LLM-based innovation dynamics analyzer: A novel approach to competitive innovation dynamics in tourism

Tourism and general management scholarship lack scalable tools for tracking how firms vie across successive innovation waves; manual or keyword coding cannot process today's news volume. We introduce the Large Language Model-Based Innovation Dynamics Analyzer (LIDA), a four-step methodological pipeline that (1) detects innovation stories with few-shot GPT-4 prompts, (2) groups them via BERTopic, (3) links them into time-bounded action–reaction chains through HDBSCAN, and (4) gauges market impact using cumulative abnormal returns. Applied to 20,000 news gathered from Restaurant SmartBrief (2011–2024), LIDA delivers the first large-sample map of competitive innovation in tourism. Results show process innovations consistently lift market-value; business-model experiments trigger short-term discounts; and product launches are value-neutral unless demand builds quickly. Absolute pioneers earn no premiums, whereas wave leaders and brief “deliberate followers” gain fleeting rewards. These patterns show that advantage depends on innovation type and timing, not historical primacy. Methodologically, LIDA converts vast unstructured corpora into analyzable competitive-action networks.