How I Manage Projects End To End (1 Hour Masterclass)

End-to-end project management “formula” (as presented)

Core premise: you start with low-resolution understanding

• In early stages, you don’t actually know what you’re building or what the market will pay for.
• Your understanding begins low-resolution and improves only through action.

Replace Waterfall and “known backlog” Agile

• Waterfall assumes the map/spec is already known.
• Agile assumes the backlog is roughly known per sprint.
• In fast-changing markets (e.g., AI and distribution), both are described as increasingly unrealistic.

Playbook: resolution model

• Start with a low-resolution plan and a minimal team.
• Use experiments (“action produces information”) to increase resolution over time.
• Build the org chart shape gradually; then scale.

---

Frameworks / playbooks explicitly referenced or implied

Low-resolution → high-resolution progression

• “Resolution increases over time as you do more.”

Team + org-graph “shape”

• Don’t scale by adding roles immediately.
• Instead, you redraw the org graph as learning occurs.

Department-first decomposition

• Break the project into departments (problems) first.
• Then derive roles (learned solutions).

Sure-shot vs R&D labeling

• Sure-shot: you know the technique; can scale by hiring more.
• R&D: you don’t know the method; requires founder attention + experiments.

Vertical slicing + Alpha/Beta/Gold

• Build small playable slices repeatedly to learn and raise resolution.
• Alpha: experimentation/learning
• Beta: near-final shape + real feedback
• Gold: bug-free playable target

Alpha → Alpha iterations

• Multiple alphas as the same slice, with increasing team depth/resolution.

Capacity math + buffer

• Use observed throughput from slices to estimate what’s needed for a full product.
• Include buffer.

Technology investment decision rule

• Invest in tech after you’ve done the “math”:
  • cost per output
  • time saved
  • break-even hires

Feedback gating + conversion targets

• Continue/stop based on willingness-to-pay or proxies (e.g., wishlists).

---

Concrete process (step-by-step)

1. Timebox learning goals

   • Example target: for the Unleash the Avatar (UTA) core team, determine whether to continue/stop within ~3 months after forming the core team.

2. Define “three outcomes” from experiments

   • Continue (great feedback)
   • Continue but need more data (unclear)
   • Stop (waste of time; no interest/paying demand)

3. Build a low-resolution “minimum viable team”

   • One-person-per-function across key departments (example given):
     • code, animation, cinematics, character art, levels, game design, VFX+sound

4. Label work as Sure-shot vs R&D

   • Ensure many departments are sure-shot (scale-ready).
   • Keep R&D limited.

5. Iterate via vertical slices across stages

   • Alpha: experimenting/learning (can have multiple alpha iterations)
   • Beta: shape nearly found; collect real feedback
   • Gold: bug-free playable target

6. Redraw org chart shape before scaling

   • Don’t scale a “graph” early.
   • First find the shape via alphas/outsourcing/contractors before committing.

7. Use outsourcing/contractors temporarily to accelerate shape

   • Outsource specific R&D-adjacent or short-duration tasks to avoid hiring too early.

---

Key metrics / KPIs / targets mentioned (business execution)

Product-market validation / demand proxies

• Goal for liking: at least 50% of people like it (explicit acceptance threshold).
• Targeted beta conversion: 5–6% conversion after beta play (proxy question: “Would you buy the game?”).
• Steam proxy metric: wishlists
  • Rule of thumb: 20–30% of wishlisters eventually buy.
• TAM framing:
  • Track Total Addressable Market progress between funding stages (guesswork reduction via trials).

Operational / planning metrics

• Buffer: ~30% buffer added to capacity estimates.
• Photogrammetry R&D example throughput
  • One scan effort: ~10 days with a dedicated team and rental cameras.
  • Earlier stages: phone/mobile scans → local street scans → film-town scans (increasing realism).

Tech investment math (example numbers)

• Animation tech break-even logic
  • Hypothetical keyframing case: ~40 animators for a full game under a “12 months” schedule.
  • Hypothetical cost: 10 lakhs per animator per year → ~4 crores/year for full keyframing.
• Motion capture tech alternative
  • ~7 animators + motion capture equipment cost ~50 lakhs
  • Investment breaks even if savings cover ~5 hires (their break-even assumption).

---

Concrete examples / case studies (UTA)

1) Vertical slice learning via trailers

• UTA released three trailers.
• Learning progression (implied):
  • Trailer 1: “cool start” but not “a great game”
  • Trailer 2: improved; major feedback still on combat
  • Trailer 3: larger playtest + more targeted combat feedback
    • Team then created roles for “meta game” balancing (e.g., skill/stat tuning)

2) Photogrammetry R&D as a department case study (Levels)

• Levels department included:
  • level design, level art, lighting art, photogrammetry
• Photogrammetry R&D escalation path:
  1. Mobile-phone scans
     • Consumer apps → nice but not game-final-quality assets.
  2. DSLR scans
     • Example: local Mumbai area/slum
     • Got ~60–70% of needed assets, but failed as “slum assets” for final game direction.
  3. Chanderei (film town) scans
     • Hired a line producer
     • Sent ~9 people (about a 10-person effort)
     • Took about ~10 days
     • Managed logistics: lodging, hotel costs, camera rentals
     • Shooting constraint: shoot between ~6–8 AM and 4–6 PM (light quality) + manual “delight it later” cleanup
     • Risk event: a drone got damaged mid-process; required shipping a replacement drone
• Decision pattern:
  • Keep photogrammetry as R&D until proven.
  • Label other level work as sure-shot.

3) AI use boundaries (workflow vs audience-facing)

• Their stance:
  • Do not generate audience-facing creative assets (dialogue/images/skill tree) via AI—these must be human-made.
  • Use AI for Lightning Way internal workflow tooling/communication/shared-state.
  • Not for “playing as an AI-generated product.”

4) Internal alignment tooling (custom CRM-style tracker)

• They built a visual skill tree / viewer tool with:
  • skill details and skill extensions
  • status tracking for workflow steps (e.g., concept clarity, motion capture, animation cleanup, code, VFX, sound, “full pass”)
• Purpose:
  • prevent misalignment as org density increases.

---

Actionable recommendations (extracted behaviors)

• Use a “3-month decision clock” after building a core team to decide continue/stop/pivot via experiments.
• Build low-res slices, not full systems—let action generate information.
• Separate departments (problems) from roles (solutions) to avoid copying other companies’ org structures.
• Cap innovation scope:
  • only innovate in 1–2 segments
  • everything else should be sure-shot or backlog-managed.
• Run R&D as fast trial loops to compress months of uncertainty into days via better iteration tools.
• Use capacity math + buffer (~30%) to forecast hires and schedule:
  • measure throughput from a playable slice
  • scale estimate + buffer.
• Invest in technology only after cost math + break-even reasoning, not before.
• Don’t scale the org graph prematurely:
  • scaling too early after raising money can force painful re-orgs and repeated “re-redraws” (including layoffs).
• After “shape found” (post Series B):
  • either start a new product or improve conversion,
  • but avoid constantly adding features/roles that create “junk.”

---

Funding / scaling guidance (high level, execution-focused)

• Bootstrap / publishers may apply for strict distribution requirements (e.g., publishing in China).
• Funding stages map to certainty about “shape”:
  • Seed: freedom to pivot 3–4 times to find the right resolution/shape
  • Series A: shape is known; scale “some of it” + invest in tech where it matters
  • Series B: shape is known; push TAM edges strategically
• Warning:
  • If you commit to the final org shape too early, then redraw after raising money, you’ll face emotional/contract commitments and likely layoffs.

---

Presenters / sources

• Presenter/source: the video speaker (no name provided in subtitles).
• Referenced external sources/entities:
  • Ubisoft
  • indie studios
  • “Paul Graham”
  • Path of Exile 2 (mentor reference)
  • Rockstar (talent reference)
  • Ghost of Tsushima (talent reference)
  • Codeex (talent sourcing example)
  • Steam (wishlists metric)
  • Billybilly World (playtest mention)
  • RealityScan (photogrammetry software)
  • Co-founders/partners not specified
  • Rocket Internet (mentioned as a playbook cloner example)