What shipping to a wholesale distributor taught me that no tutorial did

Every tutorial teaches you to build the feature. Nothing teaches you what happens when the person using it has never used software like it before, has a business to run, and does not have time for your onboarding flow.

The gap between "works" and "usable"

MyLedger technically worked the first week it was live. It was also, in retrospect, built for someone who thinks like a developer — implicit assumptions about what a "session" is, what happens if you close the app mid-entry, what a sensible default should be. None of that matches how a busy distributor actually uses an app between customers walking in. The fixes that mattered most in the first month weren't features — they were removing steps, tightening defaults, and making the one or two actions someone does fifty times a day take one tap instead of three.

Trust is earned per feature, not once

Getting a real business to migrate off a paper notebook isn't a single decision — it's a hundred small ones. The owner didn't trust the automatic due tracking until they'd manually cross-checked it against the notebook for two weeks straight. That's not a failure of the feature; it's the correct level of caution for someone whose actual money is on the line. Software that wants to replace a trusted manual process has to earn that trust at the pace the user is willing to give it, not the pace the roadmap wants.

The question that mattered most: "who changed this?"

The first real support request wasn't a bug report. It was "why does this price look different from what I set" — a staff-access question, not a code question. That's when the activity log went from a nice-to-have to the most-used feature in the app. Any system with more than one person touching shared data needs an audit trail before it needs almost anything else on the feature list, and I'd learned that in the abstract before — it hits differently when it's a real owner asking a real question about real money.

What I'd tell myself before starting

Ship the smallest version to the real user faster than feels comfortable. Every week I spent polishing a feature in isolation was a week I wasn't learning what the actual failure modes would be once real data and a real, skeptical user were in the loop. The notebook it replaced had thirty years of institutional trust behind it. Software earns that the same way — by being right, consistently, in front of the person whose business depends on it.