PFI Framework: Who Profits from the Innovation?
When you need this method
You have built something of your own, and other parties still earn from it: the systems integrator who delivers your offering, the data holder without whom it shows no benefit, the model provider it runs on. The usual tests only ask whether your capability is proprietary. They do not tell you which part of the chain you have to own for the profit to reach you, and when contracting is enough.
Approach
- 1Assess appropriability: is the knowledge written down or does it sit in people and routines, does a legal right actually bite, and how long would a copyist realistically need.
- 2Determine the market phase: are several designs still competing, or has a standard taken hold. After a standard emerges, scale, cost and access decide, no longer the better idea.
- 3List every asset additionally needed for a customer to actually experience the benefit: distribution, manufacturing, certification, data, the customer relationship, service, adjacent technology.
- 4Classify each of those assets: generally available, specialized (only you depend on it), or mutually specialized (both sides have to invest).
- 5For each specialized asset, establish who controls it, whether real alternatives exist, and how fast an imitator could secure the same access.
- 6Separate the critical assets from the dispensable ones and decide only for the critical: weak protection plus mutual specialization plus available capital argues for ownership; otherwise contract or license, but with bounded dependency and a negotiated exit.
- 7Cross-check for AI initiatives: does the bottleneck sit with you, with the model provider, with the data holder, or with the partner through whom your offering reaches the customer.
Typical application
A typical B2B SaaS in the HR field builds a review assistant for employment contracts on top of a bought-in language model. The framework shows: protection is weak, because the instructions given to the model can be reproduced within weeks, and no standard has settled yet. Among the complementary assets, compute and interface are generally available, while the vetted clause library and access through payroll systems are specialized, and the payroll interface is mutually specialized, since both sides would have to invest. The company drops the idea of its own model, brings maintenance of the clause library in house, and replaces the open partnership with the payroll vendor by a fixed-term contract covering data release and an exit right.
Limits and counter-indications
The framework sorts, it does not calculate. It will not tell you what a complementary asset is worth, how long your lead will hold, or what acquiring it may cost. Classifying an asset as specialized is contestable in advance and always plausible in hindsight, which makes the framework prone to justifying an acquisition decision that had already been taken. The empirical support comes from manufacturing and consumer goods plus one historical industry study, not from subscription software. Where technology cycles are short, owning a specialized asset can itself become a burden once the bottleneck moves. And the framework is silent on demand, price and channel: it settles who earns, not whether anyone buys.
How to measure impact
Measure the share of revenue that would not exist without an asset held by someone else, and estimate, for each such asset, the time it would take to replace it with an equivalent. Review both annually, since both figures shift with the market.
Related methods
Sources
- 1.Teece, David J.: Profiting from technological innovation: Implications for integration, collaboration, licensing and public policy, Research Policy 15(6), 285-305, 1986 (opens in a new tab) · Research Policy, Elsevier / North-Holland (Volltext-Faksimile über edegan.com) · 1986 · academic and scholarly literature · describes the methodCarries the procedure with its three building blocks: appropriability regime, dominant design, and complementary assets, which the paper divides into generic, specialized (unilateral dependence) and cospecialized (bilateral dependence). Section 5 supplies the decision rule: contracting beats integration when the appropriability regime is tight and complementary assets are in competitive supply; under weak appropriability with specialized assets, controlling them becomes the key success factor, and the paper grades the choice by time required and capital required, naming a minority position as a compromise for cash-constrained innovators. Limit: a conceptual framework with case illustrations, not a statistical test, and it says nothing about subscription software.
- 2.Teece (1986), Profiting From Technological Innovation: Zusammenfassung mit ausformulierter Entscheidungsregel (opens in a new tab) · edegan.com, Wissensbasis zur Innovations- und Unternehmerökonomie (Ed Egan) · undatiert · practitioner source · supports the underlying mechanismCondenses the three building blocks into a compact decision path: if no complementary assets are required, commercialize immediately; if they are required and protection is weak, buy the critical specialized assets provided capital and lead time allow, otherwise contract for access. It also records that integration bundles incentives and control, while contracting lowers risk and capital needs but creates hold-up exposure as soon as a partner has to invest specifically. Limit: a teaching summary with no evidence of its own, it does not replace the paper and adds nothing to it.
- 3.Tripsas, Mary: Unraveling the Process of Creative Destruction: Complementary Assets and Incumbent Survival in the Typesetter Industry, Strategic Management Journal 18 (Summer Special Issue), 119-142, 1997 (opens in a new tab) · Strategic Management Journal, John Wiley & Sons (Volltext über edegan.com) · 1997 · academic and scholarly literature · evidence of effectivenessTests the core of the framework against the history of the typesetter industry from 1886 to 1990, using firm- and product-level data for every company in the industry. It finds that incumbent products were technologically inferior in all three competence-destroying shifts, yet incumbents were displaced in only one of the three: where specialized complementary assets retained their value across the shift, incumbents held their market position despite weaker technology. Looking at investment or technical capability alone would, the paper argues, have been misleading. Limit: a single industry over a long period, observational rather than experimental, so it is no evidence of transferability to software.
- 4.Teece, David J.: Profiting from Innovation in the Digital Economy: Standards, Complementary Assets, and Business Models in the Wireless World, Research Policy 47(8), 1367-1387, 2018 (Autorenfassung) (opens in a new tab) · Research Policy, Elsevier (Autorenfassung über eScholarship, University of California) · 2018 · academic and scholarly literature · provides the contextThe original author extends his 1986 framework into the digital economy and names two shifts. First, with general purpose and enabling technologies the bottleneck can be someone else's technology rather than a conventional asset, so the innovator must acquire, license, or ally with its owner. Second, platforms and ecosystems were absent from the 1986 paper, even though profit distribution along the value chain cannot be understood without them. Limit: the evidence comes from wireless and standard essential patents, not from AI applications; carrying it over to today's model providers is a Convios reading, not a claim of the source.
Origin: David J. Teece (UC Berkeley, Haas School of Business)