3 types of product improvements

The main role of a product manager to identify and prioritize product investments within your area that'd have the most impact on the overall business goals (aka roadmap). 

Usually, you don't make just one investment. You make a portfolio of bets. You can place bets across multiple core pillars or themes as I have suggested in a previous post on product strategy

It's also helpful to assess your portfolio mix across the type of product improvements: 

1. Ah, finally improvements (bugs, annoyances) 

These are fixes for obvious annoyances or broken parts of the experience. Users know it and product teams usually know it. These experiences can cause casual users to churn immediately and fans to churn eventually. 

Regularly identifying and fixing these before they snowball is a good defense and good for building trust, pride, and quality (reduce broken windows!). 

2. Yeah, that's better improvements (iterative improvements)

Products aren't perfect out of the gate (and if it's perfect, you probably launched too late :)). You need to iterate and keep improving them over time. Initially, these iterations will produce a lot of big wins. But eventually, the yields decline and that's a sign to invest elsewhere. 

3. Oh Wow! improvements (big bets) 

These are step-level changes - completely new products or experiences like when Apple launched the iPhone, or significant improvements to an existing experience like when Google launched instant autocomplete search suggestions

As you can imagine, these are the trickiest kind of investments. They are high-risk and high investment bets. But they are also essential (nothing ventured, nothing gained). As Jeff Bezos puts it, "If the size of your failures isn't growing, you're not going to be inventing at a size that can actually move the needle." 

Depending on the size of your team, you should have one or more big bets brewing at any point in time. But be careful not to spread yourself too thin or rushing in before thoughtful strategizing and validation.