HearAdvisor

Independent testing of Fortell published today: HearAdvisor , our industry’s premier sound lab, ranks us well above every competitor. The attached charts are proof — the fruits of intense labor compound continuously. Thank you Abram Bailey , Andrew Sabin , and Steve Taddei for the role you play in helping patients and audiologists navigate the morass of superlative claims each manufacturer makes about its own products. Congratulations Fortell team, and thank you to our patients. More great technology coming down the pike. (Reminder: this is our first hearing aid!) (Data current as of today. Speech-in-noise scores show initial, out-of-box fit.)

Our groundbreaking paper is finally up @ arXiv 🚀 Congrats to the team HearAdvisor! Andrew Sabin / Steve Taddei Abstract: HearAdvisor aims to provide hearing-aid consumers with audio-performance metrics and recordings that reflect real listening experience. For speech-related metrics, HearAdvisor has historically used HASPIv2, a metric designed to predict objective intelligibility and validated primarily under simulated distortions. Its relationship to consumer-rated ease of understanding for commercial hearing aids is uncertain. Here we introduce a large-scale perceptual dataset and learned metric for listener-rated perceived benefit for speech understanding. Website visitors with self-reported hearing loss completed a blind, MUSHRA-inspired listening test in which they rated recordings of commercial hearing aids on a five-point "Ease of Understanding" scale. The dataset contains 151,608 ratings, 104,298 after quality screening, spanning 10,394 binaural acoustic-manikin recordings from 83 commercial products across 72 realistic acoustic scenes. To predict these ratings, we pass aided audio and a matched clean-speech reference through a frozen Whisper encoder, subtract their internal representations, and train a small MLP head on the resulting difference embedding. On devices held out of training, the learned metric substantially outperforms HASPIv2 at the scene level (overall r = 0.92 vs. 0.83; loud = 0.89 vs. 0.75; quiet = 0.79 vs. 0.58). In loud scenes, performance reaches the split-half reliability of the listener ratings; in quiet scenes, it approaches that ceiling. The model also responds sensibly to controlled gain and SNR manipulations. Together, the dataset and model provide a new way to predict listener-rated ease of speech understanding for real commercial hearing-aid recordings.

Since launching the our Blind Listening Challenge for #hearingaids, HearAdvisor has collected over 144,000 human ratings across more than 11,000 audio files—creating one of the largest perceptual datasets ever assembled for hearing aid evaluation. We’ve used this massive dataset to train a new deep-learning model (leveraging OpenAI’s Whisper) that can now predict human preference with incredible accuracy. The model significantly outperforms the industry-standard metrics we've used in the past, reaching a device-level correlation of 0.91 in loud scenes. We even have the data on which flagship products are currently coming out on top and which ones are struggling. Andrew Sabin is going to break all of this down—the data, the model, and what this means for the future of hearing aid transparency—this Friday morning. Join the talk: Session: 5.B (Featured Talk @ #VCCA2026) Date: Friday, June 26, 2026 Time: 9:10 AM – 9:25 AM CT (Chicago Time) If you’re interested in how we can use large-scale human data to make better recommendations for listeners, we’d love to see you there.