Trade-offs all the way down

Building Software in the Finance Industry.

1. Trade-offs all the way down Building Software in the Finance Industry Aitor García Rey - @_aitor - aitor.is
2. Sponsors
3. Software is NOT a platonic/ideal object.
4. We are constantly making trade-offs, consciously or unconsciously.
5. Trade-off: Restaurants/Bars ┌────────────────┬────────────────┐ │ │ │ │ │ │ │ │ │ │ │ │ Hospitality │ │ │ workers' life ├ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ┤ │ │ │ │ │ │ │ │ │ ╭───╮ │ │ │ │ X │ │ │ ╰───╯ │ └────────────────┴────────────────┘ Fitting my life style
6. Trade-off: Mars Rovers ┌────────────────┬────────────────┐ │ ╭───╮ │ │ │ A │ │ │ │ ╰───╯ │ │ │ │ │ │ Terrain │ │ │ Adaptability ├ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ┤ │ │ │ │ │ │ │ │ │ ╭───╮ │ │ │ │ D │ │ │ ╰───╯ │ └────────────────┴────────────────┘ Speed
7. Trade-off: 1.5°C Paris Agreement Goal ┌────────────────┬────────────────┐ │ │ │ │ │ │ │ │ │ │ │ │ Resources │ │ │ Frugality ├ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ┤ │ │ │ │ │ │ │ │ │ ╭───╮ │ │ │ │ X │ │ │ ╰───╯ │ └────────────────┴────────────────┘ Convenience
8. Trade-offs expose how most desired features live on a spectrum.
9. Spectrum: Reliability Your Average Massive Space Blog Custom Soft. E-comm Utilities Mission │ │ │ │ │ □────┴────────┬──────┴────────┬─────────────┴───┬───────┴───────┬─┴───□ │ │ │ │ Average Average Bank Nuclear E-comm SaaS Plant
10. Spectrum: Resource Availability Your Average Scale-up Multinational Blog Startup Startup Corps. States │ │ │ │ │ □────┴────────┬──────┴────────┬──────────┴──────┬───────┴───────┬─┴───□ │ │ │ │ Starting Established Series Bank Trad Company Trad Company C/D
11. Spectrum: Performance Your Average Trad Ultra Average API Video Blog biz interaction specialized team Games │ │ │ │ │ □────┴────────┬──────┴────────┬──────────┴──────┬───────┴───────┬─┴───□ │ │ │ │ Bank Average Average Industrial franchise webapp Software interaction
12. BUT consciously making a trade-off implies saying NO to something
13. AND often we end up cargo-culting others' trade-offs.
14. Countermeasure: be aware and constantly reevaluate your context.
15. Our context - Devengo 1.0 • Started as a Salary Advance solution to fight payday-loans. • First commit on Jul 21, 2019. • Rails API + Native apps (iOS & Android) + Custom integrations with HR/Payroll software. • Got smacked in the face by COVID-19 but experienced constant growth from 2021 onwards. • Spent 2 ½ years building the product...just to discover that it took us in the wrong direction.
16. Our context - Devengo 2.0 • We reflected on how we were doing all the heavy lifting of moving money • ...and deploying it (too early) in a very narrow niche. • Pivoted to Payments Provider on 2022 Q2. • Special focus on EU and Instant/Real-time A2A payments.
17. ╭─────────────╮ ╭─────────────╮ │ Originator │ │ Beneficiary │ ╰──────□──────╯ ╰──────△──────╯ │ │ │ │ ╭ ─ ─ ─▽─ ─ ─ ╮ ╭ ─ ─ ─ ─ ─ ─ ─ ╭ ─ ─ ─ ─ ─ ─ ╮ ╭ ─ ─ ─□─ ─ ─ ╮ 1. SCT Inst 2. SCT Inst │ 3. Relay 5. Funds │ Instruction │ │ Transaction │ Transaction │ │ available │ ─ ─ ─ □ ─ ─ ─ ─ ─△─ ─ ─ □ ─ ╯ ─ ─△─ ─ ─ ─□─ ─ ─ ─ △ ─ ─ ─ │ │ │ │ └───┐ │ │ │ │ │ │ │ ╭──────▽──────────╮ │ ╭──▽──────────□───╮ ╭──▽───────□──────╮ │ │ │ │ │ │ │ │ Originator PSP □──┘ │ CSM │ │ Beneficiary PSP ◁─┐ │ │ │ │ │ │ │ ╰───────△─────────╯ ╰────□───□───△────╯ ╰─────────□───────╯ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ ╭ ─ ─ ─ ─ ─ ─ ─ ─▽─ ─ ─ ─ │ │ │ │ └──□ 4a. Communicates OK/KO │ │ │ │ │ ╰ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ │ │ │ │ │ ─ ─ ─ ─□─ ─ ─ ─ ─ │ │ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ╮ │ │ 6. Relay Result ◁───────────┘ └──────▷ 4b. ACK Communication □─┘ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ╯
18. Our context - Devengo 2.0 • Still an early-stage startup, just 7 developers (including me) running everything. • But already moving a volume of hundreds of thousands of transfers per month for a value of dozens of million euros (even from old competitors). • We'll (finally!) get a PSP license by the Central Bank of Spain this summer.
19. Given that context, how do we choose the right trade-off?
20. Devengo Engineering Principles We make trade-offs by applying our principles to our current context. • Finesse • Managers Of One • Zen Pragmatism • Taking Care Of Business • Do The Right Thing • Trust, But Verify
21. Devengo Engineering Principles • We materialized them in supporting practices. • We paranoidly monitor associated red flags. • We provide self-evaluation topics.
22. Finesse Trading speed for refinement of execution.
23. Finesse / Supporting practices • Continuous refinement of the Developer Experience. • Failing gracefully on unexpected scenarios. • Continuous improvement of features. • Continuous learning. • Expansive testing.
24. Finesse / Red flags • Quality defects reaching production. • Small known bugs are present for too long. • Unexpected collaterals
25. Finesse / Self-evaluation • Does this piece of work represent my best shot at the problem? • Have I thought deeply about the consequences of this implementation? • How can I make this more robust without making it more complex?
26. Zen Pragmatism Trading idealistic approaches for hardened working solutions.
27. Zen Pragmatism / Supporting practices • Iterative implementation • Scope reduction • Consensus built on definition stages • Proof of Concepts • Minimum Valuable Product • ADRs (Architectural Decision Records)
28. Zen Pragmatism / Red flags • Projects agonizing for months • Under-used, overly complex implementations • Premature optimization (of the business problem not performance) • Bike Shedding & Yak-shaving
29. Zen Pragmatism / Self-evaluation • Is this the most cost-effective way to solve our problems? • Am I building the solution of the core part of the problem first? • Is the chosen approach reversible and flexible?
30. Do The Right Thing Trading short-term optimization for long-term relationships.
31. Do The Right Thing / Supporting practices • Thinking like a customer • Cross-org transparency • Frequent interactions with support
32. Do The Right Thing / Red flags • “Not my problem” mentality • Customers feeling unheard. • Slow reaction times.
33. Do The Right Thing / Self-evaluation • Are we doing this just because it is easy? • Are we doing this just because “everybody does it”? • Is this the best thing for our customers?
34. Managers Of One Trading (micro)control for trust.
35. Managers Of One / Supporting practices • Proof of Concept • Designs open to feedback • Developer-led initiatives • Praising dev success stories and heroes.
36. Managers Of One / Red flags • Decision bottlenecks • Impostor syndrome • Analysis paralysis
37. Managers Of One / Self-evaluation • Do I have all the information to act on this? • If not, do I know how to get it? • Do I feel confident to commit to this solution?
38. Taking Care Of Business Trading individual goals for group success.
39. Taking Care Of Business / Supporting practices • Shared, cross-org language • Direct, frequent communication with OPS and customers • Everybody rotates to on-call • Prioritize boring technology • Provide tooling to other orgs inside Devengo • Buy vs Build evaluations
40. Taking Care Of Business / Red flags • “Not invented here” syndrome • Recurring bugs • Feeling stuck in a specific approach
41. Taking Care Of Business / Self-evaluation • Is there a no-coding solution to this problem? • Do I understand the point of view of the other parties? • Can I serve the mission in other ways?
42. Trust, But Verify Trading simplicity for reliability.
43. Trust, But Verify / Supporting practices • Exhaustive monitoring • Automatic & manual checks • Feature flags • Soft-launching features/products • Expansive testing • Incident management
44. Trust, But Verify / Red flags • “No deploys on Friday” attitude. • Manual processes. • Siloed knowledge.
45. Trust, But Verify / Self-evaluation • Do I feel confident about this proposal? • Do I understand the ramifications of my work? • If this implementation fails, what is the worst-case scenario?
46. Since death alone is certain, and the time of death uncertain, what should I do?
47. 𝕏 @_aitor / aitor.is 𝕏 @devengoapi / devengo.com Thank you! Share your feedback, love & insults! l.aitor.is/scpna