WEEK 2 MILESTONE: ROUND 2 INTERVIEWS + PATTERNS
Team 3: Olivia Chiang, Krisha Singhani, Jaycee Wang, Li Ji
NORMS
- Communicate about deadlines, restrictions, busy weeks etc, and ask for help as needed
- Be direct and solve conflicts before they get worse
- Prioritize and ensure that we are aligned about common goals
- Low ego/don’t take criticism personally
- No judgement/assume the best from others
- If you say “no” to someone’s idea, propose a counterproposal
NEW QUESTIONS
- What do relationships look like in your role?
- What does coordination look like in your role? What does working with eng and design look like day to day?
- Do you feel fulfilled in your role? Why or why not?
- What does job security look like in your role? What core skills do PMs need now?
- What impact has or will AI have on your role? How has AI changed your own work in the last year? Which part of the PM role do you think AI affects most?
- How do you figure out what users actually need when things change so fast?
We clustered the struggles we identified from the first round of cold interviews into five themes, and wrote a new interview question for each theme. All second round interviewees received the new questions. Second round interviewees that were not interviewed in the first round also received the three standard prompts (average, good, and bad day).
| Theme | Struggles | New Question |
|---|---|---|
| Relationships | Huge component of the role; heavily dependent on each other; spend a lot of time talking | What do relationships look like in your role? |
| Coordination | Information arrives too late/ changes after work starts; shifting priorities; convincing engineers | What does coordination look like in your role? |
| Fulfillment | Escaping grunt work; finding fulfilling work in a structured job | Do you feel fulfilled in your role? Why or why not? |
| Job Security | Whether PM jobs will shrink or change | What does job security look like in your role? |
| AI | AI slop; new responsibilities; quality suffering | What impact has AI had, or will it have, on your role? |
Torres would call our new questions too broad (“What does X look like?”). She would suggest for us to use follow-ups such as “When did that last happen?” or “What do you do in that case?” to encourage stories over just opinions.
INTERVIEW DETAILS
| Code | Interviewee | Interviewer | Round | Transcript/ Notes |
|---|---|---|---|---|
| P1 | (Transcript Missing) | Mateo | 1 | (Transcript Missing) |
| P2 | Design and AI Consultant | Jaycee | 1 | Transcript |
| P3 | Design PM @ HUEGRID | Olivia | 1 | Transcript + Notes |
| P4 | (Transcript Missing) | Mikey | 1 | (Transcript Missing) |
| P5 | PM Intern @ Roblox | Olivia | 2 | Transcript + Notes |
| P6 | PM @ Walmart | Krisha | 1 | Transcript + Notes |
| P7 | PM @ European AI Startup | Jaycee | 2 | Transcript |
| P8 | PM Intern, Chinese Big Tech | Jaycee | 2 | Transcript |
| P9 | PM Intern @ Startup and Chinese Big Tech | Jaycee | 2 | Transcript |
| P10 | PM @ Walmart | Krisha | 2 | Notes |
| P11 | PM-SWE Hybrid @ Google | Krisha | 2 | Transcript + Notes |
| P12 | PM Intern @ Databricks | Olivia | 2 | Transcript + Notes |
| P13 | AI Strategy PM Intern @ TikTok | Li Ji | 2 | Transcript (Translated) |
| P14 | PM Intern @ Minimax | Li Ji | 2 | Notes |
We are unsure if Mateo and Mikey are still enrolled in this class. We have select quotes and themes from their Round 1 interviews, but have been unable to find any further interview content.
AFFINITY MAP
AI COMPARISON
What did it catch that you missed?
Claude identified a few patterns that we didn’t: 1. Checking whether AI output is right is slow (six interviewees: P3-8 and P12). 2. Keeping up with how fast AI changes is difficult (three interviewees: P2, P4, and P8). 3. Context gathering eats most of the job (three interviewees: P1, P3, P8).
What did it flatten into something generic?
Claude lumped the issues of a two-week review wait (P6) and skills gaps (P3) under “work stalls.” Claude also grouped the issues of Claude making up a role (P3) and AI designs being too generic (P2) under “AI output is generic or made up.” These groupings erased some of the specificity identified in the interviews.
Did it assert anything that isn’t in the transcripts?
Claude avoided hallucinations entirely. Every quote matched the source word for word.
LONG LIST
| # | Problem | People | Did/ Said | Quote |
| 1 | Depending on other teams’ process or capacity | 6 (P3, P4, P5, P8, P9, P12) | Did | “If the review was yesterday and I finish a new request today, I have to wait two weeks.” (P9) |
| 2 | Convincing other teams without authority | 6 (P3, P4, P5, P8, P11, P12) | Did | “The hard part wasn’t the technical problem. It was that I had no authority over either team.” (P4) |
| 3 | More responsibilities with AI, but no more time | 3 (P2, P4, P7) | Said | “Now I’m expected to do much more in eight hours. So the bar just keeps rising.” (P7) |
| 4 | Information arrives late, or priorities change after starting | 5 (P1, P4, P5, P11, P13) | Did | “The hardest days are when I don’t get enough information to get a project started, or when I receive information that later ends up changing.” (P1) |
| 5 | Conflicting priorities and too many stakeholders | 5 (P3, P4, P6/P10, P11, P12) | Mixed | “Too many cooks in the kitchen” (P6/P10, describing a recent bad day) |
| 6 | Stuck with grunt work | 3 (P3, P9, P13) | Did | “What I didn’t like was doing only grunt work, because I was working at an agency at that point.” (P3) |
| 7 | AI output that’s generic or made up | 2 (P2, P3) | Did | “Claude was gaslighting us on a Google Meet call.” (P3) |
| 8 | Checking whether AI output is right | 7 (P3, P4, P7, P8, P9, P12, P13) | Mixed | “Having AI write the ideal answer for an AI is kind of dumb. A person still has to check every single one.” (P8) |
| 9 | Keeping up with how fast AI changes | 3 (P2, P4, P7) | Mixed | “When a new model’s weights drop, my whole week reshuffles.” (P4) |
| 10 | Context gathering eats most of the job | 3 (P3, P7, P12) | Mixed | “Context gathering is 80% of my job.” (P3) |
TAM
TAM = number of PMs x what gets spent each year on tools for them
Number of PMs Calculation
| Estimate | Scope | Source |
|---|---|---|
| ~850,000 | Worldwide, LinkedIn Recruiter data | Shcheglov, “How many product managers are there?” (Corporate Waters) |
| ~200,000 | United States (about 47% of the active pool in the top hubs) | Shcheglov, “How many product managers are there?” (Corporate Waters) |
| ~1,060,000 | Worldwide, LinkedIn search (2023) | Yanez, “How many Product Managers are in the World? Updated!” (LinkedIn) |
| 1,000,000+ | Worldwide, LinkedIn (2021) | Product Focus, “Product Managers across the world” |
We estimate anywhere between 850,000 to 1,060,000 global active PMs.
Single PM’s Tool Stack Cost
| Tool Category | Example (Price/ User/ Month) | Source |
|---|---|---|
| Work Tracking | Jira Standard, $7.91; Linear Business, $16 (annual) | Atlassian and Linear pricing pages |
| Roadmaps and Feedback | Productboard, $19-59 per maker; Aha! Roadmaps, $59-149 (annual) | Productboard and Aha! pricing pages |
| Docs and Wiki | Notion Business, $20 (annual) | Notion pricing page |
| Design and Prototyping | Figma Professional full seat, $16 (annual) | Figma pricing page |
| AI Assistant | ChatGPT Business, $20 (annual) | OpenAI Help Center |
We estimate a lean PM tool stack (Jira + Productboard Plus + Notion + ChatGPT) to cost ~$67 per month, or $800 per year.
We estimate a fuller PM Tool stack (Linear + Aha! Premium + Notion + Figma + ChatGPT) to cost ~$131 per month, or ~$1,570 per year.
Total TAM
| Low Stack (850K PMs x $800) | High (1.06M PMs x $1,570) | |
|---|---|---|
| Worldwide TAM | ~$0.68 Billion/ Year | ~$1.67 Billion/ Year |
| US only (200K PMs) | ~$0.16 Billion/ Year | ~$0.31 Billion/ Year |
We checked our TAM calculation against Spherical Insights, which estimates the global product management software market to be ~$8.38 billion in 2025. That is several times our calculation. We suspect the gap to arise from tools bought for the whole product and engineering organizations, not just PMs. Other paid reports also disagree with each other’s estimates by several times or decimal points.
Some limitations we face in calculating total TAM including the following. First, list prices tend to overstate what large companies pay after discounts, and understate add-ons and AI credits. Second, some tools (like Jira) are mostly bought for engineers, so it may distort the TAM with respect to PMs. Additionally, the stack excludes analytics and user research tools, which interviewees described as both valuable and expensive. For example, P3 mentioned that the research tool Dovetail can cost “at least $40,000 per year.”
Sources:
- Shcheglov, M. “How many product managers are there?” Corporate Waters. https://corpwaters.substack.com/p/how-many-product-managers-are-there
- Yanez, P. “How many Product Managers are in the World? Updated!” LinkedIn. https://www.linkedin.com/pulse/how-many-product-managers-world-pau-yanez
- Product Focus. “Product Managers across the world.” https://www.productfocus.com/product-management-resources/infographics/product-managers-across-the-world/
- Atlassian. Jira pricing. https://www.atlassian.com/software/jira/pricing
- Linear. Pricing. https://linear.app/pricing
- Productboard. Plans & pricing. https://www.productboard.com/pricing/
- Aha! Roadmaps pricing. https://www.aha.io/roadmaps/pricing
- Notion. Pricing. https://www.notion.com/pricing
- Figma. Plans & pricing. https://www.figma.com/pricing/
- OpenAI Help Center. ChatGPT Business overview. https://help.openai.com/en/articles/8792828-chatgpt-business-overview
- Spherical Insights. Top 15 companies in global product management software market. https://www.sphericalinsights.com/blogs/top-15-companies-in-global-product-management-software-market-statistics-report-till-2035
INTERVIEW ROUNDS COMPARISON
Compared with the cold interviews, round 2 produced more specific answers when we asked for a particular moment or example rather than a general description. Jaycee’s P2 cold opener drew mostly secondhand takeaways, while in P7, “Walk me through your work today” got an hour-by-hour day and “Could you give a concrete example?” got a full project story. Olivia saw the same pattern: P3’s bad day stayed abstract (“It’s mainly people issues that make it feel bad”), while P5 described a missed code cut that wasn’t disclosed until the next day and delayed the release two weeks, which showed us the real problem was late information. Krisha found that targeted questions like coordination surfaced the cause of a bad day (lack of communication) instead of just its feel (messy). What didn’t change is that broad or opinion questions (core skills, “better or worse,” and P12’s “typical day”) still got opinions or generalities, not stories.
READING USAGE
Torres recommends grounding interviews in specific past instances, such as “Walk me through what you did yesterday,” which is one of our three standard prompts. It works because the interviewee isn’t shaping their answer to sound a certain way, so we get an accurate picture of their real day-to-day. It also lets them choose which details to include, and those choices show what stuck with them over the past 24 hours, both good and bad, which points us toward the problems they actually face.
