Playbooks and field guides from shipping eight products
How we run discovery, write specs, size markets, and decide what to build, with every claim traced to a case study on this site. Written by the human, checked against what actually shipped.
The page that never hit the cache: a Next.js lab you can run
A static page that quietly became dynamic. Rebuilt in a reproducible Next.js 15 lab with pinned versions, measured, and fixed three ways. Two of the three fixes, on their own, change nothing.
Copy the Homework: Using a Competitor's Decisions as Free Discovery
Starbucks pays for location research before every store. That research is visible to anyone who looks. How to read a competitor's decisions as discovery data, and the trap in copying them.
Product Discovery Interview Questions: The Script We Actually Use
The exact questions we ask in a discovery interview, in order, with the reasoning behind each one and the two we stopped asking.
Prototype vs MVP: Which One Should You Build First?
What a prototype is for, what an MVP is for, and the one question that tells you which to build first.
How to Write User Stories with Acceptance Criteria Engineers Can Build From
The story format we use, why each part is there, how MoSCoW keeps the list honest, and real examples from three shipped products.
How to Size a Market Before You Build (TAM, SAM, SOM Without Fooling Yourself)
Bottom-up numbers, footnoted assumptions, and the sanity checks that catch inflated figures. How PavIQ's $40B TAM was built and what it does and doesn't mean.
How to Run Product Discovery Before Building an MVP
Who to interview, what to ask, and the three questions that decide whether an MVP should be built at all. Examples from Dwell, GridSentinel and Squint.
Fractional Product Manager vs Agency vs Full-Time Hire: What a Startup Actually Needs
When to hire a fractional product manager, when to use an agency, and when a full-time hire is the only answer. Cost, speed, ownership and risk, compared honestly.
How Much Does an MVP Cost to Build in Canada? An Honest Breakdown
The price ranges by approach, the five decisions that move the number most, and how doing discovery first is the cheapest way to lower the bill.
Why ShiftFlow's break planner refuses to use AI
We built an AI scheduling tool, then deliberately kept its entire compliance layer away from the model. The rule I use to decide what AI should never touch.