Two founder exits in healthcare technology. Multiple firms started, scaled, and either sold or carried forward. The research below formalizes what forty years of building taught, tested at scale on a longitudinal venture database of more than four million firms. Dissertation, peer-reviewed conference proceedings, and the theoretical contribution behind both are first. The second research stream, on high-stakes decision architecture, follows.
Two questions drive this research stream. What predicts whether a startup moves from idea to funded venture? What makes a funded venture produce a transformative outcome rather than an average one? The peer-reviewed conference proceedings below answer the first. The dissertation answers the second. Both draw on the longitudinal venture database that anchors the work. I do not study entrepreneurship through case studies. I study it in large numbers.
Using GenAI-Augmented Content Analysis of S-1 and F-1 Filings. Daniels College of Business, University of Denver. Dissertation in progress, expected 2027.
The dissertation extends Schumpeter's creative destruction by specifying what must simultaneously be true about the ventures that achieve transformational outcomes. The dominant literature in entrepreneurship and strategy, drawing on Christensen's disruptive innovation, Barney's resource-based view, Kim and Mauborgne's blue ocean strategy, Hambrick and Mason's upper echelons theory, Sarasvathy's effectuation, and others, treats innovation success as decomposable into separable, additive factors. The implicit logic is compensatory. Strength in one dimension offsets weakness in another.
Petersen's Theory of Innovation Alignment rejects that logic. The theory specifies a system of constructs, each representing a necessary condition for transformational success, and predicts that a critical weakness in any single dimension constrains the probability of transformational outcomes regardless of strength elsewhere. Conjunctive, not compensatory. This is not a claim about marginal improvement. It is a claim about the architecture of extraordinary results.
The empirical work operationalizes the theory through AI-augmented content analysis of S-1 and F-1 IPO filings, with multi-model validation, applied to a longitudinal sample of post-IPO technology companies. Cox proportional hazards survival analysis examines the speed of transformation across post-IPO valuation milestones. The central question is which combinations of innovation alignment constructs predict transformational outcomes, and whether a conjunctive model outperforms a compensatory one.
One empirical study on the timing and selective criticality of market-facing functions in early-stage venture success. Accepted for presentation and proceedings publication at two peer-reviewed academic conferences in 2026.
Also presented as: Signaling Market Readiness: Early-Stage Startup Behavior and Venture Funding Progression
Venues
Abstract
Early-stage ventures begin with founder-led commercial activities but must eventually establish specialized departments to achieve scalable growth. This study examines which market-facing functions (marketing, sales, growth, product, business development, and corporate development) most influence early-stage success and when they should be established. Using time-series Cox proportional hazards models tracking 265,509 global ventures founded between 2010 and 2025 across 700,487 organization-period observations, the analysis finds that sales, product, and corporate development functions accelerate success by 27 percent, 26 percent, and 16 percent respectively, while marketing and growth functions show non-significant or negative effects. Founder prior entrepreneurial success moderates these relationships, with experienced founders' networks and market knowledge partially substituting for formal structures. The work reframes organizational design as an intertemporal marketing capability: when startups institutionalize market-facing functions proves as critical as whether they do so.
Key Finding
Sales (+27%), product (+26%), and corporate development (+16%) accelerate startup success. Marketing and growth, in early stages, do not. Founder prior success partially substitutes for formal structure. Timing matters as much as structure.
The research stream is anchored in a longitudinal venture database of more than four million firms. The database integrates founder histories, funding round timelines, organizational composition, IPO and exit outcomes, and post-IPO valuation trajectories. It supports the kind of large-sample empirical work that case studies and small-n research cannot do. It is the methodological asset that makes the conjunctive predictions of Petersen's Theory of Innovation Alignment testable at scale.
Petersen's Theory of Innovation Alignment is the theoretical contribution the dissertation operationalizes and the thread running through the broader research program. It rejects compensatory models of organizational success in favor of a conjunctive architecture in which transformative outcomes require the alignment of multiple necessary conditions. The same logic frames the High-Stakes Decision Architecture stream below: when the weakest link constrains the outcome, averaging across signals produces structurally wrong predictions. The formal manuscript establishing the theory is in preparation as part of the dissertation, with extensions in working papers and applied work.
The second empirical stream is about outcomes that are decided by a single critical weakness. Some results do not come from being good on balance. They come from satisfying every one of a small set of necessary conditions, where missing even one is fatal no matter how strong the rest are. A system can do almost everything right and still fail on the one thing it got wrong.
The three papers below show this pattern in production AI. Across dozens of documented deployments, the failures were rarely caused by weak technology overall. They were caused by one missing necessary condition: sound governance in the first paper, unbiased data and honest evaluation in the second, and cost discipline with a stopping rule in the third. In each case the organization was strong in most respects and still failed, because the outcome was set by its single weakest necessary part. The papers are listed in order of relevance to a first-time reader.
White paper
Why high-stakes AI decisions require a different architecture, and why most platforms cannot express it.
An executive guide to why standard AI platforms fail in high-stakes contexts, and what decision architecture is required instead.
Read the summaryWhite paper
Twelve case studies, two analyst forecasts, and one architectural diagnosis.
Forensic analysis of twelve documented AI implementation failures from 2016 through 2025, mapped to the architectural stages where each one broke. The companion to The High Stakes Decision: where that paper makes the architectural argument, this paper provides the evidence.
Read the summaryWhite paper
A decade of discrimination in production AI systems. Eight cases, five mechanisms, the Colorado standard, and a framework for executives.
Synthesizes eight documented cases of algorithmic discrimination from 2016 through 2025, identifies the five mechanisms by which bias enters production AI systems, and maps the findings onto Colorado's AI Act (SB24-205), which makes bias prevention the statutory standard of care.
Read the summaryWhite paper
A decade of cost and time-to-value failures in production AI. Five cases, one economic law, and a framework for executives.
Five documented cost and time-to-value failures from 2015 through 2026, read through Herbert Simon's bounded rationality and mapped onto the architectural stages where each one broke. The third paper in the When AI Fails series shows that AI cost failure is not a technology problem but a failure to satisfice, and gives executives a single-page gate to bound the spend before it bounds them.
Read the summaryWorking paper, in preparation for submission
Four architectures of causation in business research, and why the methods you were trained in may be answering a different question than the one you asked.
A methodological argument for researchers on when standard analytical methods answer the wrong question. Foundation for Innovation Alignment Theory.
Request the paperStrategic Briefing
How three compounding cost curves have rewritten who can do serious research, serious analysis, and serious consulting work, and what that means for competitive position.
When the economics of discovery shift this far this fast, the question for every CEO is which capabilities the organization no longer has to buy and which competitors can now reach the same answers without the same overhead.
Read the summaryWhite paper
A research-based analysis of generative AI competency requirements across disciplines in higher education.
A framework for what AI competencies graduates need across disciplines, with implications for curriculum design and institutional strategy.
Read the summaryWhite paper, in development
A reference for higher education curriculum design across management, business, and technical AI roles.
A reference document mapping emerging AI roles to curriculum requirements for higher education planning. Companion to the Colorado Graduates paper.
Request the paperEmpirical work applying alignment-based and weakest-link analytical architectures to reveal the structures that drive outcomes in high-stakes and transformative contexts. Method consultation, coauthorship, or full research design.
Active research collaboration is one of the engagements I value most, because it is where the academic side of my work compounds with the empirical work other scholars are already doing. The research I am most interested in collaborating on involves the application of non-averaging analytical architectures (weakest-link, conjunctive, configurational, alignment-based, top-versus-bottom contrast, hazard-based, event-based time series) to research questions where averaging methods have been the field's default.
Specific collaboration types include: methodology consultation on existing data and existing studies, full coauthorship on empirical work that benefits from a non-averaging analytical lens, research design partnership on new studies, and senior advisory roles on doctoral committees in business, healthcare, or related domains.
I am especially interested in collaborations that extend or test Innovation Alignment Theory, the conjunctive theoretical contribution currently in development as part of my doctoral research.
Doctoral candidates whose dissertations would benefit from a non-averaging analytical perspective. Faculty members working on entrepreneurial finance, organizational performance, healthcare outcomes, or innovation outcomes where the averaging assumption is doing more work than the methodology acknowledges. Methodologists working on configurational analysis, fsQCA, survival analysis, or similar approaches.
Research collaborations typically begin with a substantive conversation about the research question, the data, and the architectural fit. Some collaborations are short (methodology consultation on a paper in progress). Some are medium (coauthorship on a single study). Some are long (a research program that produces multiple papers over years). I am open to all three shapes, with preference for collaborations where the methodological match is strong and the research question is genuinely interesting.