DDDEU2023 Keynote - Systems thinking in large-scale modeling

To understand and impact a large-scale environment (a team, an enterprise, a society), we need effective modeling. Over the years DDD has gained a visible foothold in software-centric corporate change agendas.

In my work as a DDD evangelist and sociotechnical architect in large organizations, I’ve lived and breathed DDD to decouple domains and systems, to kickoff greenfield initiatives, to deploy brownfield modernization, and to design reteaming using the Inverse Conway Maneuvre and Team Topologies.

All that hard work has brought me many a pat on the shoulder. But time after time I’ve been surprised by how I/we could make things worse by attempting to make things better. I’ve been frustrated to witness bubbly energy and DDD enthusiasm from event storming workshops and alike evaporate like a deflated balloon. Improvements remained temporary and local. New practices failed to gain widespread adoption. I have to admit with absolute candor that it is so very hard to make lasting impact with DDD, or whatever else technology or tooling du jour.

In this talk, I will share my observations through the systems thinking lens, of what happened in a large change initiative in a big bank, where DDD played a central role. You will hear examples of sociotechnical systems dynamics and episodes about shifting dominance of competing feedback loops. You will be invited to partake in a postmortem forensic of what could have happened, what unfortunately didn’t happen, and the many what-if’s of alternate realities, or pasts. For people new to systems thinking, a positive side benefit can be a sneak peak into systems modeling building blocks and systems archetypes.

As a community, we’ve come a long way modeling software-centric systems, thanks to practice like DDD, architecture and Team Topologies. But what about the runtime dynamics of large human systems in our working environment, especially when such systems are full of fixes that don’t work? As software professionals, can we just tell jokes about SAFe, about policy makers in our company who can copy the Spotify Model, but not paste it? Or should we give the usual shrug – “it’s the system’s fault”?

From a systems standpoint, we are part of the same mess. If we can become better at seeing and naming the elephants in the room, not as random events, but as systemic patterns described through a consistent systems language, then the current messy reality is no longer our enemy but our ally. It can become a generative force for us to identify leverage and sustain influence in large organizational spaces.

I hope this talk will spark interest in system dynamics modeling. It is a call to action for boundary spanners and design-oriented change agents to explore abstractions, heuristics, modeling and visualization techniques to articulate and reason about dynamic system behavior.

How can we architect shared experiences to surface mental models so they can be challenged, improved and no longer stand in the way of change? With the rising initiative fatigue in large enterprises, how do we collectively discover leverage points, design and deploy systemic interventions with long lasting impacts?

Transcript

1. Systems Thinking in Large-​Scale Modeling #Midjourney DDD Europe, June 2023
2. Contextualizing DDD & architecture in large change efforts Conversation facilitator Boundary spanner, connector Idea spreader, change maker About me 01 02 03 Xin Yao Independent consultant Lives in Copenhagen, Denmark Loves gardening, yoga, sci-​ fi, theater, good food, hanging out with friends and family @settling_mud xin.yao.dk@gmail.com @xinyao@mathstodon.xyz /in/xinxin/
3. A tale of transformation at a large bank Systems thinking basics System dynamics modeling System leverage What we will cover today
4. A tale of transformation at a large bank Systems thinking basics System dynamics modeling System leverage
5. Context: A large Nordic bank 3,5 m+ customers (B2C + B2B) 22,000+ employees 4,000+ in engineering (10 countries) +150 years old 2000+ IT systems
6. Organizations as sociotechnical systems Social systems (Complex) Engineered systems (Complicated) [software] [organization] The environment
7. The next great frontier is to understand human systems. ~Jay Forrester
8. Systems building blocks: feedbacks & delays Reinforcing loop Balancing loop Births Population + delay + Deaths + delay -
9. Nonlinearity: Cause and effect can be far apart in time and space linear cause & effect Non-​ linear feedback loop There’s a fundamental mismatch between the nature of reality in complex systems and our predominant ways of thinking about that reality. The first step is correcting that mismatch is to let go of the notion that cause and effect are close in time and space. ~Peter Senge now in one year
10. The eventual consistency time span in a social system is much larger than that in an engineered system. Temporal and spatial systems modeling
11. Graphic: Juliet Young Higher leverage at a different time, and place
12. A tale of transformation at a large bank Systems thinking basics System dynamics modeling System leverage
13. 2020: Reorg & change initiative DDD & API evangelism Cloud & DevOps evangelism Agile coaching Decoupled & reusable APIs Software delivery speed Agility, autonomy + + + + Engineering maturity + + + + Enabling tribe Enabling tribe Enabling tribe Journey/Platform tribes Source: Henrik Kniberg & Anders Ivarsson delay delay delay
14. DDD reinforcing feedback loop Facilitated DDD workshops Eagerness to practice DDD Perceived value of DDD API delivery speed Domain modeling quality Domain understanding + + + + + + delay delay
15. Facilitated DDD workshops Eagerness to practice DDD Perceived value of DDD API delivery speed Domain modeling quality Domain understanding + + + + + + Availability of DDD enabling capacity Utilization of DDD enabling team Demand on DDD workshop facilitators + + + - DDD enabling capacity limits DDD "growth" Reinforcing loop Balancing loop
16. Pressure to increase WIP for DDD enabling team Pressure on DDD evangelists to rush domain discovery Premature convergence DDD enabling team attrition Stress & burn-​ out in DDD enabling team Idle API engineers in enabling team API specification velocity per domain Intervention makes matters worse Facilitated DDD workshops Eagerness to practice DDD Perceived value of DDD API delivery speed Domain modeling quality Domain understanding + + + + + + - - - + + + + +
17. System archetype: Limit to growth/success Source: William Braun - "The System Archetypes" Structure Trend
18. Systems vocabulary System variables Causal links Delays Feedback loops now in one year problem symptom Symptomatic solution Fundamental solution + - + - side effect + - nouns, verbs sentences, scenes Recurring plot lines System archetypes example
19. Why do we keep seeing the same problems recur over time? What leverage do we have in impacting the system's underlying structure? diagnostic use prospective use How do system archetypes help us • Limits to Growth (aka Limits to Success) • Shifting the Burden • Eroding Goals • Escalation • Success to the Successful • Tragedy of the Commons • Fixes that Fail • Growth and Underinvestment • Accidental Adversaries • Attractiveness Principle ... System archetypes
20. Domain-​ driven APIs Challenge to deliver DDD-​ enabled domain APIs Reliance on DDD enabling team + Perceived need to act on DDD skill gaps + Invested time to improve team's own DDD capacity + - + delay Alleviated resource shortage - - root cause Dependency High workload
21. System archetype: Shifting the burden Source: William Braun - "The System Archetypes" Structure Trend
22. Fast flow of change Value delivery rate + Team independence (X-​ as-​ a-​ service) + Stable platform APIs + + Changes in environment (e.g. regulatory requirements) Need for platform rethink (different boundaries?) Need for coordination across teams Risk to miss fast flow KPI (embarrassment) + + Pressure to find a workaround (no interface change) + Accidental complexity (Business debt) + + - Is "flow" a non-​ negotiable goal DDD to discover platform boundaries + Local optimization = negative externality KPI delay constraint = accidental complexity +
23. Slowing down growth gives the balancing loop with limits and delays, time to function ... The optimal rate in a system is far less than the fastest possible growth. ~Donella Meadows Optimizing the whole
24. What is Systems Thinking Interdependencies Living systems Many parts Synthetic holism Analytical reductionism Synthetic holism Analytical reductionism Graphics: Systems Innovation Network Mechanistic thinking Systems thinking
25. A tale of transformation at a large bank Systems thinking basics System dynamics modeling System leverage
26. How do fish explain the water they are in?
27. Systems dynamics modeling: Moving to higher levels of abstraction problem symptom Symptomatic solution Fundamental solution + - + - side effect + - now in one year Structures of which we are unaware hold us prisoner. We can only grasp a complex system through abstract modeling. Reality Models
28. Hurray! Now we can use systems dynamics modeling to understand and impact systems.
29. You are only seeing my view of the world My experiences My data selections My perspectives My assumptions My conclusions My beliefs "The optimist sees the doughnut, the pessimist sees the hole." ~ Oscar Wilde
30. You are not even seeing my view of the world
31. Systems understanding: The Iceberg Model Graphics: Academies For Systems Change API delivery speed = business value APIs in prod pr. month Systems archetype Our team delivered no API this month. Observable Hidden
32. Low leverage point: event-​ driven systems A system input output Event handler Gain achieved Pain relieved events Trends Structures Mental models Credit: Systems Innovation Network Event issue, problem, goal
33. React High leverage point: mental models events Trends Structures Mental models Anticipate Design Transform Credit: Systems Innovation Network
34. A transformation isn't an event. It's a progressive accumulation of habits. -​Andrew Clay Shafer
35. Culture (collective habits) for an organization is like personality for a person. Your energy to try to change it will tend to make it stronger. ~Peter Senge
36. My worst event storming workshop
37. Workshop retrospective You are too in love with your model :-). We don't see forest, only trees You=Me model=event storming
38. We perceive the world through our individual filters Source: Jeff Patton
39. Externalize our thinking visually "Make the implicit explicit" Source: Jeff Patton
40. Combine & refine - integrate perspectives Source: Jeff Patton
41. discussion Reflective conversation Credit: David Bohm, Peter Senge, Chris Argyris advocacy advocacy inquiry Here is how I see it. Here are the facts that support how I see it I know I don't see the whole picture, how do you see it? Here is what I think this data/episode means... Do you see it differently? Oh, that's a perspective I haven't considered Do you see gaps in my reasoning? Here is my view and here is how I arrived at it. dialogue
42. What is it about this situation, and about me or others, that is making open exchange difficult? The questions that got us unstuck What are you most passionate about contributing to in this domain? What is the future we dream of for this domain?
43. What happened after event storming Collective reflection Inquiry & advocacy
44. To DDD or not to DDD Q: Why do you still want to finish the DDD Workshop series? A: You were trying to convince us before. Now we are testing ideas together.
45. To enable is to lead "Will you follow me?" Task-​ oriented Skill-​ oriented Relationally oriented
46. Undercover systems change with DDD DDD consultancy/ facilitation DDD output + + Reflective converations + Generative, adaptive capacity + + The ideal Present (or future?) "DDD change agent"
47. Life is not a problem to be solved, but a reality to be experienced. ~ Søren Kierkegaard
48. A tale of transformation in a large bank Systems basics Systems dynamics modeling Systems leverage What we've covered today
49. Thank you @settling_mud xin.yao.dk@gmail.com @xinyao@mathstodon.xyz /in/xinxin/ Contact me if you need an undercover change agent :-)