63 Innovation, New Products, and Diffusion Seminar

Innovation is the engine of growth, and new products are where marketing, engineering, and finance collide. This chapter is a graduate seminar in how new products are conceived, developed, launched, adopted, and—if they survive—diffuse through a market until they remake or destroy the firms that compete to supply them. It organizes the field not as a checklist of stage-gate steps but as a tour of the questions a doctoral student must be able to pose and answer: how does a new product spread through a population, and can its sales be predicted before its peak? how is the development process itself managed, from the voice of the customer to a finished design? when does pioneering pay, and when does it merely advertise the opportunity to a fast follower? does radical innovation reward the firms that bet on it—and do capital markets price that bet correctly?

The chapter matters because innovation is simultaneously the most celebrated and the least controllable strategic lever. A single diffusion curve can be fit with two behavioral parameters, yet those parameters are notoriously unstable before the sales peak; a development process can be mapped into clean decision modules, yet the hardest choices—how radical to be, when to cannibalize, when to enter—resist optimization. The seminar’s central tension runs through every week: there are three distinct intellectual projects here, and they pull in different directions. Predicting and modeling diffusion (the Bass tradition and its descendants) treats the new product as an aggregate sales trajectory to be forecast; managing the development process (the design and voice-of-the-customer tradition) treats it as a sequence of engineering and marketing decisions to be optimized; and valuing innovation (the marketing–finance tradition) treats it as a risky asset to be priced by capital markets. A student who finishes the course should be able to estimate a diffusion model and state what breaks its identification, design a concept test, and read an event study of an innovation announcement with equal fluency.

This seminar is the doctoral reading-map that organizes the substantive material of the innovation chapter (Chapter 25) into a full field course, and it shares a border with the market-entry chapter (Chapter 26): timing of entry, pioneer advantage, and order effects are studied here as facets of the innovation problem rather than as a separate competitive-strategy question. The course is also a strong fit for the book’s advisor corpus. Gerard J. Tellis has shaped the modern empirical study of new-product growth—international takeoff, the saddle, radical versus incremental innovation, the incumbent’s curse, technological evolution, and the stock-market returns to innovation—and his work anchors several weeks below. These are woven where genuinely canonical, not gratuitously: the seminar’s spine would look much the same in any top program, and it happens to run through a great deal of one advisor’s research.

63.1 Semester arc

The arc begins with diffusion as an aggregate phenomenon. The first four weeks build the modeling backbone—the Bass model and its generalizations, cross-country and international diffusion, and the empirical regularities of takeoff and the “saddle”—because a doctoral student must be able to write down, estimate, and critique a diffusion model before doing anything else with new products. This block establishes the field’s first dependent variable, the sales trajectory, and its recurring identification problem: the curve’s parameters are weakly identified from pre-peak data, so forecasts made when they are most useful are least reliable.

The middle of the semester turns from the market’s behavior to the firm’s behavior: the new-product development process itself. Weeks on the development- decisions framework, the voice of the customer and concept testing, and preference measurement and product-design optimization treat the new product as something to be engineered and tested, not merely forecast. Here the seminar connects to the preference-measurement and conjoint machinery used throughout the book, and it foregrounds a different methodological tradition—survey design, perceptual mapping, and choice modeling rather than time-series estimation. The block on really-new products and entry timing then bridges process and strategy: when the product is genuinely new, both customer learning and competitive order-of-entry effects change the rules.

The seminar closes with innovation as strategy and as a priced asset. Weeks on radical versus incremental innovation and the incumbent, technology generations and network effects, user and open innovation, innovation and firm value, technological evolution and disruption, and a diffusion frontier (social contagion and machine learning) move the course from forecasting and process toward the questions that animate current dissertations: does radical innovation pay, and for whom? do incumbents really suffer a curse? do markets price innovation efficiently? Throughout, the pedagogy is method-forward—each module names the identification challenge (parameter instability, selection into entry, endogenous innovation choice, event-study and factor-model confounds, contagion-versus-homophily) that stands between a pattern and a finding.

The reading map uses two tags: [F] = Foundational (canon an innovation scholar is expected to know cold) and [R] = Frontier/Recent (an active research front, refreshed as the literature moves). Each week pairs at least one foundational anchor with one frontier paper. DOIs are reproduced as verified against Crossref; works without a DOI-verified record (chiefly books and a pre-DOI economics classic) are named without a link and flagged.

63.2 Week 1 — Diffusion and the Bass model

Topic. How a new product spreads through a population of potential adopters; the canonical aggregate model of first-purchase diffusion.

Subtopics. Innovators vs. imitators; the adopter categories; the $S$-curve; coefficients of innovation and imitation; the qualitative theory of diffusion behind the algebra.

Methods. Aggregate (macro) diffusion modeling; nonlinear least squares and maximum likelihood on cumulative-adoption data; the sociological theory of adoption.

Key readings.

Bass (1969), “A New Product Growth for Model Consumer Durables,” Management Science. doi:10.1287/mnsc.15.5.215 — the founding two-parameter diffusion model; the anchor for the entire course. [F]
Rogers, Diffusion of Innovations (Free Press; 1st ed. 1962, 5th ed. 2003) — the qualitative theory of adopter categories and innovation attributes behind the algebra; a book, cited without DOI. [F]
Mahajan, Muller & Bass (1990), “New Product Diffusion Models in Marketing: A Review and Directions for Research,” Journal of Marketing. doi:10.1177/002224299005400101 — the field-defining review that frames the whole modeling tradition (read here as orientation, developed in Week 2). [F]

Debate. Is the $p$/$q$ split a behavioral truth (external vs. internal influence) or a convenient functional form? Do innovators and imitators name real people or just curvature?

63.3 Week 2 — Bass extensions and generalized diffusion

Topic. Relaxing the basic model: marketing-mix effects, repeat purchase, flexible curves, and the conditions under which extensions actually improve fit and forecasts.

Subtopics. The generalized Bass model with price and advertising; why decision variables often add little to fit; flexible and multi-stage diffusion; refinement vs. parsimony.

Methods. Estimation under added covariates; nested-model comparison; forecasting evaluation and the bias of pre-peak estimates.

Key readings.

Mahajan, Muller & Bass (1990), “New Product Diffusion Models in Marketing: A Review and Directions for Research,” Journal of Marketing. doi:10.1177/002224299005400101 — the canonical taxonomy of extensions and open problems. [F]
Bass, Krishnan & Jain (1994), “Why the Bass Model Fits Without Decision Variables,” Marketing Science. doi:10.1287/mksc.13.3.203 — the generalized Bass model, and the subtle argument that mix effects enter multiplicatively without disturbing the basic shape. [F]
Norton & Bass (1987), “A Diffusion Theory Model of Adoption and Substitution for Successive Generations of High-Technology Products,” Management Science. doi:10.1287/mnsc.33.9.1069 — extends diffusion to technology generations (revisited in Week 10). [F]

Debate. When does adding price/advertising to a diffusion model buy real forecasting power, and when is it overfitting? Is the generalized Bass model an explanation or a rationalization of the basic model’s robustness?

63.4 Week 3 — Cross-country and international diffusion

Topic. How diffusion parameters and takeoff timing vary across countries, and what economic, cultural, and institutional factors drive the variation.

Subtopics. Cross-national differences in $p$ and $q$; lead–lag effects between countries; developing-country diffusion; culture and country innovativeness as moderators.

Methods. Hierarchical/Bayesian estimation across countries and products; meta-analytic synthesis of parameters; hazard models of takeoff timing.

Key readings.

Tellis, Stremersch & Yin (2003), “The International Takeoff of New Products: The Role of Economics, Culture, and Country Innovativeness,” Marketing Science. doi:10.1287/mksc.22.2.188.16041 — what makes new products take off faster in some countries than others. [F]
Talukdar, Sudhir & Ainslie (2002), “Investigating New Product Diffusion Across Products and Countries,” Marketing Science. doi:10.1287/mksc.21.1.97.161 — a Bayesian decomposition of diffusion parameters across products and developed/ developing markets. [F]

Debate. Are cross-country differences in diffusion driven by economics (income, infrastructure) or culture (uncertainty avoidance, cosmopolitanism)? Do later-adopting countries diffuse faster because they “learn” from earlier ones?

63.5 Week 4 — Takeoff, slowdown, and saddles

Topic. The non-monotone features of real sales curves: the long flat introductory stage, the sudden takeoff, and the post-takeoff sales dip (“saddle”) that the smooth Bass curve cannot produce.

Subtopics. Modeling takeoff as a hazard; really-new vs. incrementally-new takeoff; the saddle and the chasm; informational cascades and heterogeneity as mechanisms.

Methods. Hazard/duration models of takeoff; threshold and two-segment diffusion models; identifying structural breaks in sales.

Key readings.

Golder & Tellis (1997), “Will It Ever Fly? Modeling the Takeoff of Really New Consumer Durables,” Marketing Science. doi:10.1287/mksc.16.3.256 — defines and predicts the takeoff point, the transition from introduction to rapid growth. [F]
Goldenberg, Libai & Muller (2002), “Riding the Saddle: How Cross-Market Communications Can Create a Major Slump in Sales,” Journal of Marketing. doi:10.1509/jmkg.66.2.1.18472 — the saddle as a consequence of weak communication between early and main markets. [F]
Chandrasekaran & Tellis (2011), “Getting a Grip on the Saddle: Chasms or Cycles?,” Journal of Marketing. doi:10.1509/jmkg.75.4.21 — large-sample evidence on how prevalent and how deep saddles are. [R]

Debate. Is the saddle a structural feature of two-segment markets or an artifact of macroeconomic cycles? Can takeoff be forecast in time to act on it?

63.6 Week 5 — The new-product development process

Topic. Turning the lens from the market to the firm: how new products are actually developed, and the decisions that structure the process.

Subtopics. The development-decisions framework (concept, design, testing, production ramp); stage-gate and its critics; platform and portfolio decisions; the marketing–engineering interface.

Methods. Decision-analytic and process frameworks; design-process modeling; case-based and survey evidence on development practice.

Key readings.

Krishnan & Ulrich (2001), “Product Development Decisions: A Review of the Literature,” Management Science. doi:10.1287/mnsc.47.1.1.10668 — organizes the sprawling NPD literature as a set of decisions across disciplines. [F]
Urban & Hauser, Design and Marketing of New Products (2nd ed., Prentice Hall,
1. — the canonical text mapping the development funnel from opportunity identification to launch; a book, cited without DOI. [F]

Debate. Is NPD best modeled as a sequence of decisions (Krishnan–Ulrich) or as an organizational/cultural capability? Does formal stage-gate help or ossify radical projects?

63.7 Week 6 — Voice of the customer and concept testing

Topic. Eliciting and structuring customer needs as the front end of development; turning fuzzy wants into engineering targets.

Subtopics. The voice of the customer (VOC); need hierarchies and relative importance; quality function deployment (QFD); concept generation and screening.

Methods. Qualitative-interview elicitation; affinity structuring; reliability of needs measurement; concept-test design.

Key readings.

Griffin & Hauser (1993), “The Voice of the Customer,” Marketing Science. doi:10.1287/mksc.12.1.1 — how many customers and analysts are needed to capture the structured set of needs; the measurement foundation of the front end. [F]
Urban & Hauser, Design and Marketing of New Products (2nd ed., 1993) — the QFD “house of quality” and concept-testing pipeline (revisited from Week 5 through the needs-measurement lens); a book, cited without DOI. [F]

Debate. Does VOC reliably surface latent needs, or only articulated ones? Are really-new products knowable from customer interviews at all (a hand-off to Week 8)?

63.8 Week 7 — Preference measurement and product-design optimization

Topic. Quantifying how customers value product attributes, and using those valuations to design and position the product.

Subtopics. Conjoint analysis (ratings-based and choice-based); attribute part-worths; product-line and design optimization; commercial practice and its evolution.

Methods. Conjoint/choice experiments; hierarchical Bayes estimation of part-worths; optimization over the attribute space; perceptual mapping.

Key readings.

Green & Srinivasan (1990), “Conjoint Analysis in Marketing: New Developments with Implications for Research and Practice,” Journal of Marketing. doi:10.1177/002224299005400402 — the authoritative synthesis of conjoint method and design. [F]
Wittink & Cattin (1989), “Commercial Use of Conjoint Analysis: An Update,” Journal of Marketing. doi:10.1177/002224298905300310 — documents how widely and how conjoint is used in practice. [F]

Debate. Do stated-preference part-worths predict real new-product choice? When does choice-based conjoint beat ratings-based, and at what cost in design efficiency?

63.9 Week 8 — Really-new products and entry timing

Topic. What changes when the product is genuinely new—and when does being first to a new market pay?

Subtopics. Measuring preferences for really-new products; pioneer advantage vs. pioneer myth; order-of-entry effects; survival risks of pioneers vs. followers.

Methods. Preference elicitation under unfamiliarity (information acceleration); survival/hazard analysis of entrants; survivorship-bias correction in pioneer samples.

Key readings.

Hoeffler (2003), “Measuring Preferences for Really New Products,” Journal of Marketing Research. doi:10.1509/jmkr.40.4.406.19394 — why conventional preference measurement fails for unfamiliar products, and what to do instead. [F]
Golder & Tellis (1993), “Pioneer Advantage: Marketing Logic or Marketing Legend?,” Journal of Marketing Research. doi:10.1177/002224379303000203 — shows that survivorship bias inflates the apparent advantage of market pioneers. [F]
Min, Kalwani & Robinson (2006), “Market Pioneer and Early Follower Survival Risks: A Contingency Analysis of Really New Versus Incrementally New Product-Markets,” Journal of Marketing. doi:10.1509/jmkg.70.1.015.qxd — pioneer survival depends on whether the market is really-new or incremental. [R]

Debate. Is pioneer advantage real once survivorship bias is corrected? Does the “really-new” distinction reverse the entry-timing prescription?

63.10 Week 9 — Radical vs. incremental innovation and the incumbent

Topic. The organizational determinants of radical innovation, and whether large incumbents are doomed to lose the radical-innovation race.

Subtopics. Willingness to cannibalize; organizational antecedents of radical product innovation; the incumbent’s curse (and its limits); size and incumbency effects.

Methods. Survey measurement of organizational constructs; historical content analysis of innovation databases; logit models of who innovates radically.

Key readings.

Chandy & Tellis (1998), “Organizing for Radical Product Innovation: The Overlooked Role of Willingness to Cannibalize,” Journal of Marketing Research. doi:10.1177/002224379803500406 — willingness to cannibalize existing assets as the key organizational driver. [F]
Chandy & Tellis (2000), “The Incumbent’s Curse? Incumbency, Size, and Radical Product Innovation,” Journal of Marketing. doi:10.1509/jmkg.64.3.1.18033 — historical evidence that incumbents and large firms innovate radically more often than the “curse” narrative claims. [F]

Debate. Is the incumbent’s curse a robust regularity or a sampling artifact of celebrated failures? Is willingness to cannibalize a cause of radical innovation or a rationalization after the fact?

63.11 Week 10 — Technology generations and network effects

Topic. Innovation as a sequence of substituting generations, and the demand-side forces—network externalities—that govern adoption of compatible technologies.

Subtopics. Successive-generation diffusion and substitution; installed base and adoption; network externalities and compatibility; standards and tipping.

Methods. Multi-generation diffusion models; structural models of adoption under network effects; analytical models of compatibility competition.

Key readings.

Norton & Bass (1987), “A Diffusion Theory Model of Adoption and Substitution for Successive Generations of High-Technology Products,” Management Science. doi:10.1287/mnsc.33.9.1069 — diffusion when a new generation cannibalizes and extends the old. [F]
Katz & Shapiro (1985), “Network Externalities, Competition, and Compatibility,” American Economic Review — the foundational analytical treatment of demand-side network effects and compatibility; a pre-DOI classic cited without a verified Crossref link (flagged). [F]

Debate. Do network effects “tip” markets to a single standard, or can incompatible technologies coexist? Does successive-generation substitution follow the same $p$/$q$ logic as first-purchase diffusion?

63.12 Week 11 — User innovation, open innovation, and crowdfunding

Topic. Innovation that originates outside the firm—among lead users and crowds—and the platforms that finance and surface it.

Subtopics. Lead users as a source of concepts; user vs. manufacturer innovation; open and distributed innovation; crowdfunding dynamics and signals of success.

Methods. Lead-user identification and field studies; analysis of platform/ crowdfunding data; hazard and regression models of funding success.

Key readings.

von Hippel (1986), “Lead Users: A Source of Novel Product Concepts,” Management Science. doi:10.1287/mnsc.32.7.791 — the founding statement that ahead-of-market users generate the most valuable concepts. [F]
Mollick (2014), “The Dynamics of Crowdfunding: An Exploratory Study,” Journal of Business Venturing. doi:10.1016/j.jbusvent.2013.06.005 — what predicts crowdfunding success, and how delivery actually plays out. [R]

Debate. Do lead users predict mass-market needs or only niche ones? Is crowdfunding a market test, a financing mechanism, or a marketing event—and does the distinction change what the data mean?

63.13 Week 12 — Innovation and firm value

Topic. Whether and how innovation translates into shareholder value and risk—the marketing–finance interface applied to new products.

Subtopics. Radical innovation and financial consequences; innovation as a driver of abnormal stock returns; value vs. risk effects of radical vs. incremental innovation; efficient vs. biased market reactions.

Methods. Content-analyzed innovation databases; event studies; calendar-time portfolios and long-run abnormal returns; risk decomposition.

Key readings.

Sorescu, Chandy & Prabhu (2003), “Sources and Financial Consequences of Radical Innovation: Insights from Pharmaceuticals,” Journal of Marketing. doi:10.1509/jmkg.67.4.82.18687 — which firms produce radical innovations and how markets value them. [F]
Sood & Tellis (2009), “Do Innovations Really Pay Off? Total Stock Market Returns to Innovation,” Marketing Science. doi:10.1287/mksc.1080.0407 — measuring the full stock-market return to innovation across the project life cycle, not just at announcement. [R]
Sorescu & Spanjol (2008), “Innovation’s Effect on Firm Value and Risk: Insights from Consumer Packaged Goods,” Journal of Marketing. doi:10.1509/jmkg.72.2.114 — distinguishes the value and risk effects of radical vs. incremental innovation. [R]

Debate. Does radical innovation pay on a risk-adjusted basis? Are the abnormal returns evidence of mispricing or compensation for risk—and does measuring returns across the whole life cycle change the verdict?

63.14 Week 13 — Technological evolution and disruption

Topic. How technologies evolve over time and across competing trajectories, and when new technologies disrupt established ones.

Subtopics. Step-function vs. smooth technological evolution; multiple competing technologies and crossings; disruptive vs. sustaining innovation; the demand-side account of disruption.

Methods. Long-run performance-trajectory data; technology S-curve and crossing analysis; demand-based modeling of segment overlap.

Key readings.

Sood & Tellis (2005), “Technological Evolution and Radical Innovation,” Journal of Marketing. doi:10.1509/jmkg.69.3.152.66361 — evidence that technologies evolve through irregular step functions with many crossings, not a single smooth S-curve. [F]
Christensen, The Innovator’s Dilemma (Harvard Business School Press, 1997) — the disruptive-innovation thesis that well-managed incumbents fail by listening to their best customers; a book, cited without DOI. [F]
Adner (2002), “When Are Technologies Disruptive? A Demand-Based View of the Emergence of Competition,” Strategic Management Journal. doi:10.1002/smj.246 — recasts disruption as a consequence of demand-side segment structure rather than supply-side technology. [R]

Debate. Is disruption a supply-side (technology) or demand-side (preference- overlap) phenomenon? Does the evidence on technological evolution support or undercut the disruption narrative?

63.15 Week 14 — Diffusion frontier: social contagion, machine learning, synthesis

Topic. The modern micro-foundations of diffusion—individual-level contagion and networks—and the methodological frontier; closing synthesis.

Subtopics. Opinion leadership and social contagion; contagion vs. homophily (the identification problem); network-based and machine-learning diffusion models; what the field knows and what remains open.

Methods. Network panel data; contagion identification (instruments, structural controls); ML/predictive diffusion modeling; conceptual synthesis.

Key readings.

Iyengar, Van den Bulte & Valente (2011), “Opinion Leadership and Social Contagion in New Product Diffusion,” Marketing Science. doi:10.1287/mksc.1100.0566 — sociometric and self-reported opinion leadership and how contagion operates through networks. [R]
Van den Bulte & Lilien (2001), “Medical Innovation Revisited: Social Contagion Versus Marketing Effort,” American Journal of Sociology. doi:10.1086/320819 — the landmark demonstration that apparent contagion can vanish once marketing effort is controlled; the identification cautionary tale. [F]

Debate. Is observed adoption clustering genuine contagion or homophily/common exposure? Do ML diffusion models forecast better while explaining less—and does the field want prediction or mechanism?

63.16 Foundational vs. frontier at a glance

Foundational core (every innovation student must know): Bass (1969); Rogers (Diffusion of Innovations); Katz & Shapiro (1985); von Hippel (1986); Norton & Bass (1987); Mahajan, Muller & Bass (1990); Green & Srinivasan (1990); Wittink & Cattin (1989); Griffin & Hauser (1993); Golder & Tellis (1993); Urban & Hauser (Design and Marketing of New Products); Bass, Krishnan & Jain (1994); Golder & Tellis (1997); Christensen (1997); Chandy & Tellis (1998, 2000); Krishnan & Ulrich (2001); Van den Bulte & Lilien (2001); Goldenberg, Libai & Muller (2002); Talukdar, Sudhir & Ainslie (2002); Tellis, Stremersch & Yin (2003); Hoeffler (2003); Sorescu, Chandy & Prabhu (2003).

Frontier / actively updated (refresh each edition): Sood & Tellis (2005, 2009); Min, Kalwani & Robinson (2006); Sorescu & Spanjol (2008); Adner (2002, read as a frontier reframing); Chandrasekaran & Tellis (2011); Iyengar, Van den Bulte & Valente (2011); Mollick (2014). The split is pedagogical, not chronological: Bass (1969) is foundational because the field still estimates its model weekly, while a 2002 demand-based theory of disruption is “frontier” because its agenda is still being executed. Each module pairs at least one foundational anchor with a live edge so students meet both the canon and its open questions.

63.17 How this chapter expands

The weekly map is a backbone, not a ceiling. It is designed to grow along several axes:

An estimation companion per week. Each module names an identification challenge—parameter instability in pre-peak diffusion, survivorship bias in pioneer samples, selection into entry, endogenous innovation choice, event-study and factor-model confounds, contagion-versus-homophily. A future edition should pair each with its remedy (Bayesian pooling, Heckman correction, instrumental variables, calendar-time portfolios, network-autocorrelation controls), so the chapter teaches how the field adjudicates, not only what it concluded. The worked Bass section below models this.
A refreshed diffusion-frontier block. The contagion and machine-learning modules turn over fastest; replace or supplement frontier readings as new network-identification methods and large-scale predictive diffusion models appear, keeping the Bass and takeoff anchors fixed.
Emerging modules as the field grows: generative AI in the development process (concept generation, synthetic concept testing); digital and platform diffusion (apps, two-sided adoption, virality); sustainability and responsible innovation; and the diffusion of services and business models rather than durables. Each should follow the template—foundational anchor + frontier paper + identification debate.

63.18 The Bass diffusion model

The single most-estimated object in this seminar is the Bass model of first-purchase diffusion doi:10.1287/mnsc.15.5.215. Its premise is that the population of eventual adopters has fixed size $m$ (the market potential), and that the probability an as-yet-non-adopter adopts at time $t$ is a linear function of the fraction who have already adopted. Let $F(t)$ be the cumulative fraction of the market potential that has adopted by time $t$, with density $f(t)=F'(t)$. The behavioral core is the hazard of adoption—the conditional likelihood of adopting now given not having adopted yet: \[ h(t) \;=\; \frac{f(t)}{1-F(t)} \;=\; p \;+\; q\,F(t), \tag{63.1}\] where $p$ is the coefficient of innovation (external influence: advertising, media, the pull of novelty, operating independently of prior adopters) and $q$ is the coefficient of imitation (internal influence: word of mouth, social contagion, proportional to the installed adopter base $F(t)$). The two parameters encode the qualitative split between innovators and imitators as a single linear hazard.

Equation Equation 63.1 is a separable ordinary differential equation. Writing $f = (p + qF)(1-F)$ and integrating with $F(0)=0$ yields the closed-form cumulative-adoption curve: \[ F(t) \;=\; \frac{1 - e^{-(p+q)\,t}}{1 + \dfrac{q}{p}\,e^{-(p+q)\,t}}, \tag{63.2}\] and the corresponding non-cumulative (sales) density: \[ f(t) \;=\; \frac{(p+q)^2}{p}\;\frac{e^{-(p+q)\,t}}{\left(1 + \dfrac{q}{p}\,e^{-(p+q)\,t}\right)^{2}} . \tag{63.3}\] Observed unit sales at $t$ are $S(t)=m\,f(t)$, and the model is typically estimated on $S(t)$ or on cumulative sales $m\,F(t)$. When $q>p$—the empirically usual case—the sales curve $S(t)$ is single-peaked and bell-shaped: imitation amplifies early innovator adoption, drives the rapid-growth phase, and then exhausts the remaining market.

Setting $f'(t)=0$ gives the peak-sales time, the moment the diffusion curve turns over: \[ t^{*} \;=\; \frac{1}{p+q}\,\ln\!\left(\frac{q}{p}\right), \tag{63.4}\] which is positive precisely when $q>p$ (imitation dominates innovation). At $t^{*}$ the cumulative penetration is $F(t^{*}) = \tfrac{1}{2}\!\left(1 - p/q\right)$, so a pure-innovation product ($q=0$) peaks at launch, while strong imitation pushes the peak later and past the midpoint of the eventual market. Expressions 1–Equation 63.4 are the working toolkit: given $(m,p,q)$ they generate a full forecast, and given a sales history they are inverted to recover the parameters.

Estimation, however, is where the model’s reputation is made and unmade. Discrete-time nonlinear least squares on the regression form $S(t)=\beta_1 + \beta_2\,Y(t-1) + \beta_3\,Y(t-1)^2 + \varepsilon_t$, where $Y(t-1)$ is cumulative sales, recovers $(m,p,q)$ from $(\beta_1,\beta_2,\beta_3)$ via $m$ as a root of a quadratic, $p=\beta_1/m$, and $q=-\beta_3 m$. Three identification caveats recur throughout the seminar. First, the estimates are weakly identified before the peak: until the data bracket $t^{*}$, the likelihood is flat along a ridge that trades off $m$ against $q$, so market potential and the imitation rate are nearly unidentified exactly when a forecast would be most valuable. Second, the parameters are unstable—the well-known critique that $\hat{p}$, $\hat{q}$, and $\hat{m}$ shift, sometimes implausibly, as each new period of data arrives, undermining the model’s use as a genuine forecaster rather than an after-the-fact fit. Third, the discrete regression form induces correlated errors and finite-sample bias in $(\beta_1,\beta_2,\beta_3)$, which maximum-likelihood and Bayesian estimation on the underlying hazard partly remedy. These caveats motivate the cross-country Bayesian pooling of Week 3 (borrow strength across products to pin $m$), the takeoff and saddle models of Week 4 (which add the structure the smooth curve omits), and the contagion identification of Week 14 (which asks whether the “imitation” coefficient $q$ even measures social influence rather than common exogenous shocks).

63.19 Key Takeaways

The seminar braids three distinct projects—predicting diffusion (Bass and its generalizations), managing development (the voice-of-the-customer and design tradition), and valuing innovation (the marketing–finance interface)—and a doctoral student must move fluently among forecasting, process design, and asset pricing.
The Bass model doi:10.1287/mnsc.15.5.215 reduces first-purchase diffusion to a linear hazard in two parameters—innovation $p$ and imitation $q$ (Equation 63.1)—with a closed-form $S$-curve
1. and peak at $t^{*}=\ln(q/p)/(p+q)$ (Equation 63.4); its pre-peak weak identification and parameter instability are the field’s defining estimation problem.
Diffusion generalizes across technology generations and countries, and real sales curves show takeoff and saddles the smooth Bass curve cannot produce doi:10.1287/mksc.16.3.256; doi:10.1509/jmkg.66.2.1.18472—each extension trades parsimony for fit and reopens identification.
On the firm side, pioneer advantage shrinks once survivorship bias is corrected doi:10.1177/002224379303000203, the incumbent’s curse is weaker than its legend doi:10.1509/jmkg.64.3.1.18033, and really-new products demand different preference-measurement and entry-timing logic doi:10.1509/jmkr.40.4.406.19394.
Innovation is ultimately a priced, risky asset: radical innovation and total stock-market returns to innovation doi:10.1287/mksc.1080.0407 link new products to firm value and risk, while disruption is plausibly a demand-side, not purely supply-side, phenomenon doi:10.1002/smj.246.
The diffusion frontier is an identification frontier: distinguishing genuine social contagion from homophily and common shocks doi:10.1086/320819 is the modern version of the Bass model’s original question about what the imitation coefficient really measures.

# Innovation, New Products, and Diffusion Seminar {#sec-innovation-npd-seminar} *Innovation* is the engine of growth, and new products are where marketing, engineering, and finance collide. This chapter is a graduate seminar in how new products are conceived, developed, launched, adopted, and—if they survive—diffuse through a market until they remake or destroy the firms that compete to supply them. It organizes the field not as a checklist of stage-gate steps but as a tour of the questions a doctoral student must be able to pose and answer: how does a new product spread through a population, and can its sales be predicted before its peak? how is the development process itself managed, from the voice of the customer to a finished design? when does pioneering pay, and when does it merely advertise the opportunity to a fast follower? does radical innovation reward the firms that bet on it—and do capital markets price that bet correctly? The chapter matters because innovation is simultaneously the most celebrated and the least controllable strategic lever. A single diffusion curve can be fit with two behavioral parameters, yet those parameters are notoriously unstable before the sales peak; a development process can be mapped into clean decision modules, yet the hardest choices—how radical to be, when to cannibalize, when to enter—resist optimization. The seminar's central tension runs through every week: there are three distinct intellectual projects here, and they pull in different directions. *Predicting and modeling diffusion* (the Bass tradition and its descendants) treats the new product as an aggregate sales trajectory to be forecast; *managing the development process* (the design and voice-of-the-customer tradition) treats it as a sequence of engineering and marketing decisions to be optimized; and *valuing innovation* (the marketing–finance tradition) treats it as a risky asset to be priced by capital markets. A student who finishes the course should be able to estimate a diffusion model and state what breaks its identification, design a concept test, and read an event study of an innovation announcement with equal fluency. This seminar is the doctoral reading-map that organizes the substantive material of the innovation chapter (@sec-innovation) into a full field course, and it shares a border with the market-entry chapter (@sec-market-entry): timing of entry, pioneer advantage, and order effects are studied here as facets of the innovation problem rather than as a separate competitive-strategy question. The course is also a strong fit for the book's advisor corpus. **Gerard J. Tellis** has shaped the modern empirical study of new-product growth—international takeoff, the saddle, radical versus incremental innovation, the incumbent's curse, technological evolution, and the stock-market returns to innovation—and his work anchors several weeks below. These are woven where genuinely canonical, not gratuitously: the seminar's spine would look much the same in any top program, and it happens to run through a great deal of one advisor's research. ## Semester arc The arc begins with **diffusion as an aggregate phenomenon**. The first four weeks build the modeling backbone—the Bass model and its generalizations, cross-country and international diffusion, and the empirical regularities of takeoff and the "saddle"—because a doctoral student must be able to write down, estimate, and critique a diffusion model before doing anything else with new products. This block establishes the field's first dependent variable, the *sales trajectory*, and its recurring identification problem: the curve's parameters are weakly identified from pre-peak data, so forecasts made when they are most useful are least reliable. The middle of the semester turns from the *market's* behavior to the *firm's* behavior: the **new-product development process** itself. Weeks on the development- decisions framework, the voice of the customer and concept testing, and preference measurement and product-design optimization treat the new product as something to be engineered and tested, not merely forecast. Here the seminar connects to the preference-measurement and conjoint machinery used throughout the book, and it foregrounds a different methodological tradition—survey design, perceptual mapping, and choice modeling rather than time-series estimation. The block on really-new products and entry timing then bridges process and strategy: when the product is genuinely new, both customer learning and competitive order-of-entry effects change the rules. The seminar closes with **innovation as strategy and as a priced asset**. Weeks on radical versus incremental innovation and the incumbent, technology generations and network effects, user and open innovation, innovation and firm value, technological evolution and disruption, and a diffusion frontier (social contagion and machine learning) move the course from forecasting and process toward the questions that animate current dissertations: does radical innovation pay, and for whom? do incumbents really suffer a curse? do markets price innovation efficiently? Throughout, the pedagogy is method-forward—each module names the identification challenge (parameter instability, selection into entry, endogenous innovation choice, event-study and factor-model confounds, contagion-versus-homophily) that stands between a pattern and a finding. The reading map uses two tags: **[F] = Foundational** (canon an innovation scholar is expected to know cold) and **[R] = Frontier/Recent** (an active research front, refreshed as the literature moves). Each week pairs at least one foundational anchor with one frontier paper. DOIs are reproduced as verified against Crossref; works without a DOI-verified record (chiefly books and a pre-DOI economics classic) are named without a link and flagged. ## Week 1 — Diffusion and the Bass model {#sec-inno-week01} **Topic.** How a new product spreads through a population of potential adopters; the canonical aggregate model of first-purchase diffusion. **Subtopics.** Innovators vs. imitators; the adopter categories; the $S$-curve; coefficients of innovation and imitation; the qualitative theory of diffusion behind the algebra. **Methods.** Aggregate (macro) diffusion modeling; nonlinear least squares and maximum likelihood on cumulative-adoption data; the sociological theory of adoption. **Key readings.** - Bass (1969), "A New Product Growth for Model Consumer Durables," *Management Science*. [doi:10.1287/mnsc.15.5.215](https://doi.org/10.1287/mnsc.15.5.215) — the founding two-parameter diffusion model; the anchor for the entire course. **[F]** - Rogers, *Diffusion of Innovations* (Free Press; 1st ed. 1962, 5th ed. 2003) — the qualitative theory of adopter categories and innovation attributes behind the algebra; a book, cited without DOI. **[F]** - Mahajan, Muller & Bass (1990), "New Product Diffusion Models in Marketing: A Review and Directions for Research," *Journal of Marketing*. [doi:10.1177/002224299005400101](https://doi.org/10.1177/002224299005400101) — the field-defining review that frames the whole modeling tradition (read here as orientation, developed in Week 2). **[F]** **Debate.** Is the $p$/$q$ split a behavioral truth (external vs. internal influence) or a convenient functional form? Do innovators and imitators name real people or just curvature? ## Week 2 — Bass extensions and generalized diffusion {#sec-inno-week02} **Topic.** Relaxing the basic model: marketing-mix effects, repeat purchase, flexible curves, and the conditions under which extensions actually improve fit and forecasts. **Subtopics.** The generalized Bass model with price and advertising; why decision variables often add little to fit; flexible and multi-stage diffusion; refinement vs. parsimony. **Methods.** Estimation under added covariates; nested-model comparison; forecasting evaluation and the bias of pre-peak estimates. **Key readings.** - Mahajan, Muller & Bass (1990), "New Product Diffusion Models in Marketing: A Review and Directions for Research," *Journal of Marketing*. [doi:10.1177/002224299005400101](https://doi.org/10.1177/002224299005400101) — the canonical taxonomy of extensions and open problems. **[F]** - Bass, Krishnan & Jain (1994), "Why the Bass Model Fits Without Decision Variables," *Marketing Science*. [doi:10.1287/mksc.13.3.203](https://doi.org/10.1287/mksc.13.3.203) — the generalized Bass model, and the subtle argument that mix effects enter multiplicatively without disturbing the basic shape. **[F]** - Norton & Bass (1987), "A Diffusion Theory Model of Adoption and Substitution for Successive Generations of High-Technology Products," *Management Science*. [doi:10.1287/mnsc.33.9.1069](https://doi.org/10.1287/mnsc.33.9.1069) — extends diffusion to technology generations (revisited in Week 10). **[F]** **Debate.** When does adding price/advertising to a diffusion model buy real forecasting power, and when is it overfitting? Is the generalized Bass model an explanation or a rationalization of the basic model's robustness? ## Week 3 — Cross-country and international diffusion {#sec-inno-week03} **Topic.** How diffusion parameters and takeoff timing vary across countries, and what economic, cultural, and institutional factors drive the variation. **Subtopics.** Cross-national differences in $p$ and $q$; lead–lag effects between countries; developing-country diffusion; culture and country innovativeness as moderators. **Methods.** Hierarchical/Bayesian estimation across countries and products; meta-analytic synthesis of parameters; hazard models of takeoff timing. **Key readings.** - Tellis, Stremersch & Yin (2003), "The International Takeoff of New Products: The Role of Economics, Culture, and Country Innovativeness," *Marketing Science*. [doi:10.1287/mksc.22.2.188.16041](https://doi.org/10.1287/mksc.22.2.188.16041) — what makes new products take off faster in some countries than others. **[F]** - Talukdar, Sudhir & Ainslie (2002), "Investigating New Product Diffusion Across Products and Countries," *Marketing Science*. [doi:10.1287/mksc.21.1.97.161](https://doi.org/10.1287/mksc.21.1.97.161) — a Bayesian decomposition of diffusion parameters across products and developed/ developing markets. **[F]** **Debate.** Are cross-country differences in diffusion driven by economics (income, infrastructure) or culture (uncertainty avoidance, cosmopolitanism)? Do later-adopting countries diffuse faster because they "learn" from earlier ones? ## Week 4 — Takeoff, slowdown, and saddles {#sec-inno-week04} **Topic.** The non-monotone features of real sales curves: the long flat introductory stage, the sudden takeoff, and the post-takeoff sales dip ("saddle") that the smooth Bass curve cannot produce. **Subtopics.** Modeling takeoff as a hazard; really-new vs. incrementally-new takeoff; the saddle and the chasm; informational cascades and heterogeneity as mechanisms. **Methods.** Hazard/duration models of takeoff; threshold and two-segment diffusion models; identifying structural breaks in sales. **Key readings.** - Golder & Tellis (1997), "Will It Ever Fly? Modeling the Takeoff of Really New Consumer Durables," *Marketing Science*. [doi:10.1287/mksc.16.3.256](https://doi.org/10.1287/mksc.16.3.256) — defines and predicts the takeoff point, the transition from introduction to rapid growth. **[F]** - Goldenberg, Libai & Muller (2002), "Riding the Saddle: How Cross-Market Communications Can Create a Major Slump in Sales," *Journal of Marketing*. [doi:10.1509/jmkg.66.2.1.18472](https://doi.org/10.1509/jmkg.66.2.1.18472) — the saddle as a consequence of weak communication between early and main markets. **[F]** - Chandrasekaran & Tellis (2011), "Getting a Grip on the Saddle: Chasms or Cycles?," *Journal of Marketing*. [doi:10.1509/jmkg.75.4.21](https://doi.org/10.1509/jmkg.75.4.21) — large-sample evidence on how prevalent and how deep saddles are. **[R]** **Debate.** Is the saddle a structural feature of two-segment markets or an artifact of macroeconomic cycles? Can takeoff be forecast in time to act on it? ## Week 5 — The new-product development process {#sec-inno-week05} **Topic.** Turning the lens from the market to the firm: how new products are actually developed, and the decisions that structure the process. **Subtopics.** The development-decisions framework (concept, design, testing, production ramp); stage-gate and its critics; platform and portfolio decisions; the marketing–engineering interface. **Methods.** Decision-analytic and process frameworks; design-process modeling; case-based and survey evidence on development practice. **Key readings.** - Krishnan & Ulrich (2001), "Product Development Decisions: A Review of the Literature," *Management Science*. [doi:10.1287/mnsc.47.1.1.10668](https://doi.org/10.1287/mnsc.47.1.1.10668) — organizes the sprawling NPD literature as a set of decisions across disciplines. **[F]** - Urban & Hauser, *Design and Marketing of New Products* (2nd ed., Prentice Hall, 1993) — the canonical text mapping the development funnel from opportunity identification to launch; a book, cited without DOI. **[F]** **Debate.** Is NPD best modeled as a sequence of decisions (Krishnan–Ulrich) or as an organizational/cultural capability? Does formal stage-gate help or ossify radical projects? ## Week 6 — Voice of the customer and concept testing {#sec-inno-week06} **Topic.** Eliciting and structuring customer needs as the front end of development; turning fuzzy wants into engineering targets. **Subtopics.** The voice of the customer (VOC); need hierarchies and relative importance; quality function deployment (QFD); concept generation and screening. **Methods.** Qualitative-interview elicitation; affinity structuring; reliability of needs measurement; concept-test design. **Key readings.** - Griffin & Hauser (1993), "The Voice of the Customer," *Marketing Science*. [doi:10.1287/mksc.12.1.1](https://doi.org/10.1287/mksc.12.1.1) — how many customers and analysts are needed to capture the structured set of needs; the measurement foundation of the front end. **[F]** - Urban & Hauser, *Design and Marketing of New Products* (2nd ed., 1993) — the QFD "house of quality" and concept-testing pipeline (revisited from Week 5 through the needs-measurement lens); a book, cited without DOI. **[F]** **Debate.** Does VOC reliably surface latent needs, or only articulated ones? Are really-new products knowable from customer interviews at all (a hand-off to Week 8)? ## Week 7 — Preference measurement and product-design optimization {#sec-inno-week07} **Topic.** Quantifying how customers value product attributes, and using those valuations to design and position the product. **Subtopics.** Conjoint analysis (ratings-based and choice-based); attribute part-worths; product-line and design optimization; commercial practice and its evolution. **Methods.** Conjoint/choice experiments; hierarchical Bayes estimation of part-worths; optimization over the attribute space; perceptual mapping. **Key readings.** - Green & Srinivasan (1990), "Conjoint Analysis in Marketing: New Developments with Implications for Research and Practice," *Journal of Marketing*. [doi:10.1177/002224299005400402](https://doi.org/10.1177/002224299005400402) — the authoritative synthesis of conjoint method and design. **[F]** - Wittink & Cattin (1989), "Commercial Use of Conjoint Analysis: An Update," *Journal of Marketing*. [doi:10.1177/002224298905300310](https://doi.org/10.1177/002224298905300310) — documents how widely and how conjoint is used in practice. **[F]** **Debate.** Do stated-preference part-worths predict real new-product choice? When does choice-based conjoint beat ratings-based, and at what cost in design efficiency? ## Week 8 — Really-new products and entry timing {#sec-inno-week08} **Topic.** What changes when the product is genuinely new—and when does being first to a new market pay? **Subtopics.** Measuring preferences for really-new products; pioneer advantage vs. pioneer myth; order-of-entry effects; survival risks of pioneers vs. followers. **Methods.** Preference elicitation under unfamiliarity (information acceleration); survival/hazard analysis of entrants; survivorship-bias correction in pioneer samples. **Key readings.** - Hoeffler (2003), "Measuring Preferences for Really New Products," *Journal of Marketing Research*. [doi:10.1509/jmkr.40.4.406.19394](https://doi.org/10.1509/jmkr.40.4.406.19394) — why conventional preference measurement fails for unfamiliar products, and what to do instead. **[F]** - Golder & Tellis (1993), "Pioneer Advantage: Marketing Logic or Marketing Legend?," *Journal of Marketing Research*. [doi:10.1177/002224379303000203](https://doi.org/10.1177/002224379303000203) — shows that survivorship bias inflates the apparent advantage of market pioneers. **[F]** - Min, Kalwani & Robinson (2006), "Market Pioneer and Early Follower Survival Risks: A Contingency Analysis of Really New Versus Incrementally New Product-Markets," *Journal of Marketing*. [doi:10.1509/jmkg.70.1.015.qxd](https://doi.org/10.1509/jmkg.70.1.015.qxd) — pioneer survival depends on whether the market is really-new or incremental. **[R]** **Debate.** Is pioneer advantage real once survivorship bias is corrected? Does the "really-new" distinction reverse the entry-timing prescription? ## Week 9 — Radical vs. incremental innovation and the incumbent {#sec-inno-week09} **Topic.** The organizational determinants of radical innovation, and whether large incumbents are doomed to lose the radical-innovation race. **Subtopics.** Willingness to cannibalize; organizational antecedents of radical product innovation; the incumbent's curse (and its limits); size and incumbency effects. **Methods.** Survey measurement of organizational constructs; historical content analysis of innovation databases; logit models of who innovates radically. **Key readings.** - Chandy & Tellis (1998), "Organizing for Radical Product Innovation: The Overlooked Role of Willingness to Cannibalize," *Journal of Marketing Research*. [doi:10.1177/002224379803500406](https://doi.org/10.1177/002224379803500406) — willingness to cannibalize existing assets as the key organizational driver. **[F]** - Chandy & Tellis (2000), "The Incumbent's Curse? Incumbency, Size, and Radical Product Innovation," *Journal of Marketing*. [doi:10.1509/jmkg.64.3.1.18033](https://doi.org/10.1509/jmkg.64.3.1.18033) — historical evidence that incumbents and large firms innovate radically more often than the "curse" narrative claims. **[F]** **Debate.** Is the incumbent's curse a robust regularity or a sampling artifact of celebrated failures? Is willingness to cannibalize a cause of radical innovation or a rationalization after the fact? ## Week 10 — Technology generations and network effects {#sec-inno-week10} **Topic.** Innovation as a sequence of substituting generations, and the demand-side forces—network externalities—that govern adoption of compatible technologies. **Subtopics.** Successive-generation diffusion and substitution; installed base and adoption; network externalities and compatibility; standards and tipping. **Methods.** Multi-generation diffusion models; structural models of adoption under network effects; analytical models of compatibility competition. **Key readings.** - Norton & Bass (1987), "A Diffusion Theory Model of Adoption and Substitution for Successive Generations of High-Technology Products," *Management Science*. [doi:10.1287/mnsc.33.9.1069](https://doi.org/10.1287/mnsc.33.9.1069) — diffusion when a new generation cannibalizes and extends the old. **[F]** - Katz & Shapiro (1985), "Network Externalities, Competition, and Compatibility," *American Economic Review* — the foundational analytical treatment of demand-side network effects and compatibility; a pre-DOI classic cited without a verified Crossref link (flagged). **[F]** **Debate.** Do network effects "tip" markets to a single standard, or can incompatible technologies coexist? Does successive-generation substitution follow the same $p$/$q$ logic as first-purchase diffusion? ## Week 11 — User innovation, open innovation, and crowdfunding {#sec-inno-week11} **Topic.** Innovation that originates outside the firm—among lead users and crowds—and the platforms that finance and surface it. **Subtopics.** Lead users as a source of concepts; user vs. manufacturer innovation; open and distributed innovation; crowdfunding dynamics and signals of success. **Methods.** Lead-user identification and field studies; analysis of platform/ crowdfunding data; hazard and regression models of funding success. **Key readings.** - von Hippel (1986), "Lead Users: A Source of Novel Product Concepts," *Management Science*. [doi:10.1287/mnsc.32.7.791](https://doi.org/10.1287/mnsc.32.7.791) — the founding statement that ahead-of-market users generate the most valuable concepts. **[F]** - Mollick (2014), "The Dynamics of Crowdfunding: An Exploratory Study," *Journal of Business Venturing*. [doi:10.1016/j.jbusvent.2013.06.005](https://doi.org/10.1016/j.jbusvent.2013.06.005) — what predicts crowdfunding success, and how delivery actually plays out. **[R]** **Debate.** Do lead users predict mass-market needs or only niche ones? Is crowdfunding a market test, a financing mechanism, or a marketing event—and does the distinction change what the data mean? ## Week 12 — Innovation and firm value {#sec-inno-week12} **Topic.** Whether and how innovation translates into shareholder value and risk—the marketing–finance interface applied to new products. **Subtopics.** Radical innovation and financial consequences; innovation as a driver of abnormal stock returns; value vs. risk effects of radical vs. incremental innovation; efficient vs. biased market reactions. **Methods.** Content-analyzed innovation databases; event studies; calendar-time portfolios and long-run abnormal returns; risk decomposition. **Key readings.** - Sorescu, Chandy & Prabhu (2003), "Sources and Financial Consequences of Radical Innovation: Insights from Pharmaceuticals," *Journal of Marketing*. [doi:10.1509/jmkg.67.4.82.18687](https://doi.org/10.1509/jmkg.67.4.82.18687) — which firms produce radical innovations and how markets value them. **[F]** - Sood & Tellis (2009), "Do Innovations Really Pay Off? Total Stock Market Returns to Innovation," *Marketing Science*. [doi:10.1287/mksc.1080.0407](https://doi.org/10.1287/mksc.1080.0407) — measuring the full stock-market return to innovation across the project life cycle, not just at announcement. **[R]** - Sorescu & Spanjol (2008), "Innovation's Effect on Firm Value and Risk: Insights from Consumer Packaged Goods," *Journal of Marketing*. [doi:10.1509/jmkg.72.2.114](https://doi.org/10.1509/jmkg.72.2.114) — distinguishes the value and risk effects of radical vs. incremental innovation. **[R]** **Debate.** Does radical innovation pay on a risk-adjusted basis? Are the abnormal returns evidence of mispricing or compensation for risk—and does measuring returns across the whole life cycle change the verdict? ## Week 13 — Technological evolution and disruption {#sec-inno-week13} **Topic.** How technologies evolve over time and across competing trajectories, and when new technologies disrupt established ones. **Subtopics.** Step-function vs. smooth technological evolution; multiple competing technologies and crossings; disruptive vs. sustaining innovation; the demand-side account of disruption. **Methods.** Long-run performance-trajectory data; technology S-curve and crossing analysis; demand-based modeling of segment overlap. **Key readings.** - Sood & Tellis (2005), "Technological Evolution and Radical Innovation," *Journal of Marketing*. [doi:10.1509/jmkg.69.3.152.66361](https://doi.org/10.1509/jmkg.69.3.152.66361) — evidence that technologies evolve through irregular step functions with many crossings, not a single smooth S-curve. **[F]** - Christensen, *The Innovator's Dilemma* (Harvard Business School Press, 1997) — the disruptive-innovation thesis that well-managed incumbents fail by listening to their best customers; a book, cited without DOI. **[F]** - Adner (2002), "When Are Technologies Disruptive? A Demand-Based View of the Emergence of Competition," *Strategic Management Journal*. [doi:10.1002/smj.246](https://doi.org/10.1002/smj.246) — recasts disruption as a consequence of demand-side segment structure rather than supply-side technology. **[R]** **Debate.** Is disruption a supply-side (technology) or demand-side (preference- overlap) phenomenon? Does the evidence on technological evolution support or undercut the disruption narrative? ## Week 14 — Diffusion frontier: social contagion, machine learning, synthesis {#sec-inno-week14} **Topic.** The modern micro-foundations of diffusion—individual-level contagion and networks—and the methodological frontier; closing synthesis. **Subtopics.** Opinion leadership and social contagion; contagion vs. homophily (the identification problem); network-based and machine-learning diffusion models; what the field knows and what remains open. **Methods.** Network panel data; contagion identification (instruments, structural controls); ML/predictive diffusion modeling; conceptual synthesis. **Key readings.** - Iyengar, Van den Bulte & Valente (2011), "Opinion Leadership and Social Contagion in New Product Diffusion," *Marketing Science*. [doi:10.1287/mksc.1100.0566](https://doi.org/10.1287/mksc.1100.0566) — sociometric and self-reported opinion leadership and how contagion operates through networks. **[R]** - Van den Bulte & Lilien (2001), "Medical Innovation Revisited: Social Contagion Versus Marketing Effort," *American Journal of Sociology*. [doi:10.1086/320819](https://doi.org/10.1086/320819) — the landmark demonstration that apparent contagion can vanish once marketing effort is controlled; the identification cautionary tale. **[F]** **Debate.** Is observed adoption clustering genuine contagion or homophily/common exposure? Do ML diffusion models forecast better while explaining less—and does the field want prediction or mechanism? ## Foundational vs. frontier at a glance {#sec-inno-ff} **Foundational core** (every innovation student must know): Bass (1969); Rogers (*Diffusion of Innovations*); Katz & Shapiro (1985); von Hippel (1986); Norton & Bass (1987); Mahajan, Muller & Bass (1990); Green & Srinivasan (1990); Wittink & Cattin (1989); Griffin & Hauser (1993); Golder & Tellis (1993); Urban & Hauser (*Design and Marketing of New Products*); Bass, Krishnan & Jain (1994); Golder & Tellis (1997); Christensen (1997); Chandy & Tellis (1998, 2000); Krishnan & Ulrich (2001); Van den Bulte & Lilien (2001); Goldenberg, Libai & Muller (2002); Talukdar, Sudhir & Ainslie (2002); Tellis, Stremersch & Yin (2003); Hoeffler (2003); Sorescu, Chandy & Prabhu (2003). **Frontier / actively updated** (refresh each edition): Sood & Tellis (2005, 2009); Min, Kalwani & Robinson (2006); Sorescu & Spanjol (2008); Adner (2002, read as a frontier reframing); Chandrasekaran & Tellis (2011); Iyengar, Van den Bulte & Valente (2011); Mollick (2014). The split is pedagogical, not chronological: Bass (1969) is foundational because the field still estimates its model weekly, while a 2002 demand-based theory of disruption is "frontier" because its agenda is still being executed. Each module pairs at least one foundational anchor with a live edge so students meet both the canon and its open questions. ## How this chapter expands {#sec-inno-expands} The weekly map is a backbone, not a ceiling. It is designed to grow along several axes: 1. **An estimation companion per week.** Each module names an identification challenge—parameter instability in pre-peak diffusion, survivorship bias in pioneer samples, selection into entry, endogenous innovation choice, event-study and factor-model confounds, contagion-versus-homophily. A future edition should pair each with its remedy (Bayesian pooling, Heckman correction, instrumental variables, calendar-time portfolios, network-autocorrelation controls), so the chapter teaches *how the field adjudicates*, not only what it concluded. The worked Bass section below models this. 2. **A refreshed diffusion-frontier block.** The contagion and machine-learning modules turn over fastest; replace or supplement frontier readings as new network-identification methods and large-scale predictive diffusion models appear, keeping the Bass and takeoff anchors fixed. 3. **Emerging modules as the field grows:** generative AI in the development process (concept generation, synthetic concept testing); digital and platform diffusion (apps, two-sided adoption, virality); sustainability and responsible innovation; and the diffusion of services and business models rather than durables. Each should follow the template—foundational anchor + frontier paper + identification debate. ## The Bass diffusion model {#sec-inno-worked} The single most-estimated object in this seminar is the **Bass model** of first-purchase diffusion [doi:10.1287/mnsc.15.5.215](https://doi.org/10.1287/mnsc.15.5.215). Its premise is that the population of eventual adopters has fixed size $m$ (the *market potential*), and that the probability an as-yet-non-adopter adopts at time $t$ is a linear function of the fraction who have already adopted. Let $F(t)$ be the cumulative fraction of the market potential that has adopted by time $t$, with density $f(t)=F'(t)$. The behavioral core is the **hazard** of adoption—the conditional likelihood of adopting now given not having adopted yet: $$ h(t) \;=\; \frac{f(t)}{1-F(t)} \;=\; p \;+\; q\,F(t), $$ {#eq-inno-hazard} where $p$ is the **coefficient of innovation** (external influence: advertising, media, the pull of novelty, operating independently of prior adopters) and $q$ is the **coefficient of imitation** (internal influence: word of mouth, social contagion, proportional to the installed adopter base $F(t)$). The two parameters encode the qualitative split between innovators and imitators as a single linear hazard. Equation @eq-inno-hazard is a separable ordinary differential equation. Writing $f = (p + qF)(1-F)$ and integrating with $F(0)=0$ yields the closed-form **cumulative-adoption curve**: $$ F(t) \;=\; \frac{1 - e^{-(p+q)\,t}}{1 + \dfrac{q}{p}\,e^{-(p+q)\,t}}, $$ {#eq-inno-cdf} and the corresponding **non-cumulative (sales) density**: $$ f(t) \;=\; \frac{(p+q)^2}{p}\;\frac{e^{-(p+q)\,t}}{\left(1 + \dfrac{q}{p}\,e^{-(p+q)\,t}\right)^{2}} . $$ {#eq-inno-pdf} Observed unit sales at $t$ are $S(t)=m\,f(t)$, and the model is typically estimated on $S(t)$ or on cumulative sales $m\,F(t)$. When $q>p$—the empirically usual case—the sales curve $S(t)$ is single-peaked and bell-shaped: imitation amplifies early innovator adoption, drives the rapid-growth phase, and then exhausts the remaining market. Setting $f'(t)=0$ gives the **peak-sales time**, the moment the diffusion curve turns over: $$ t^{*} \;=\; \frac{1}{p+q}\,\ln\!\left(\frac{q}{p}\right), $$ {#eq-inno-peak} which is positive precisely when $q>p$ (imitation dominates innovation). At $t^{*}$ the cumulative penetration is $F(t^{*}) = \tfrac{1}{2}\!\left(1 - p/q\right)$, so a pure-innovation product ($q=0$) peaks at launch, while strong imitation pushes the peak later and past the midpoint of the eventual market. Expressions @eq-inno-cdf–@eq-inno-peak are the working toolkit: given $(m,p,q)$ they generate a full forecast, and given a sales history they are inverted to recover the parameters. Estimation, however, is where the model's reputation is made and unmade. Discrete-time nonlinear least squares on the regression form $S(t)=\beta_1 + \beta_2\,Y(t-1) + \beta_3\,Y(t-1)^2 + \varepsilon_t$, where $Y(t-1)$ is cumulative sales, recovers $(m,p,q)$ from $(\beta_1,\beta_2,\beta_3)$ via $m$ as a root of a quadratic, $p=\beta_1/m$, and $q=-\beta_3 m$. Three identification caveats recur throughout the seminar. First, the estimates are **weakly identified before the peak**: until the data bracket $t^{*}$, the likelihood is flat along a ridge that trades off $m$ against $q$, so market potential and the imitation rate are nearly unidentified exactly when a forecast would be most valuable. Second, the parameters are **unstable**—the well-known critique that $\hat{p}$, $\hat{q}$, and $\hat{m}$ shift, sometimes implausibly, as each new period of data arrives, undermining the model's use as a genuine forecaster rather than an after-the-fact fit. Third, the discrete regression form induces **correlated errors and finite-sample bias** in $(\beta_1,\beta_2,\beta_3)$, which maximum-likelihood and Bayesian estimation on the underlying hazard partly remedy. These caveats motivate the cross-country Bayesian pooling of Week 3 (borrow strength across products to pin $m$), the takeoff and saddle models of Week 4 (which add the structure the smooth curve omits), and the contagion identification of Week 14 (which asks whether the "imitation" coefficient $q$ even measures social influence rather than common exogenous shocks). ## Key Takeaways - The seminar braids three distinct projects—**predicting diffusion** (Bass and its generalizations), **managing development** (the voice-of-the-customer and design tradition), and **valuing innovation** (the marketing–finance interface)—and a doctoral student must move fluently among forecasting, process design, and asset pricing. - The **Bass model** [doi:10.1287/mnsc.15.5.215](https://doi.org/10.1287/mnsc.15.5.215) reduces first-purchase diffusion to a linear hazard in two parameters—innovation $p$ and imitation $q$ (@eq-inno-hazard)—with a closed-form $S$-curve (@eq-inno-cdf) and peak at $t^{*}=\ln(q/p)/(p+q)$ (@eq-inno-peak); its pre-peak weak identification and parameter instability are the field's defining estimation problem. - Diffusion generalizes across **technology generations** and **countries**, and real sales curves show **takeoff** and **saddles** the smooth Bass curve cannot produce [doi:10.1287/mksc.16.3.256](https://doi.org/10.1287/mksc.16.3.256); [doi:10.1509/jmkg.66.2.1.18472](https://doi.org/10.1509/jmkg.66.2.1.18472)—each extension trades parsimony for fit and reopens identification. - On the firm side, **pioneer advantage** shrinks once survivorship bias is corrected [doi:10.1177/002224379303000203](https://doi.org/10.1177/002224379303000203), the **incumbent's curse** is weaker than its legend [doi:10.1509/jmkg.64.3.1.18033](https://doi.org/10.1509/jmkg.64.3.1.18033), and **really-new products** demand different preference-measurement and entry-timing logic [doi:10.1509/jmkr.40.4.406.19394](https://doi.org/10.1509/jmkr.40.4.406.19394). - Innovation is ultimately a **priced, risky asset**: radical innovation and total stock-market returns to innovation [doi:10.1287/mksc.1080.0407](https://doi.org/10.1287/mksc.1080.0407) link new products to firm value and risk, while **disruption** is plausibly a demand-side, not purely supply-side, phenomenon [doi:10.1002/smj.246](https://doi.org/10.1002/smj.246). - The diffusion frontier is an **identification frontier**: distinguishing genuine social contagion from homophily and common shocks [doi:10.1086/320819](https://doi.org/10.1086/320819) is the modern version of the Bass model's original question about what the imitation coefficient really measures.