Data Product Manager
1. Applying Data Science to Product Management
- Introduction to Data Product Management
- Granularity, Distribution, and Modeling Data
- Trends, Enrichment, & Visualization
- Project - Flying Taxi Business Case Part 1
- Setting Product Objectives & Strategy
- Proposal Synthesis & Design
- Project - Flying Taxi Business Case Final
2. Establishing Data Infrastructure
- Introduction to Data Infrastructure
- Introduction to Data Pipeline
- Data Consumers
- Data Producers
- Data Strategy
- Project - Build a Scalable Data Strategy
3. Leveraging Data in Iterative Product Design
- Introduction to Iterative Design
- Choose & Measure KPIs for Live Products
- Evaluate User Conversion Funnels
- Apply User Segment Analysis
- Quantitative and Qualitative Data
- A/B Test and Multivariate Test
- Project - Create an Iterative Design Path
1.1 Intro to Data Product Management:
Product Management - uncover user problems, leverage data/user research, set objectives and metrics to align cross functional teams, lead engineers/designs to build right solutions, execute and launch product, represents customer and understand risks and data possibilities.
Types of Data Product Managers: build data products for external users, build data platforms for internal users, leverage data science for PM decisions.
Apply data science to product management: size and nature of company affects data scientist analysis availability, do it yourself and be strategic with limited resources, self-service product analytics, have pros focus on deep statistical analyses and ML.
Use Case: PM for app, users aggregate data in dashboard, but manually input some data instead of using data source integration, investigate manual input data, if most frequent data is not with a partner, integrate with said partner.
Business Stakeholders: data science (determine success metrics, stat analyses, build ML models), design (physical/digital interactions), product marketing (determine pricing, market position, asses market, establish brand, estimate addressable market, rollout stat, collateral comms), user research (recruit/interview customer segments for product viability and completeness), data engineering (collect, transform, store batch and stream data into desired output & location for analyses), client engineering (Web, Android, iOS, build intuitive interfaces that display data for end user, and collect event data), server engineering (backend engineers building databases, stores, and caching for frontend CRUD, get save client data and send to data warehouse), quality assurance (break product, expose bugs, you should leverage analytics on issue priority), localization (dynamic text translation for international adoption), sales success support (evangelize limitations and capabilities of product, provide roadmap, engineering feature exclusion reasons), finance & business ops (measure unit economics, business growth, financial health metrics, quantify opportunities for regional/global expansion), legal privacy compliance (determine data processing regulations requirements and ensure data products are compliant with consumer protection laws).
Product Management Fundamentals: understand user pain, use data to understand macro/micro-trends, create solutions with eng and design, prioritize features/bugs, build roadmap, write specs, create marketing material. Product managers and different from project management and don't do technical mentorship, technical architecture, write code, operational management.
Course overview: understand granularity, distribution and data modelling, analyze trends for enrichment and visualizations, measure formulate implement product objectives and KPIs, proposal synthesis & design.
1.2 Granularity, Distribution, Modeling Data
Granularity: what is being measure in the dataset (each row)
Distribution: how spread apart is the data, where the center of the data is for weighted numerical and ruw number of data points.
Modeling: how is the data spread across multiple tables, and how the data aggregated for query efficiency.
Big Picture: Don't take data at face value, tables can come pre-filtered, understand what is relevant.
Granularity: What is being measured? Size matters, may have extra data, may be too much for visualization. Need primary key to join with other datasets and select count(distinct(*)) should mach select count(*) or use describe tableName. Primary keys can be composite (two or more joined column values).
Data Distribution: to understand price elasticity, outliers as edge cases, identifying segments, mean avg, median (think median (middle) lane), standard deviation (how wide data is, 1std 68%, 2std 95%, 3std 99.7%), long tail distribution (large cluster of data far from median).
The course uses Tableau but I'm using Data Studio. I used a filter to show only product_id 2 reviews and created a metric to show standard deviation STDDEV(review) of the review as a Calculated Field.
The more data the more unlikely you are to query all of it (time and money cost), use aggregations to stored in the DB as a dimensional table which stores metadata about different slices of data in the fact table, while joining them by a common key.
Ex. total_sales | product_id | sale_month |