 text

Case Study Optimising serverless for BBC ~ Johnathan Ishmael, Lead technical Architect, BBC

Slide 26

Slide 26 text

No content

Slide 27

Slide 27 text

Unexpected Growing Line of Business Scalability challenge 26 https://www.infoq.com/news/2021/01/bbc-serverless-scale/ • Complexity of running such an extensive service. • Unpredictable traffic (e.g. Olympic, WW FG) • Tons of Millions users watching and reading content at once • Combined with essential features – SEO, accessibility, personalization • Multiple channel support – Web, Mobile • Compete with Apple new, Netflix with less budgets

Slide 28

Slide 28 text

Motivation of driving refactor to Serveless 60 million DAU 20K RPS 3x grows QPS / new breaking news story erupts

Slide 29

Slide 29 text

Legacy architecture of BBC 28 • Timeline : During 2020 • 165M unique requests from WW • “Web Core” stack averaged 80K request/min, peaking at 120K/min • Edge traffic over CDN 2.5M PV/min ; 41K qps

Slide 30

Slide 30 text

Success factors 29 https://medium.com/bbc-product-technology/shipping-progressive-web-apps-everywhere-190a421c606a Value Chain Visible invisible Serving context: - 160M unique PV - 80K RPM, Peak 120K RPM - 2.5M PV/Min; 41K QPS over CDN Offer WW content (Sports, RADIO, News, TV) World Wide Access Multiple platform support Unpredictable traffic Serving Content Distribution Network React App Storage Cache Elastic Compute Environment inertia Serverless EC2/Container/Kubernetes Web App

Slide 31

Slide 31 text

Architecture Decision Record 30 https://medium.com/bbc-product-technology/shipping-progressive-web-apps-everywhere-190a421c606a Value Chain Visible invisible Serving context: - 160M unique PV - 80K RPM, Peak 120K RPM - 2.5M PV/Min; 41K QPS over CDN World Wide Access Elastic Compute Environment inertia Serverless EC2/Container/Kubernetes Container (Kubernetes) Lambda - Heavy delay in scaling - Don’t tolerate large bursts of traffic well enough CPU bound compute Lived up < 2s EC2 No persistent requirement - limited by underlying compute instance - Has to scale out to meet the demands of running container - Suitable for long lived, stateful sessions, good to porting - Good to handle short lived traffic - non-persistent, stateless connections - Leave provisioning tasks to providers

Slide 32

Slide 32 text

BBC Serverless architecture 60 million DAU | 20K RPS | 3x growth up QPS/new breaking news story erupts Base scenario : No traditional architecture, go with API Gateway, customzie the entry point from EC2 • Internal handle the traffuc forwarding • Fastest caching operation and fail backs • High network, high memory, low CPU usages

Slide 33

Slide 33 text

Overall gained performance 32 A graph showing the average and p90 response times of our presentation React app serverless functio over a 4 week period. X axis is time, Y axis is duration in milliseconds.

Slide 34

Slide 34 text

Heuristics of practice

Slide 35

Slide 35 text

Goal Alignment Facilitating Problem Domain Resources alignment

Slide 36

Slide 36 text

1. Value Flywheel 2. Wardley Mapping 3. Architecture Decision Record • Persona • North Star • Don’t embrace lagging metrics • Invest on core sub domain • Know the momentum • Mitigate the inertia • Make it traceable • Align with business goals • Take case stakeholders

Slide 37

Slide 37 text

Thank you! 36