A：Upstart's key achievements highlighted include being a stronger company with better technology, business, and teams; having AI leadership and precise macro handling; achieving 80% year-on-year growth in transaction volume and 71% revenue growth; growing consumer demand with over 30% increase in applications; managing risk models to respond to macroeconomic signals; maintaining profitability and a strong credit performance; new product offerings such as small dollar loans, auto, and home loans; continued innovation in processes and partnerships; strong capital market execution with new partner agreements; strong balance sheet with excess capacity and new logo additions; and issuing a securitization with strong demand and oversubscription.

A：Upstart's AI and credit models have made significant progress, particularly in responding with speed and precision to macroeconomic changes. The model's ability to adjust has been enhanced by a calibration methodology that is expected to reduce unwanted monthly volatility in model calibration. Further, new product offerings like auto retail and secured personal loans have been developed, showcasing the effectiveness of these models in different financial segments.

A：Upstart encountered challenges in Q3 with transaction volume being less than anticipated despite high consumer demand. This was addressed by risk models that reduced approvals and increased interest rates in response to macroeconomic signals, leading to a decrease in the conversion rate. The company believes this was a temporary 'speed bump' as there was no material deterioration in consumer credit strength and recent signs of improvement were observed.

A：Upstart is continuing to innovate in personal loans and customer acquisition by reducing end-to-end latency, launching a true machine learning model to optimize take rates, creating a framework that allows the use of their underwriting algorithms in partnership ecosystems, and developing a technique to target marketing spend based on causal impact for better return on investment. These innovations are designed to enhance the efficiency and effectiveness of the personal loan and customer acquisition processes.

A：Upstart has made advancements in the home equity loan process by using multimodal AI to expedite the review of documents typically required in home loans. This innovation is expected to lead to an industry-leading home equity product. Additionally, improvements in small dollar relief loans include instant funding capabilities, allowing most borrowers to see funds in their bank account shortly after approval. These developments are part of Upstart's ongoing efforts to refine and enhance its newer product offerings.

A：Third-party capital in the core unsecured lending segment remains readily accessible, outstripping the company's borrower supply and not impeding growth. Compression of spreads on third-party capital indicates a competitive funding environment and investor confidence in the company's credit performance.

A：The company prioritizes accurate credit performance, believing AI models better suited for a complex environment. Recent caution in the model, due to factors like rising repayment speeds and a slight drop in consumption growth, is seen as an indicator of imminent credit improvement and potential acceleration of growth prospects.

A：Q3 had total revenue of approximately $277 million, up 71% year over year and 8% sequentially. Revenue from fees was approximately $259 million, up year over year but short of internal expectations. Servicing revenue grew 10% sequentially. Net interest income was approximately $19 million, resulting from strong return performance. GAAP operating expenses were around $253 million, with variable costs related to borrower acquisition, verification, and servicing up sequentially.

A：The volume of loan transactions was approximately 428,000, up 128% from the prior year and representing approximately 75,000 new borrowers. The average loan size was approximately $6,670, down 12% from the prior quarter. The contribution margin, a non-GAAP metric, was 50% in Q3, down approximately 2 percentage points from the prior quarter.

A：The company is focusing on new product introductions and aims to put third-party capital arrangements in place to shift away from balance sheet funding and release invested capital. Progress on this front is being made, with the expectation of having multiple agreements in place for new product lines in 2026.

A：For Q4, the company expects total revenues of approximately $288 million, with revenue from fees of approximately $262 million and total net interest income of approximately $26 million. The contribution margin is expected to be approximately 53%, with GAAP net income of approximately $17 million and adjusted net income of approximately $52 million. Adjusted EBITDA is expected to be approximately $63 million. For the full year of 2025, total revenues are expected to be approximately $1.035 billion.

A：Application demand grew about 30% quarter on quarter, which was ahead of the origination transaction volume growth. This growth is attributed to successful marketing programs and cross-selling efforts. The increase in applications is viewed positively, despite being slightly conservative in the third quarter, as it highlights a robust business outlook even if the transaction volume didn't meet the highest expectations.

A：The impact of fraudulent activity on the market is considered to be limited and not widespread according to the speaker. While it has created some caution and increased diligence among banks and senior financing providers, there has been no major issue reported.

A：The company has been rigorous in building processes to effectively underwrite the dealership and mitigate risks associated with dealer activity. They have not seen any major issues despite the challenges highlighted in the market.

A：Originations in the super prime segment were down sequentially, and it is suggested that model tightness and macro signals are influencing this segment. The low to mid 700s FICO score range showed a relatively low UMI, and there is a model impact on this segment. The segment is also described as very competitive with price impact as a factor of competition.

A：The AI system has had some positive wins in customer acquisition by selecting people with a high propensity to apply and convert. The likelihood of converting through the funnel is mentioned as a factor, but the overall model impact and the choice to be more conservative in credit selection led to marketing to less likely approved individuals. Despite this, the model is still expected to influence Q4.

A：There has been very good credit performance in auto, and the models are working well. A potential transition or inflection point in the market may be positive for the company as it expands into the auto business. The company feels optimistic about the future of the auto business and believes that any disruption or noise in the market can present opportunities.

A：The improvements in UMI have materialized and were conservative. Even though they are subsiding, the model impact from Q3 will continue to affect Q4. As of the month of October, some amount of Q4 was impacted by the UMI rise.

A：Refinancing credit card debt continues to be the dominant use case for personal loans, but personal loans are also used for a wide variety of purposes. They are particularly useful for fast, easy, and competitive financing, which sometimes competes with secured loans for certain types of purchases.

A：Conversations with potential funding partners for the RD products are trending positively. There is a large deal appetite with good progress across various new product areas, although the timeline for these large, multibank deals is not perfectly predictable. There is no reported contraction in demand from private credit partners and discussions are ongoing regarding deal processes, legal processes, financing, and bank relationships.

A：The primary driver of the change in conversion rate is the conservatism in the model, which results in marginally fewer people being approved, higher rates of approved loan size, and an overall less responsive model.

A：Other factors influencing the conversion rate include the mix of applicants and macroeconomic conditions. The latter affects the financial health of borrowers and is a significant contributor to the overall conversion rate.

A：The model's response to the latest macro signals has been managed to reduce variance in conversion rates. Improvements were made to how the model reacts to changes in macro conditions, aiming for faster and more precise responses while minimizing unwanted variance caused by sampling and measurement error.

A：Faster loan repayment speeds are generally a positive sign of improving underlying consumer health and are adversely correlated with defaults over time. In the short term, however, faster repayments result in less interest earned to offset defaults, leading the model to become more conservative in pricing by increasing the coupon in loans.

A：Yes, the conversion rate changes have predominantly been related to the model's level of conservatism, with the most significant impact coming from changes in approvals.

A：The projected take rate for the HELOC product is expected to be roughly half the amount of the corporate average but on much larger loan sizes.

A：The primary performance metrics for the model are model separation and model calibration.

A：Model calibration is about the credit performance of the model. It has been exceptionally strong during the time period discussed.

A：The model's ability to approve or convert people weakened as a result of increased conservatism stemming from observations of elevated risk signals in various pockets of borrowers.

A：The model has evolved to have a more conservative approach in response to risk signals, which is seen as a feature rather than a bug. The model's ability to detect and respond to changes is considered strong.

A：The company is quite optimistic about the fourth quarter, with good growth rates, appropriate conservatism, a strong pipeline of model improvements expected to drive conversion rates up, and a belief in the overall strength of consumer health and the model.

A：The speaker considers the conservative approach of the model as a feature because it signifies the model's strength in making different decisions or taking a different stance on the market, leading to a belief in the model's overall health and resilience.

A：The UMI suggests that there is a segmentation in the economy where the sub-660 population, as measured through traditional credit scores, is in good shape with respect to default trends pre-Covid. Conversely, the 720 to 750 segment has elevated default rates compared to pre-Covid levels, indicating a U-shaped economy with different performance levels across various segments.

A：HELOC loans are general-purpose loans used for various reasons such as home improvement, other types of debt, retirement, or personal needs. The company views HELOC loans and personal loans as potential substitutes for each other.

A：The model has been intentionally built to respond faster than traditional credit metrics, providing real-time insights into credit performance by controlling for various borrower population changes and risk factors. This allows the model to detect precise and segment-specific patterns or broad trends that may not be visible with conventional metrics. The goal is to be faster and more precise than others in detecting market changes.