Product Management
Overview
Product Management Principles
Concepts From "The Lean Startup"
Prioritization
Thinking Big
Recommender Systems
80/20 Rule vs The Long Tail
Latency Analysis
Emerging Markets
Product Management Principles
Borrowed from "How Google Works"
(44) Focus on the user, and the money will follow. The primary objective of product teams is to create new, surprising, radically better products. Do this, and any smart company will figure out how to make money from it. Products that are highly valuable and differentiated on a scalable basis will generate great revenue.
[Put this first! PMs should be the voice of the user. What problem are we trying to solve for the user?]
(22) Know your competition, but don’t copy it. Copying competition leads to only incremental improvement. It can shift market share, but doesn’t grow the market. While you are busy fighting over share points, someone else will come in and create a new platform for the entire industry.
[Differentiate]
(43) Think 10X, not 10%. Global scale is available to just about everyone. But too many people are stuck in the old, limited mindset. Thinking big gives people much more freedom, since it pushes them to remove constraints and spurs ideas that were previously not considered. And it is a powerful tactic to attract and retain the very best people, who are usually drawn to the biggest challenges.
(47) Optimize for speed. Product cycles need to be quick: ship new products and features as quickly as possible, then gather market data to rapidly iterate and improve. When you optimize for speed things will go wrong, so tolerate some messiness and empower smart creatives to adjust for problems. When you fail, do it quickly and without stigma and salvage the technology and expertise gained in the effort.
(48) Fail well. Failure is permissible as long as the team was focusing on the user and thinking big. But it must be done well: fail quickly (be ruthless, don’t throw good money after bad), learn from your failure, morph any valuable assets into other projects, and don’t stigmatize the team that failed.
[aka Fail fast]
Concepts From "The Lean Startup"
But NOT just for startups!
Iterative Feedback Loop
Qualitative Metrics
Monthly Active Users
Revenue/Customer
Quantitative Metrics
Early stage
Interviews
Surveys
Later stage
App store ratings but take with a grain of salt
YouTube, blog & media reviews (e.g. Android Police)
Vanity Metrics
Make us feel good but don't help us make decisions (e.g. number of site visits)
Nigerian scammers are wise to ignore vanity metrics; their goal is to get the most gullible to respond so that the scammers don't have to deal with the "noise" of the non-gullible[*]
Cohort Analysis
Cohort: A group of users defined by something in common across the analysis period (e.g. signup month)
In the following example, Month 3 cohort remains engaged[*]
Revenue/User Example
1st month Rev./User is increasing, and rate of decline is decelerating,[*]
Testing
A/B - Changing one thing (e.g. color) and measuring the result (e.g. revenue); e.g. Google's 41 Shades of Blue
Multivariate - Changing several things at once to see which correlates with a result
Minimum Viable Product (MVP)
MVP - Least amount of features to validate a product assumption; For example:
Start a food truck before restaurant
Here's brilliant, "Battlestar Galactica" example:[*]
Product Phases
Prioritization
Ways to prioritize:
Bottom-line - How much will the project cost vs. how much revenue will it generate (or cost save)?
User-centric - Do what's best for the user. "Focus on the user & the money will follow."
Growth-centric - Prioritize based on growth potential. "Think 10X, not 10%"
etc.
Bottom-line Prioritization Example
Here is an example of a Bottom-line prioritization spreadsheet. It ranks projects based on "Business Value", "Project Size", and "Percent Complete".
Thinking Big
Think 10X, not 10%
How Google Works:[*]
The Internet provides all of the world’s info to everyone, mostly for free
Mobile devices & networks make global reach available to everyone
Cloud computing puts practically infinite computing power at everyone’s disposal
Thinking Big - Estimations
Here are some numbers to help with large scale projections (as of 2017):
Population
US: 300 million; Avg people per household: 3
World: 7 billion
Europe: 700 million [more than 2x US!]
Asia: 4 billion
Life Expectancy
US: 80 yrs
World: 65-70 yrs
Internet & Email
Internet users: 3 billion [almost half the world pop.]
Email users: 3 billion [almost every Internet user has an email account]
Gmail users: 1 billion [⅓ of Internet users]
Emails sent/day: 200 billion
Emails per user per day:
Biz: Received – 92; Sent – 32; growing[*]
Personal: Received – 20; Sent – 2; static
Avg email size: 75 KB
Thinking Big - Storage
Typically, largest cost is storage
Cost is minimized with Distributed File System like Hadoop Distributed File System (HDFS) or Google File System (GFS)
Makes use of commodity servers configured as a cluster, where each server contains inexpensive internal disk drives
HDDs (spinning) are better than SSDs (solid-state) for mass storage[*]
Storage Redundancy[*]
The following example shows the redundancy for a Cassandra distributed data store; example shows 6 copies!
Storage - Other Considerations
Need to store point-in-time backups (e.g. 500 GB requires 3900 GB or 8x of storage capacity)
Use "cold storage" for rarely-accessed data
Potentially leverage compression &/or deduplication
Recommender Systems
Recommender System - predict the "rating" or "preference" that a user would give to an item
Music Examples:
Types:
Content-based Filtering
Collaborative Filtering
Content-based Filtering
e.g. use song metadata to recommend additional songs with similar characteristics
Collaborative Filtering
Find users with similar preferences
Given similar users A & B, if A liked a song, B is likely to like that song too
NOTE: not as good as content-based filtering b/c users have very complicate preferences!
Enhance similarity analysis with:
Demographic data (e.g. age, gender, location)
Social graph... but how much do you really have in common with Aunt Suzy
Collaborative Filtering Example[*]
80/20 Rule vs The Long Tail
80/20[*]
20% of products account for 80% of the sales
a small number of bestsellers account for most of the business
when shelf space is limited, as in brick-n-mortars, you only want to carry the top 20%.
Long Tail Analysis[*]
Where shelf space is unlimited, many products with low sales volumes can accumulate sales that exceed the few 20% blockbusters with high sales volumes (many products w/ low sales > few products w/ high sales).
A significant portion of Amazon.com's sales come from obscure books that are not available in brick-n-mortars.
Latency Analysis
Latency - The time it takes to send & receive (round-trip) a packet of data (e.g. 20ms)
Can never be faster than the physical limitations of the network and hardware but can always be much worse
A tail latency, like P99 or P95, is used to represent the worst-case scenario and is often the single most important latency metric.[*]
e.g. P99 = 75ms: For the 99th percentile, the latency was 75ms. Therefore, 99% of the requests were faster than 75ms.[*]See more about Performance Metrics
Emerging Markets
Challenge – Poor cellular and wifi signals
Case Study - YouTube
Save video for offline; low res option to save disk space
Ability to preview video in thumbnail view
Ability to share via Bluetooth
India & China are promising markets b/c they have a lot of total Internet users but low Internet users per capita; however, has a challenging political climate
Data Sources
DMR (ExpandedRamblings.com)
Radicati.com
