The PM's job is deciding what not to build

The most effective product work I've done was subtraction. When we rebuilt merchant onboarding for small businesses at Juspay, the breakthrough wasn't a feature. It was a list of things we refused to ship.

The original onboarding flow asked merchants for everything a payments company might conceivably need: configurations, credentials, preferences, priorities. Each field had a constituency inside the company. Each field was, individually, defensible. Together they were a wall, and small merchants, who don't have an integrations team, bounced off it.

Defaults are decisions you make so customers don't have to

The uncomfortable realization was that for most merchants, there is a correct answer to almost every configuration question, and we knew what it was. Asking anyway wasn't flexibility; it was cowardice, exporting our indecision to the customer as a form field.

So the work became: for every input, either derive it, default it, or defend it in a meeting. Deriving and defaulting went into a simplified API and an automated checklist that walked a merchant through only what genuinely required them. Defending was the political work, because every removed field had an internal owner who could describe the one enterprise customer who once needed it. The phrase that won those meetings, eventually, was: "They can still change it later. They just don't have to know about it now."

One metric, chosen for honesty

We made onboarding friction the metric: the time and steps between "merchant decides to integrate" and "first successful real transaction." It's a brutal metric because you can't fake it with a redesign. Progressive disclosure that hides the same forty questions behind an accordion doesn't move it. Only genuine subtraction moves it.

It moved. SMB adoption grew tenfold, and the automation freed about a fifth of the engineering resources that had been absorbed by hand-holding onboardings. A good product metric is one that fails to flatter you. Most dashboards are mirrors; this one was a scale.

What I'd tell a new PM

Your backlog will always be a list of additions, because additions have advocates and subtractions don't. Customers ask for features in words; they ask for simplicity only in churn data, which arrives late and unsigned.

So the discipline has to be yours. For everything in the spec, ask: who absorbs the complexity if we ship this, us or the customer? The companies that serve small businesses well are the ones that consistently answer "us." That's the entire trick, and it's hard precisely because it produces nothing to demo at the all-hands.