Analyzing large-scale text corpora is a core challenge in machine learning, crucial for tasks like identifying undesirable model behaviors or biases in training data. Current methods often rely on costly LLM-based techniques (e.g. annotating dataset differences) or dense embedding models (e.g. for clustering), which lack control over the properties of interest. We propose using sparse autoencoders (SAEs) to create SAE embeddings: representations whose dimensions map to interpretable concepts. Through four data analysis tasks, we show that SAE embeddings are more cost-effective and reliable than LLMs and more controllable than dense embeddings. Using the large hypothesis space of SAEs, we can uncover insights such as (1) semantic differences between datasets and (2) unexpected concept correlations in documents. For instance, by comparing model responses, we find that Grok-4 clarifies ambiguities more often than nine other frontier models. Relative to LLMs, SAE embeddings uncover bigger differences at 2-8x lower cost and identify biases more reliably. Additionally, SAE embeddings are controllable: by filtering concepts, we can (3) cluster documents along axes of interest and (4) outperform dense embeddings on property-based retrieval. Using SAE embeddings, we study model behavior with two case studies: investigating how OpenAI model behavior has changed over time and finding "trigger" phrases learned by Tulu-3 (Lambert et al., 2024) from its training data. These results position SAEs as a versatile tool for unstructured data analysis and highlight the neglected importance of interpreting models through their data.
Innovation proficiency is widely recognized as essential to organizational competitiveness; yet, how product managers and their teams develop these proficiencies across organizational contexts remains understudied. This study examines six innovation proficiencies—Customer Empathy, Insight Generation, Idea Generation, Idea Selection, Experimentation and Learning, and Mobilizing and Executing—using a longitudinal dataset of 15,842 survey responses collected across 1066 organizations over nine years (2016–2024). Responses were analyzed using descriptive statistics, nonparametric group comparisons, correlation analysis, reliability testing, longitudinal analysis, and systematic qualitative theme extraction from nearly 50,000 responses. Insight Generation is consistently the lowest-rated proficiency despite being ranked high in importance. Organization-weighted analysis finds significant, but modest, improvement in just one of the six proficiencies (Idea Selection) over the study period and positive but non-significant trends in the others. Qualitative analysis identifies customer centricity, lack of formalized processes and data-based decision making as persistent challenges over time. These findings suggest that improvement of innovation practices depends not only on Product Managers and their teams, but on organization-wide infrastructure changes to facilitate and support their innovation work.