Strategy and validation for product managers

Top-down starts with industry data and shrinks it. "There are 50 million small businesses, 10% match our profile, of those 5% will pay us $1,000 a year." Bottom-up starts with your actual customers. "We sell at $1,000 a year, we have 200 customers in our beachhead segment, the segment has 50,000 companies." Top-down is faster but easy to fudge. Bottom-up forces you to know your customer and value per customer. Investors trust bottom-up more for that reason, and most pitches use both as a cross-check.

The honest answer: depends on your funding model. A bootstrapped business can thrive on a $50M TAM if it captures 20%. A venture-backed startup needs the TAM to plausibly support a billion-dollar revenue line, so most VCs want to see a TAM north of $1B and a credible path to a few hundred million in SOM within five years. The trap is multiplying a huge TAM by an arbitrary "we'll get 1%" to manufacture the number. Anchor SOM in something real. Comparable players, segment-by-segment buildup, or current pipeline.

Sean Ellis surveyed users with one question: "How would you feel if you could no longer use [product]?" If at least 40% answer "very disappointed," you have product-market fit. He arrived at the threshold by benchmarking hundreds of startups. Above 40%, growth got easier. Below, it stayed grindy. Superhuman used the same survey to climb from 33% to 58% over a year. Ask the survey of users who've experienced the core flow at least twice in the last two weeks. Then segment ruthlessly to see who the very-disappointed group actually is.

No. PMF is a composite. The Sean Ellis score is the single sharpest survey question, but it's a leading indicator at best. Pair it with retention curve flattening, organic growth, and word-of-mouth. The PMF Score Calculator combines multiple signals so you don't bet the company on one survey.

ROI is the net benefit divided by the cost of getting it, expressed as a percent. The hard part is the inputs. Be honest about both engineering cost and ongoing maintenance, and be ruthless about which of the projected benefits would actually hit the P&L. Use a payback period alongside ROI. Payback shows how many months until the feature pays for itself; ROI shows total return over a period. SaaS payback periods commonly land in the 9 to 14 month range for new customer acquisition, with top performers under 12.

Two to three years is a common ceiling for strategic feature work. For features tied directly to acquisition or retention, you want payback inside 18 months and preferably 12. The math is the same as CAC payback, applied to the feature spend. If a $500K feature build saves $20K a month in churn and $30K a month in expansion, payback is 10 months. That's a green light. Shorter payback also frees up cash to reinvest.

Talk to users first. Then put the rough version in front of them and watch what they do. Surveys lie, behavior doesn't. The signals that have held up well: do users come back without prompting, do they pay, do they tell other people. Run the Sean Ellis survey once you have at least 40 active users to ask. If you can't get 40 users to try the rough version, that's also a signal.

Yes, and you usually should anyway. Bottom-up TAM only needs three things: your price, the number of companies or people in your target profile, and a reasonable estimate of who'd buy. You can get the second number from public sources (LinkedIn search, Census data, industry directories) without paying for a report. Bottom-up estimates are easier to defend in a pitch because you can show the math.