Fit and behavioral questions
How to build a PM behavioral story bank
In this lesson5 sections
An unofficial lesson drawing on the methods in the Behavioral Questions 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
A PM behavioral story bank is a set of true stories from your own work, prepared so that any "tell me about a time" prompt finds a strong answer in seconds. This lesson builds one from a résumé and fixes the thin stories.
We'll follow Maren. She has been a PM for two years at Quarrybill, an invented company that sends invoices for about 9,000 small wholesalers and chases the shops that owe them money. Before that she spent three years as an operations analyst at a freight broker and two as a support team lead at a payroll software company. On weekends she runs fixtures for an amateur football league. She has three PM loops coming up, one of them at Amazon.
For telling one story well, see the STAR method lesson and how PM behavioral answers are scored and delivered. The general story bank lesson covers every role; this one is for PM loops.
Let the Amazon sheet pick the flagships
Maren starts with Amazon, because it publishes the values its interviewers score. Each interviewer there probes different Leadership Principles and compares notes afterwards, so a story told twice tends to earn credit once. She makes a sheet that gives every principle at least one story, and the principles she expects most a backup too. Its rows for those principles:
| Leadership Principle | Best story | Backup |
|---|---|---|
| Customer Obsession | Payment-status feed | Support first reply time |
| Dive Deep | Cutting a payment reminder | Peak-season forecast miss |
| Have Backbone; Disagree and Commit | Sales and the one-off template | Dispatch lead |
| Ownership | Late-fee setting | Wrong tax-form article |
| Deliver Results | Moving the acquired suppliers | Carrier scorecard pilot |
| Bias for Action | First segment with no usage data | Tax-deadline week |
No story repeats, so each interviewer can hear something new. A loop with several behavioral interviewers calls for 8 to 12 prepared stories, four to six of them flagships. The sheet sets Maren's count: she expects interviewers to probe these six principles most, so the best story in each row becomes a flagship. She'll know them cold and steer toward them whenever a prompt gives her room. Her other two loops draw on the same bank, and behavioral questions there often turn up in more than one round.
The Amazon PM interview guide covers how the loop runs, and our Googleyness guide covers Google's version of this round.
To practise this step, try Amazon Leadership Principles – Behavioral Deep Dive. Assign a story from your bank to each prompt, mark every forced repeat, and fill those gaps first.
Put the scorecard's number up front
Maren tried the carrier scorecard pilot, her Deliver Results backup, on a former colleague, then asked for the situation, her action and the result back in their own words. The retelling was "You built a spreadsheet and the VP liked it." The pilot design and the drop in late loads, from 14% to 9% in one region, never registered, so she moved both to the front. Try every story you'll lean on this way, on a friend or a partner from a peer mock interview site. If their version loses your action or the number, refine the story or replace it.
Record yourself too. Many first rounds are video calls or one-way recorded screens, and a recording shows filler you can't hear live. An AI mock interviewer can draw out follow-ups you didn't expect, but keep your own words and numbers. Polished phrasing sounds like everyone else's.
Log what each flagship lost in practice
The scorecard's gap took a listener to find. Maren puts every flagship through a retelling and a recording, then logs what turned up beside the row each story has to win:
| Flagship | Row it must win | Lost in the retelling or recording | Survives "how was that built?" |
|---|---|---|---|
| Payment-status feed | Customer Obsession | nothing | yes |
| Cutting a payment reminder | Dive Deep | The stakes, her call and any number | yes |
| Sales and the one-off template | Have Backbone | nothing | yes |
| Late-fee setting | Ownership | Which parts were hers: "we" 11 times, "I" twice | Only if she claims the spec |
| Moving the acquired suppliers | Deliver Results | nothing | yes |
| First segment with no usage data | Bias for Action | nothing | yes |
The row is what that interviewer grades, so a story that shows nothing about its row is filler, however well it answers the prompt. You can run the last column's check without a listener: one follow-up about how something was built is all an interviewer needs to expose a claim you borrowed.
To practise this step, try Behavioral Stories: Technical Depth, Data Experimentation, Failure. Map each of its parts to one of your flagships, then have a friend retell each story and ask how one piece was built.
Fix the two flagships the log flagged
Her late-fee recording, the one with "we" eleven times, raises a question. If every "we" turns into "I", is enough left that is hers? If not, she tells its backup, the wrong tax-form article, instead. The recording began:
"We shipped a setting that let suppliers add a late fee, and we didn't preview how it would look on invoices already sent."
Enough is left, because the spec was hers, and so was the missing preview. Reworded:
"I wrote the spec for the late-fee setting, and I left out a preview of how the fee would appear on the shop's invoice."
Then she runs the build follow-up on it. Asked how the setting was built, she can walk through her spec and credit the engineers for the code. She claims only what she did, so the question has nothing to expose.
Her recording of the reminder story:
"Some suppliers weren't happy with our payment reminders, so I recommended a shorter sequence. The team launched it and late payments went down."
It gives a listener nothing to keep. "Weren't happy" gives no stakes, "went down" gives no number, and "recommended" leaves the decision with someone else.
Before you drop a story like this, ask what the short version hides, such as an analysis you ran or a call you made and defended. Maren finds both: she grouped the invoices herself, and she made the call on which reminder to cut. Rebuilt:
"One in eight of our suppliers had switched payment reminders off, and on-time payment had slid from 66% to 61%. I grouped nine months of invoices by the last reminder a shop opened before paying. The earliest one, ten days before the due date, came before only 1 in 20 on-time payments but drew most of the complaints shops sent their suppliers. I decided to cut the default from three reminders to two. Sales worried payments would slow, so I first tested it on half our suppliers for six weeks. Their invoices were paid on time 64% of the time against 61%, half as many switched reminders off, and I rolled it out to everyone."
The stakes now come first, the grouping leads to a call that is plainly hers, and the half-of-suppliers test gives the number.
Then she writes the story's core in one line: the problem, her move, the number and what she learned. Many current rubrics expect a closing lesson, so decide yours now rather than in the room. Hers reads: "One reminder annoyed shops so much that suppliers were switching all of them off. I found it and cut it after a test. On-time payment 61% to 64%. Taking something away can beat adding something."
With the core fixed, the length becomes a choice: 30 seconds on a recruiter screen or two minutes at an onsite. The core also tells her where to open. Asked about a conflict, she starts at the sales objection. Asked about using data, she starts at the grouping.
To practise this step, try Innovation, Root Cause, and Deadline Management Stories. For each part, write the stakes, your own call and the number before anything else, and rebuild or replace any story missing one.
Find a second ambiguity story
Her practice prompts ask about ambiguity, and Maren hoped her rate-quote model could back up the first-segment story there. To follow it, a listener needs spot and contract freight rates, empty return miles and why three months of data skews a lane. Four sentences of freight economics in, she still hadn't reached her own decision. When a listener can't reach your part without a long stretch of background, retire the story. Their attention goes on the background, and they lose track of you.
That leaves her one ambiguity story. To find another, she lays her résumé against the prompts she expects, taking the columns from her practice prompts and the Amazon sheet:
| Impact | Failure | Conflict | Ambiguity | Deadline | Customer | |
|---|---|---|---|---|---|---|
| Quarrybill, PM | Cutting a payment reminder | Late-fee setting with no preview | Sales and a one-off template | First segment with no usage data | Moving an acquired rival's suppliers | Payment-status feed |
| Freight broker, analyst | Carrier scorecard pilot | Peak-season forecast 20% low | Dispatch lead | Rate-quote model (retired) | ||
| Payroll software, support lead | Wrong tax-form article | Tax-deadline week | First reply time from 9 hours to 2 | |||
| Football league, volunteer | A season with no pitch clashes | Two clubs, one pitch |
Blanks in older rows are normal. Interviewers ask most about your latest role, so a gap there is worth an evening with old docs and tickets. Maren's Quarrybill row is full, but the ambiguity column now has only one live story, so she goes back to her Quarrybill planning docs for a second.
Kept current, the grid also answers narrow prompts. Asked about her hardest disagreement at the freight broker, she looks at one cell and finds the dispatch lead.
One lesson in this course covers what failure, influence and teamwork answers need. "Why this company?" needs no story from the grid; prepare it separately.
To practise this step, try Describe impact, ambiguity, and conflict. Fill in your own grid, then pull a different story for each part of the prompt.