Ads forecasting at Netflix and Spotify, how to build your personal moat - Jeff Li - The Data Scientist Show #069
Sep 14, 2023
auto_awesome
Jeff Li, a senior data scientist at Netflix and former data science manager at Spotify, discusses ads forecasting, career paths as a manager vs IC, and the culture differences at Spotify, Netflix, and Doordash. They also talk about the challenges of forecasting in finance and ads, detecting and accounting for seasonality and black swan events in advertising, transitioning from manager to senior data scientist, comparing company cultures, changes in tech stacks and data visualization tools, the future of forecasting, the importance of mentors in career growth, and the role of communication skills for data scientists.
Having experience in both advertising and forecasting is highly valued in the industry, helping individuals build a unique and valuable career profile.
Forecasting is applied in various industries, such as financial forecasting for revenue estimation and inventory optimization for accurate sales forecasting and shelf stocking adjustments.
Financial forecasting primarily focuses on revenue estimation and business valuation, while ads forecasting involves predicting ad inventory to optimize sales decisions and directly impact revenue.
Deep dives
Importance of Having Ads Experience and Forecasting Expertise
Having experience in both advertising and forecasting is highly valued in the industry. Netflix recognized the importance of a candidate with expertise in both areas, which helped the speaker secure their role. Finding the intersection of skill sets that are sought after in the industry can help individuals build a unique and valuable career profile.
Understanding the Use Cases of Forecasting
Forecasting is commonly applied in financial forecasting and inventory optimization. Financial forecasting involves estimating revenue and properly valuing a company for investors. Inventory optimization helps businesses accurately forecast sales and adjust shelf stocking in supermarkets. These are just a few examples of the versatile applications of forecasting across different industries.
Differences Between Financial and Ads Forecasting
The key distinction between financial and ads forecasting is the nature of the use cases. Financial forecasting primarily focuses on estimating revenue for valuation purposes and considering business initiatives. Ads forecasting involves predicting ad inventory to facilitate sales and operational decisions, directly impacting revenue. Both require accuracy but serve different purposes in the business context.
Challenges in Forecasting and the Importance of Interpretability
Financial forecasting faces challenges such as making accurate guesses about the future based on past trends and adjusting forecasts to meet stakeholder expectations. Ads forecasting carries risks of under-forecasting, leading to inventory shortage and revenue loss, or over-forecasting, leading to unfulfilled advertiser requests. Interpretability of forecasting models is crucial in understanding and explaining forecast outcomes, especially to gain stakeholder trust and ensure informed decision-making.
Scaling Impact Through Management
Being a manager allows for scaling impact across a team, as getting team members to perform well can have a 10X impact compared to doing individual contributor (IC) work. Each company culture differs in terms of intensity and work-life balance. At DoorDash, the focus was on quick execution and efficiency due to slim profit margins. Spotify offered a more relaxed work culture with good work-life balance and a collaborative environment. Netflix falls in between, with a unique culture that values informed decision-making and fast-paced work without excessive hours.
Tech Stacks and AI Tools
Tech stacks varied across the companies, including the use of Redshift, Snowflake, Google Cloud, Python, SQL, and others. When it comes to MLOps tools, Spotify and Netflix rely more on internally built platforms like Hendrix, while DoorDash uses a combination of internal and third-party tools for productionization systems. As for AI tools, chatbots like ChatGPT have been useful for debugging code, parsing error messages, and accessing command prompts. The future of forecasting leans toward transformer models, surpassing univariate time series models. Data collection and incorporation of external signals are pivotal for improving forecasting performance.
Jeff Li is a senior data scientist at Netflix, focusing on Ads forecast. Previously he was a data science manager at Spotify, worked on supply forecasting, demand forecasting, and data infrastructure. He studied business at the University of Southern California. We talked about Ads forecasting, career path as a manager vs IC, culture in Spotify vs Netflix vs Doordash.
Subscribe to Daliana's newsletter on www.dalianaliu.com for more on data science and career. Jeff Li’s LinkedIn: https://www.linkedin.com/in/lijeffrey/
Daliana's Twitter: https://twitter.com/DalianaLiu
Daliana's LinkedIn: https://www.linkedin.com/in/dalianaliu
(00:00:00) Introduction
(00:00:45) Got into data science from poker and consulting
(00:07:54) Ads forecasting at Netflix and Spotify
(00:13:30) From IC to manager to IC
(00:14:53) how to measure forecasting models
(00:21:58) collaborating with stakeholders in sales
(00:29:44) how he became an expert in ads forecasting
(00:49:57) impact sizing at Doordash
(00:57:34) Company culture differences (DoorDash, Spotify, Netflix)
(01:12:47) how he wants to grow his career
Get the Snipd podcast app
Unlock the knowledge in podcasts with the podcast player of the future.
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode
Save any moment
Hear something you like? Tap your headphones to save it with AI-generated key takeaways
Share & Export
Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode