ATHAN DIAL.PHD
← All Case Studies

Build vs Buy: Strategic Analysis for Analog Generation

product strategy · organization · high

Quantitative analysis guiding $250k+ partnership decision — ROI modeling revealed internal model needed 50× improvement, recommended hybrid approach balancing near-term progress with long-term IP

Context

Late 2025 brought Montai to a strategic crossroads. Our core IP hinged on generating novel analog compounds (“Anthrologs”) through proprietary AI models — but the internal generative model produced ~360M virtual compounds with uncertain synthetic feasibility. Meanwhile, XtalPi offered curated, higher-confidence AI-suggested compounds from external sources.

Facts:

  • Late 2025: Montai’s core IP = generating novel analog compounds (“Anthrologs”)
  • Problem: Internal generative model produced ~360M virtual compounds, but many not synthesizable/uncertain value
  • Alternative: XtalPi (external partner) offered curated AI-suggested compounds (more drug-like)
  • Stakes: Resource allocation - invest in internal model improvement OR buy external suggestions?

The CSO (Margo) needed an evidence-based recommendation by December 1, 2025. This wasn’t a philosophical debate about build vs buy — it was a portfolio allocation decision with measurable ROI implications. I had three weeks to model the tradeoffs quantitatively and make a clear recommendation.

Ownership

I owned:

  • Comparative analysis design (internal vs external compound quality)
  • ROI modeling (compounds accessible per $ investment)
  • Visualization strategy (UMAP plots, complexity charts for exec communication)
  • Presentation to CSO (Dec 1 deadline)

I influenced:

  • Strategic direction (hybrid approach recommendation)
  • ACN model improvement priorities (with Duminda, ML scientist)
  • Partnership terms with XtalPi (data informed negotiations)

Decision Frame

Problem statement:

Allocate resources between internal generative model improvement and external compound partnerships to maximize discovery speed while managing IP and cost tradeoffs, constrained by:

  • Internal generative model accuracy too low (needed ~50× improvement)
  • XtalPi partnership cost vs potential value unclear
  • Time pressure (programs need compounds NOW, not in 1 year)

Options considered:

Option A: Rely internally (double down on ACN model)

  • Pros: IP stays in-house, potentially massive unique space
  • Cons: Model needs significant improvement, compounds have uncertain synthetic feasibility
  • Risk: Wasted time on low-quality suggestions, delays programs

Option B: Outsource analog suggestions (XtalPi)

  • Pros: Immediate high-quality suggestions, less internal R&D
  • Cons: Cost, reliance on external, less proprietary
  • Risk: Dependence on partner, potential quality gaps

Option C: Hybrid (external for near-term + internal for long-term)

  • Pros: Don’t miss opportunities today while investing in tomorrow
  • Cons: More complex, requires patience for internal improvements
  • Risk: Internal improvements never materialize (sunk cost)

Decision: Recommended Option C because:

[FACTS from archaeology p.14-15:]

Quantitative analysis:

  1. Head-to-head comparison (STAT6, OX40L, TL1A):

    • Filtered both Montai Anthrologs + XtalPi space to top quality
    • Compared predicted activity and novelty via UMAP diversity plots
    • XtalPi covered chemical areas Montai didn’t (complementary)

  2. ROI model (“napkin math”):

    • $250k building block investment → 0.9-12.4M Anthrologs accessible
    • Scales with higher budgets, but diminishing returns without model improvement
    • Calculated ACN precision gain needed: ~50× improvement for full generative utility

  3. Strategic implication:

    • External (XtalPi) fills gap NOW while ACN improves
    • Set goal: If cACN v2 achieves precision by mid-2026, reduce external dependency
    • Hybrid de-risks both options (not all-in on unproven internal model)

Constraints:

  • Dec 1, 2025 deadline to CSO (time-boxed analysis)
  • 360M Montai compounds vs unknown XtalPi set size (asymmetric comparison)
  • Internal model improvement timeline uncertain (Duminda’s capacity)

Outcome

Primary outcome:

Guided strategic decision that balanced near-term progress with long-term IP:

  • Immediate: Montai allocated budget to XtalPi-sourced compounds for 2-3 programs
  • Long-term: Tasked data/ML team to focus on cACN v2/v3 improvements (precision goal)
  • Validation: Some XtalPi compounds became hits in early 2026 (justified spend)

The hybrid approach avoided overcommitting in either direction. Going all-in on internal models would have delayed programs by 6-12 months while we chased a 50× accuracy improvement. Going fully external would have ceded our core IP advantage. The quantitative analysis made it clear that both paths had merit — and that combining them de-risked the strategy while keeping options open.

Metrics:

  • Analysis timeframe: ~3 weeks (Nov-Dec 2025)
  • Executive decision: Made by Dec 1, 2025 (met deadline)
  • Outcome validation: XtalPi hits in Q1 2026 + cACN v2 improved 2× by mid-2026 (on track to 50×)

Strategic artifacts:

  • ROI model spreadsheet (investment scenarios)
  • UMAP diversity plots (XtalPi vs Montai chemical space)
  • Complexity charts (synthetic accessibility comparison)
  • Presentation deck to CSO (evidence-based recommendation)

Second-order effects:

  • Template for future build vs buy decisions (quantitative framework reused)
  • Strengthened partnership with XtalPi (data-driven collaboration)
  • Internal team focused on high-leverage improvements (not boiling ocean)

Limitations acknowledged:

  • Analysis timeboxed (deeper investigation possible but not needed for decision)
  • Model improvement uncertainty (50× gain ambitious, may take longer)
  • Hybrid complexity (managing two paths simultaneously)

Reflection

What I’d do differently:

Three strategic analysis lessons emerged in retrospect:

  • Start internal model improvements earlier (not wait for crisis/comparison)
  • Engage chemists more in generative model design (synthetic feasibility blindspot)
  • Pilot XtalPi with one program before org-wide (reduce risk)

The first was a failure of proactive planning — we knew the ACN model had accuracy issues but didn’t prioritize improvements until external pressure forced the comparison. The second repeated a pattern from the nomination scaling work: treating chemist expertise as validation rather than design input. The third was standard risk mitigation we should have applied but didn’t.

What this taught me about decision-making:

This analysis validated three principles about strategic framing:

  • Quantitative framing transforms opinions into evidence — the ROI model gave the CSO a concrete basis for decision rather than competing gut feelings about build vs buy
  • Build vs buy is rarely binary — hybrid approaches often dominate pure strategies, especially when timelines and uncertainty favor hedging
  • Executive decisions need clear options plus a recommendation — presenting “here’s the data, you decide” abdicates leadership; executives want your synthesis and recommendation backed by evidence

How this informs future decisions:

These meta-takeaways now guide my approach to strategic analysis:

  • Always model tradeoffs quantitatively when stakes are high — even “napkin math” beats handwaving about strategic direction
  • Build vs buy is a portfolio decision, not an all-in bet — maintaining optionality through hybrid approaches preserves strategic flexibility
  • Strategic patience requires near-term wins to buy time for long bets — the hybrid worked because XtalPi compounds delivered wins while we improved internal models

Factual Evidence Citations:

  • Project Inventory p.9 (XtalPi collaboration entry)
  • Decision Systems p.14-15 (detailed analysis narrative)
  • Slack DM (William→Athan analysis goal)
  • Jake Ombach Slack (quantitative scenario results)