An inside look at Mixpanel’s product journey | Vijay Iyengar (Head of Product)
Vijay Iyengar, Head of Product at Mixpanel, discusses Mixpanel's journey from multi-product expansion to refocusing on core analytics, sharing lessons on product strategy, prioritization, and common mistakes in setting up product analytics.
Deep Dive Analysis
13 Topic Outline
Unlearning 'No' as an Engineer in Product Leadership
Mixpanel's Product Journey: Expansion and Refocusing
The Dangers of Expanding Beyond Core Product
Mixpanel's Six-Month Product Planning Cycle
RICE Prioritization: When to Ignore Confidence and Effort
Estimation Challenges and 'Appetites Over Estimates' Approach
Keeping Product Teams Connected to Customers via Slack
Mixpanel's SaaS and Internal Data Stack Tools
Biggest Mistakes in Product Analytics Setup
The Rise of Data Warehouses and Event-Based Analytics
Career Growth Through Adopting a Product Mindset
Recommended Books and Podcasts
Favorite Interview Question and Thought Leaders
4 Key Concepts
Appetites Over Estimates
Instead of estimating how long a project will take, you define a fixed timebox (appetite) you're willing to invest in solving a problem. This forces teams to scope hammer the solution to fit the time, ensuring a complete, shippable solution within the allocated period rather than an open-ended estimate.
Event-Based Analytics
This approach models all customer interactions—whether with sales, marketing, or product—as events, capturing 'user did this action at this time.' It provides a granular, intuitive, and universal data model for understanding user behavior and asking complex questions about sequences of actions.
Product Mindset
This involves getting closer to customers, consuming raw customer context, and actively seeking opportunities to talk to them. Adopting this mindset, even for engineers, leads to a deeper understanding of problems and ultimately better products and services.
Theory of Constraints
This concept, described in 'The Goal,' focuses on identifying the single most limiting factor (constraint) in a system and then systematically improving or removing it to increase overall productivity and output.
7 Questions Answered
Engineers can try to sincerely make 'yes' work for new ideas for at least 10 minutes before defaulting to 'no.' This approach helps explore potential paths and avoids prematurely killing high-reach, high-impact directions.
A company should only expand to new ventures by investing profits, not by diverting people from the core product. If you're the leader in a core product, continue to out-invest competitors in that core to avoid disruption, and be wary of entering adjacent categories where you might only be an 'nth best' solution.
The core problem with estimation is being asked to estimate things before fully understanding what the 'thing' is, leading to inaccurate and often extended timelines. Estimates tend to be unreliable because they are produced too early in the discovery process.
Teams can pipe all customer feedback (from success, sales, Twitter, NPS, etc.) directly into a shared Slack channel and Notion database. This allows engineers and designers to consume raw feedback, react with an 'email emoji' to reach out directly to customers for more context, and build a culture of direct customer engagement.
The biggest mistake is setting up analytics using client-side SDKs (web and mobile tracking). This leads to poor data quality due to dropped events (web) and maintenance difficulties, duplicate events, and outdated tracking (mobile) because updates depend on client app versions.
It is recommended to track events from your servers instead of your clients. This ensures 100% cross-platform reach, allows for instant tracking updates for all users, and leverages engineers' existing familiarity with tracking logs, leading to higher quality data.
The data warehouse has become the center of gravity for all company data (product, marketing, sales), ensuring all teams operate from a single, consistent version of the truth. It serves as a loading dock for data, which can then be easily modeled as events and pushed out to various tools.
19 Actionable Insights
1. Invest Profits, Not People
If you lead in a core product, continuously out-invest competitors in that core, then use profits (not people or venture capital) to fund new ventures to avoid distracting your core team and leaving yourself vulnerable to disruption.
2. Try to Make Yes Work
As an engineer or product leader, instead of immediately saying “no” to new ideas, earnestly try to make “yes” work and document your efforts; this approach protects fragile ideas and leads to better outcomes.
3. Avoid Nth Best Products
Be cautious about expanding into new product categories where you’ll only be the “nth best,” as these bolt-on products often contribute little revenue while distracting resources from your core offering.
4. Optimize for Speed in Crisis
In highly competitive situations with critical missing features causing churn, prioritize speed to address table stakes problems quickly, focusing on immediate fixes over long-term holistic design.
5. Prioritize System Architecture Design
After addressing immediate feature gaps, prioritize a design-led initiative to define the product’s system architecture and core building blocks, ensuring consistency and maximizing the reach and discoverability of new features.
6. Grant Designers Strategic Breathing Room
To enable strategic design work, temporarily decouple designers from tactical, engineering-led projects, giving them dedicated time (e.g., three months) to focus on holistic product architecture without immediate feature demands.
7. Delay RICE Confidence/Effort
When using the RICE framework, initially ignore the “Confidence” and “Effort” scores for high-reach, high-impact ideas, allowing engineers and designers to explore solutions for a week before re-evaluating with C and E.
8. Use Appetites, Not Estimates
Instead of estimating project duration, set a timebox (appetite) as the input for planning, then explore how the solution would differ if given more or less time to find an efficient frontier of cost and impact.
9. Provide Raw Customer Feedback Access
Create a direct, ungated feed of raw customer feedback (e.g., in Slack) for all engineers and designers, empowering them to consume context, reach out to customers directly, and develop a product mindset.
10. Default to Server-Side Tracking
Avoid client-side SDK tracking for analytics due to poor data quality and maintenance issues; instead, default to tracking events from your servers for cross-platform consistency, environment control, and easier developer adoption.
11. Centralize Data in Warehouse
Utilize a data warehouse (e.g., Snowflake, BigQuery) as the central hub for all company data (product, marketing, sales) to ensure all teams operate from a single, consistent source of truth.
12. Model All Data as Events
Adopt events (time series of user actions) as the universal data model for all analytics, as every customer interaction can be granularly and intuitively modeled this way, enabling powerful sequence-based questions.
13. Cultivate Engineer Product Mindset
Encourage engineers to adopt a product mindset by getting closer to customers, consuming raw customer feedback, and actively engaging with them to foster innovation and build better products.
14. Six-Month Strategy & Bets
Plan on a six-month horizon, starting with a leadership strategy memo, then have teams develop “bets” (problem, solution hypothesis, plan to win, and success metrics) based on this strategy and customer context.
15. Leadership Joins Bet Ideation
Leaders should actively participate in team ideation and solution discovery during planning, contributing ideas and thoughts to foster high-bandwidth communication and ensure alignment on bets.
16. Linked Planning Artifacts
Document planning outcomes using linked artifacts: a Notion database for detailed bets (problem, evidence, impact, success, hypothesis, plan), a summary presentation, and an execution plan for sequencing and staffing.
17. Organize Teams by Paired Tensions
Organize cross-functional teams around long-lived, paired problems that have inherent tensions (e.g., power vs. simplicity), forcing the team to confront and resolve these trade-offs to create better solutions.
18. Enrich Customer Feedback Data
Continuously enrich customer feedback feeds with relevant account information (e.g., ARR, CSM) to provide engineers with necessary context, enabling them to make informed decisions before contacting customers.
19. Build Internal Tools with Reverse ETL
Leverage a data stack with reverse ETL tools (e.g., Census) to push modeled data from your data warehouse to other tools, enabling the creation of low-code internal tools and automating information flow to teams.
5 Key Quotes
If you're the leader in some core product, our takeaway here is you should continue to out-invest everyone else in that core and then invest the profits that come out of that core into the next venture. Like invest profits and not people.
Vijay Iyengar
Ideas are fragile in their infancy. And it's, you know, a hard no can really kill a whole direction that you could potentially go that could be very high reach and high impact.
Vijay Iyengar
You can't mow your lawn while your house is on fire. You kind of put out the fire and then deal with everything else.
Vijay Iyengar
There's not that many people that need the sixth best CDP or the eighth best feature flagging or the tenth best message targeting tool.
Vijay Iyengar
Engineers have been tracking events from servers forever. It's called logs, right? And events are just logs with a user ID in them.
Vijay Iyengar
4 Protocols
Mixpanel's Core Product Refocusing Strategy (2018)
Vijay Iyengar- Make the hard call to say 'no' to secondary product categories and focus the entire engineering team on the core product.
- Discard all existing planning and planned work.
- Collect all churn reasons painstakingly gathered by customer success and sales teams.
- Group churn reasons by category (roughly product features needed).
- Sort grouped reasons in descending order by Annual Recurring Revenue (ARR).
- Select the top 10 items to form the immediate roadmap.
- Grant every engineer direct access to customers and assign them a problem bucket to work on, optimizing for speed and clarity.
Mixpanel's Six-Month Product Planning Cycle
Vijay Iyengar- Leadership team writes a strategy memo outlining company direction for the next year and product's contribution, establishing key pillars.
- Teams combine the strategy memo with quantitative and qualitative customer context.
- Teams ideate and develop a series of 'bets' for the next six months, defining the problem, solution hypothesis, plan to win, and success metrics.
- Product leadership (Head of Product, Head of Design) actively participates in ideation and solution discovery with each team, contributing ideas and thoughts.
- Teams finalize bets, creating three linked artifacts: a Notion database page for each bet (problem, evidence, reach/impact, success metrics, hypothesis, rough plan), a summary presentation (one slide per bet), and an execution-focused plan for sequencing and staffing.
- Teams present their plans in a 'roadshow' to the rest of the company to gather feedback.
RICE Prioritization Adjustment for Innovation
Vijay Iyengar- Initially ignore the 'Confidence' (C) and 'Effort' (E) components of the RICE framework.
- Focus on 'Reach' (R) and 'Impact' (I) to identify potentially high-reach, high-impact, and innovative bets.
- Spend dedicated time (e.g., a week) with engineers and designers, committed to earnestly trying to solve these high-RI ideas.
- During this exploration, work to understand confidence and effort better, and potentially find higher-confidence, lower-effort ways to achieve the goals.
- Reintroduce the 'Confidence' and 'Effort' components.
- Apply the RICE framework as usual to achieve a balanced mix of innovative, incremental, and technical/product debt bets.
Server-Side Event Tracking for Analytics
Vijay Iyengar- Default to tracking events from your servers instead of client-side SDKs (web/mobile apps).
- Ensure all server-side logs include a user ID to transform them into trackable events.
- If necessary, supplement server-side tracking with client-side context for data only available on the client.
- Pipe all event data from servers into a data warehouse (e.g., BigQuery, Snowflake, Redshift).
- Use reverse ETL tools (e.g., Census) to push modeled and joined data from the data warehouse to analytics tools and other internal systems.