committing before you understand

a program is 9 months late. the team is good, and the budget was real. when we get called into a program like that, the first thing we do is not look at the schedule. we ask one question: who is this for, and what changes for them when it works?

5 people give 5 slightly different answers. that is the problem, and it is older than the slipped date. months earlier, someone committed to building a thing before anyone had agreed on who it was for. everything since has been built on top of that guess.

the early call nobody remembers making

the decisions that sink a program are made in its first weeks, quickly, and on very little evidence. a feature list gets written from what the sales team heard. a timeline gets set from what the board wants to hear. a vendor gets picked because a demo looked good. none of these feel like big decisions at the time. by the time they show up as a late delivery, half a year of work is sitting on them.

outside help can make it worse. people who buy consulting tell us, in pharma especially, that they walked a firm through a deck about their problem and got the same deck back a few weeks later as the diagnosis, with an invoice. nobody in that room had met the person the product is for.

in a large company that mistake shows up as a program that is 9 months late. in a startup it is fatal, which is why it gets counted there. cb insights read the post-mortems of 431 venture-backed companies that shut down from 2023 onward. most ran out of money, which cb calls the way the story ends rather than the cause. the cause it found most often: 43 percent had built something the market did not need¹.

4 things to ask for before you commit

you do not have to run the build to protect yourself from this. you have to ask for 4 things, in order, and refuse to commit until you have them.

1. the sentence. before anything is built, the team writes one sentence: who this is for, and what changes for them when it works. if the team cannot agree on the sentence, the program is not ready, whatever the timeline says. a team that starts without it spends its daily standup solving 5 different problems at once.

2. the fake. a clickable prototype, a fake of the product you can tap through on a phone, takes a few days to make and a week to put in front of 5 real users. a full build answers the same question months later, after the money is spent. teams skip this because a fake does not look like progress to whoever is watching. ask to see it anyway. learning that an assumption was wrong before engineers start is the most productive week a program can have.

3. the one feature. pendo, a company that measures how software gets used, studied 615 software products in 2019. about 12 percent of features drove 80 percent of daily use, and roughly 80 percent of features were rarely or never touched². that is why a first release is called minimal. the point of a minimum viable product is to win adoption on the one feature that carries the value, and to learn from real use before anything else gets built.

4. a small first audience. ask that whatever ships goes live for a small group first, so a mistake reaches 50 people instead of 50,000, and that the team can change it cheaply once it is out. understanding the user keeps happening after launch that way.

the person at the other end

go back to the sentence. the person most often left out of it cannot see the screen well, or cannot use a mouse, or reads with software that speaks the page aloud. in a patient-facing program that person is not an edge case. webaim, a nonprofit, scans the top million home pages every year. its 2026 report found accessibility failures on 95.9 percent of them, worse than the year before³, in the year producing a web page became easier than it has ever been. a team that has written the sentence knows who that person is. a team that skipped it does not know they exist.

so before your next program commits, ask for the sentence, the fake, the one feature, and the small first audience. if the team cannot produce the first 3 in 2 weeks, the program is not ready to commit, and finding that out now costs 2 weeks rather than 9 months.