Technology Life Cycle Bibliometrics

Because technologies typically diffuse through three phases that together trace an S-shape: slow early growth while a few pioneers work on the idea, rapid acceleration as the technology proves itself and many actors adopt it, and a plateau as opportunities are exhausted. Ernst showed that cumulative patenting in CNC technology followed exactly this logistic pattern. Fitting an S-curve captures all three phases with a few parameters, lets you read off the current stage from where activity sits relative to saturation, and supports forecasting by extrapolating toward the estimated ceiling, which simpler linear trends cannot do.

Through carefully constructed queries and classifications. Analysts combine patent classification codes, keyword searches, and journal filters to assemble a corpus that captures the technology while excluding unrelated work. Watts and Porter stressed that this delineation is the most consequential and error-prone step, because the entire life-cycle signal depends on the boundary drawn. Best practice involves iterating the query with domain experts, validating samples for relevance, and testing how sensitive the resulting trend is to reasonable changes in the search, since an ill-defined corpus produces a misleading curve.

Only partially, and Watts and Porter argued it should not be used alone. Extrapolating a fitted S-curve gives a forecast of activity and saturation, but it assumes smooth logistic growth and is unreliable when disruptive jumps or new waves occur, and recent-year data are truncated. Their innovation-forecasting framework therefore combines the life-cycle indicator with contextual influences and value-chain or market indicators, plus content analysis of success factors. Used as one component of this richer assessment, life-cycle bibliometrics is a powerful forecasting aid; used in isolation, it can mislead.