Answers are grounded in this paper and cite their sources.
Recent surveys show that only 1 in 400 startups secure funding and achieve post-fund sustainability. Startup assessment is becoming increasingly complicated due to the evolving entrepreneurial landscape, which is affected by too many factors, compelling venture capital firms to harness state-of-the-art Artificial Intelligence (AI). Natural Language Processing (NLP) offers a promising approach for startup assessment, by analyzing documents that summarize key factors of long-term success. Although startup sustainability research is active, modern NLP for assessing business plans has received limited attention. Specifically, the majority of previous works consider only short text descriptions of startups instead of complete business plans, and do not offer adaptability to the varying investment criteria of different investors. In this paper, we propose ForeStart, a startup sustainability forecasting system powered by a transformer-based language model that offers seamless adaptability to the particular needs of users. The proposed system can analyze business plans of unlimited length, including also startup-related content such as pitch deck presentations and company websites. ForeStart is based on a hierarchical Distilled Bidirectional Encoder Representations from Transformers–Bidirectional Long Short-Term Memory (DistilBERT–BiLSTM) architecture trained using both unsupervised domain adaptation and supervised learning, effectively leveraging both unlabeled and labeled business plans. In addition, the proposed system addresses issues arising from privacy concerns and low-resource settings of the startup environment. Our experiments on actual business plans achieve 79% F₁ score in forecasting startup sustainability, even with a limited number of labeled business plans as training data. As an assisting screening tool for analysts, ForeStart can significantly reduce the time required for startup assessments, providing value to Venture Capital (VC) organizations. Moreover, startup founders can also use ForeStart for self-assessment.
No takes yet. Share an insight, caveat, or question.
Takas et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: