Getting into PM
PM side projects that interviewers believe
In this lesson8 sections
An unofficial lesson drawing on the methods in the Getting the Right Experience chapter of Cracking the PM Interview by Gayle Laakmann McDowell and Jackie Bavaro (2014). PracHub is not affiliated with the authors or publisher. Buy the book
For a career switcher with no product work, a side project is often the strongest evidence you can bring to a PM interview, and for a student it ranks just behind a real internship. Interviewers dig into it more than most lines on the page, so it works only when it fills a real gap in your record and you can answer every follow-up honestly.
We'll follow one candidate the whole way. Petra has spent five years as an operations analyst at a regional freight company, matching trucks to loads and patching schedules when drivers call in sick. She wants a PM job. Her resume proves she's good with data and scheduling. It says nothing about whether she has ever found a problem by talking to people and then decided what to build for them.
Pick the idea that proves what your resume can't
An interviewer reading Petra's resume has one doubt: has she ever made a product decision that nobody handed her? Her job can't settle that, however well she describes it. A project she started on her own can. Most of that proof sits around the product, not in it: who you talked to first, and which number showed it worked.
To find your version of that doubt, read three PM postings you'd apply for and underline every requirement your resume can't back up. Then ask of each idea what it would let you prove that your resume can't.
Petra weighed a route planner for small delivery firms against an app for school carpools. The route planner is a common trap for career switchers: it's her day job again, a strength the interviewer can already read on her resume, so it adds nothing new. Carpools would make her find users, make product calls and ship software, all evidence her resume lacks. The idea also had to survive twenty minutes of follow-ups, which is easier when you live with the problem. When an interviewer asks why she built it, the gap is her answer:
"My resume shows I can schedule trucks. It didn't show I could find a problem by talking to people and get a product used, and I wanted that evidence before applying for PM roles. Also, I sat in that pickup line every afternoon."
Students have other routes too, covered in Becoming a PM as a student, new grad or MBA.
Let twelve parents talk you out of the first idea
Before writing any code, Petra went to the people with the problem, because her last project skipped that step. A year earlier she set up a shared spreadsheet for neighbors in her building to lend each other tools. Six people joined, and nobody touched it after the second week: she never asked anyone whether they wanted to borrow tools. It still goes on her site, with one honest line on why it failed, because a failure you can explain shows you learn from evidence.
Her first idea this time was an app for taking turns driving among families who already carpool. Over four afternoons at the school gate she asked 12 parents to walk her through last week's pickups. Ask about a real recent week, not what people would like: people are poor at predicting what they'd use. Only 3 of the 12 carpooled, and they handled their turns fine in a group chat. The other 9 had a different problem: they didn't know which nearby families had a child in the same class with the same hours. Several said they'd only let their child ride with a family whose kid they knew. Also, 10 of the 12 had muted or deleted the school's app, yet all 12 had read that week's school text about a schedule change.
So the real problem was finding a family you'd trust, and turn-taking was already solved. Had Petra built first, she'd have shipped a tidy answer to something those parents had handled long ago. Plenty of side projects and startups end that way, and interviewers have seen it. Out loud:
"My first idea was a turn-taking app for existing carpools. Twelve conversations at pickup changed it: the parents who carpooled were fine, and the ones who didn't couldn't find a family they'd trust."
To practise choosing, take Describe a 0-to-1 project you led and pick the project you'd answer it with. Write the requirement it lets you back up and the first people you spoke to, then stop.
Let a paper card and 41 sticky notes set the rules
If software isn't the gap you need to fill, a design project can be the whole thing. Petra's first month was one. It came before the sign-up form, and it would have stood on its own, because it produced two rules with evidence behind each.
The class goes on every match, and the street address never does. Petra printed a half-page card showing a proposed match: the other parent's first name, their child's class, their neighborhood and the days they could drive. At a school evening event she showed it to 8 parents she'd never met. Six asked about the child's class before anything else. Nobody needed the other family's street address to decide, and several didn't want to share their own. Paper was the cheapest way to learn what a match must show.
The wait belongs to the school. Petra wrote each complaint she heard on a sticky note, 41 in all, and grouped them into piles: finding a family nearby, trusting the other driver, schedules that shift, and the wait itself. The wait was the biggest pile, and on those four afternoons at the gate she'd counted about 120 cars queued each day, the last ones waiting around 25 minutes. But only the school can change how the line runs, so she set it aside. Say why you dropped a pile, not just which one you kept.
That month lacked only usage, which the next two weeks supplied.
Two hand-run weeks, not a sign-up count
Would parents put their child in another family's car? That belief could sink the whole idea, and Petra had to choose evidence that would settle it. A sign-up costs nothing, so a sign-up rate is weak evidence. Parents giving and getting rides is strong evidence, so she ran the service by hand for two weeks before building anything, a method often called a concierge test.
She posted a sign-up form in the school's parent group chat, which has about 260 families, and 31 filled it in, about 12%. Quote the share of people who saw it, not the raw count. Every Sunday evening she matched them by hand in a spreadsheet, using rules from her interviews: same grade (same class, which parents asked for, left most families with nobody nearby), homes within about 2 km, overlapping days. Then she texted each family its match. Over two weeks, 22 families landed in 8 carpools, and 9 had nobody suitable nearby. She counted weekly active parents, meaning parents who gave or got at least one ride that week: 19 in the first week, 24 in the second.
So parents would hand over rides. By then every rule in the product had evidence behind it, which is what an interviewer wants when they ask about your choices:
"Texts instead of an app, because 10 of the 12 parents had muted or deleted the school's app, but all of them had read the school's text that week. Same grade, not same class: matching by class left most families with nobody nearby, so I widened it and show the class on every match. Neighborhood instead of street address, because nobody who saw the paper card needed it, and several didn't want to share theirs."
Build with AI, then check the script by hand
Only then did she build. Over three weekends she used an AI coding assistant to make a sign-up form, a Sunday-night matching script and texts that confirm each ride. She wrote the matching rules herself.
A working prototype is now within anyone's reach, so the app alone proves less than it used to. The project is often the one place an interviewer sees your judgment with nobody to share the credit, so expect the first follow-up to be what you did yourself:
"I ran the interviews, wrote the matching rules and did two weeks of matching by hand. An AI coding assistant wrote most of the code. I read the matching part line by line and tested it against my spreadsheet."
Her last sentence invites the next question: did the script get it right? She'd compared it with her week-two hand matches. It agreed on 7 of 8 carpools, and on the eighth she was the one who'd erred.
If a loop asks you to build something small live, our guide to the PM vibe coding interview covers that round.
Let the app's log find the Friday failure
Next comes whether anyone used it. Name your number and why it beats a flattering one:
"Weekly active parents: a parent who gave or got a ride that week. Sign-ups only measure curiosity. It went from 19 in the first hand-run week to 33 in the fourth week after the app launched."
A number that only rises invites the next question: what didn't work? Have a real answer with a number in it.
"Fridays. Monday to Thursday, about 1 ride in 10 was cancelled. On Fridays it was 21 of 52. The school lets out early on Fridays, a lot of parents work from home that day, and my form asked for one schedule for the whole week. During the hand-run weeks I blamed it on one family's sick kid. The app's log showed the pattern."
Once you've named a cause, the next question is what you'd change. Tie it to the failure and say what result would tell you it's fixed:
"Ask for each weekday separately, and text every family on Thursday evening to confirm Friday. I'd call it fixed when Friday cancellations fall to the weekday rate, and only then would I try a second school."
To practise the whole chain, take Pick your favorite past project and walk through your role, decisions and results with your own project. Answer in two minutes, then have a friend ask each follow-up from your last answer.
Some interviewers turn the what-next question into a full design round. Practise with Product Design From Your Own Resume: Turn a Past Project Into a New Product Proposal, out loud, using the project you just prepared.
Put 47 families on the resume, not hundreds
Give the project its own entry on your resume and LinkedIn profile, plus a one-page write-up on a personal site or portfolio page you can link to. Write its bullets the way you'd write bullets for a job: the result first, the method in a few words, one line each where you can. Four weeks after launch, 47 families had signed up. "Used by hundreds" would fall apart at the first follow-up, while "47 families at one school" holds up. Petra's entry:
Carpool matching for one elementary school (side project; text-message app built with an AI coding assistant)
Reached 33 weekly active parents from 47 families four weeks after launch; first matched 22 by hand
Found Friday rides cancelled four times as often as other days, traced to a one-schedule-per-week form
The write-up holds what doesn't fit on a resume: the idea that died after twelve conversations, the paper card, the Friday problem and the next step. For the rest of the page, see PM resume for students and career switchers.
Name the belief that would sink the next idea
For the carpool project, the belief that could sink it was whether parents would put their child in another family's car. Start your next project the same way: test that belief before you build. To practise choosing what to test first, try Amazon Product Launch: Choose an MVP to Test a Home Improvement Services Business: write that belief for the business and the cheapest test of it, then stop.