Case 02 · Relate
Rewriting the founders' plan before we built it
I joined a YC-backed CRM to build a 0 to 1 product and was handed a plan. Two weeks of research later I told the founders the plan described a product their own users would not have bought, and brought the evidence for what to build instead.
- Role
- Product Manager.
- Timeframe
- January to May 2026.
- Product
- Sales-deck sharing with prospect view-tracking.
- Company
- Relate, YC S22, B2B sales CRM.
If you read one thing
- I disagreed with the founders and brought evidence, not an opinion. A feature and pricing teardown of 12+ competing tools plus direct research on what reps needed, against a plan that had been written from intuition.
- The output of the disagreement was a smaller roadmap, not a bigger one. I cut the scope to the three features reps actually named. Waitlist enrollment rose 45%.
- Then I made the smaller thing ship fast. Rewriting the build spec as acceptance criteria per screen instead of descriptions took an estimated three-week build down to five days.
01Context
Relate is a YC S22 B2B sales CRM. The 0 to 1 product I owned was a sales-deck tool: reps create a deck, share it with a prospect, and see who opened it and how long they spent on each page. Think DocSend.
I came in to a sitemap and a plan already written by the founders. That is a normal way to hand off work at an early-stage company. It is also written by people who spent the last month raising money and hiring, not talking to sales reps.
02Why I did not just build it
The plan described a fairly complete deck platform: creation, templates, sharing, tracking, routing, analytics. Nothing in it was wrong, exactly. What bothered me was that I could not point at one thing in it and say which rep problem it solved first.
For a company with runway and two engineers, broad is not ambition. It is the most expensive way to find out you were wrong.
So before building I tore down what already existed, and went and found out what reps actually complained about.
- 12+Competing tools torn down on features and pricing, to find what was genuinely uncovered rather than merely absent from our plan.
- 18Screens the original plan implied, once I mapped the feature matrix into an actual sitemap.
- 3Features reps named unprompted when asked what would make them switch tools.
The gap between eighteen screens and three features was the entire finding.
03What I considered
| Approach | Why it was on the table | Why I did or did not take it |
|---|---|---|
| Build the plan as written | It is what I was asked to do, it avoids a conflict in week three of an internship, and the founders know their market. | Rejected. The most expensive way to be wrong, and I would have had no answer when it did not land. |
| Build a thinner version of the same plan | Ships faster, avoids the disagreement, still gets something in front of reps. | Rejected. A smaller version of the wrong shape is still the wrong shape. It would have tested nothing. |
| Re-scope from rep evidence, then present the case | Turns a disagreement into a decision the founders can actually evaluate, because it comes with the research attached. | Taken. Cost two weeks of build time up front and changed what we built. |
04The decision, and the trade-off
I cut the roadmap to the three features reps named, and made prospect view-tracking the centre of the product instead of one capability among many. The pitch stopped being "a better way to build decks" and became "see which prospects opened it, and what they read."
Before writing production code I put up a landing page for the narrower product and watched whether revenue teams signed up. They did. That test cost a day and would have saved weeks if it had failed.
What the cut cost: no templates, no analytics beyond the core view data, none of the routing logic. Every one was a real request from someone. I judged that three things done properly would get a rep to switch and seven done partially would not. Waitlist enrollment rose 45%.
05Making the smaller thing ship fast
Two weeks of research meant I owed the team speed. I specified all eighteen screens and built the first working version with Claude Code.
The first attempt was not good. I had written the spec the way most people do, as a description of each screen: what it contains, what it is for. The output came back plausible and consistently slightly wrong, because a description can be satisfied several ways and the model kept picking a different one than I meant.
So I rewrote it as acceptance criteria per screen. Not "the deck list shows the user's decks with their view counts," but the conditions that had to be true for the screen to be correct. Descriptions are interpretable. Acceptance criteria are testable.
The useful version of "prompting" for a PM is not phrasing, it is specification. A good prompt and a good spec are the same document.
Estimated at three weeks. It took five days, and a pre-launch QA pass caught twenty-odd gaps before any rep saw it.
06Where it got to
200
Reps on the product, its first real users.
+45%
Waitlist enrollment after the roadmap was cut to the three named features.
+70%
Visitors, after an SEO series written against the topics reps were already searching.
The number that mattered most is not on that row. It is that the founders took the re-scope. An intern with an opinion is worth nothing. An intern with a teardown, rep research and a smaller roadmap is making an argument the people who own the company can check.
07What I would do differently
I would have brought the disagreement sooner. I waited until I had a complete case. That felt rigorous and it cost about a week, because the founders would have engaged with a partial finding and pointed me at the right reps faster.
I would have instrumented before launch, not after. I can tell you what the funnel did. I cannot tell you which of the three features the first reps came back for, because the events went in too late. That is the question I most want answered.
I would keep a written record of what we cut and why. We re-litigated templates twice from memory. A one-page list of explicit non-goals would have ended both conversations in a minute.