You need to sign in or sign up before continuing.
Reflection — An Honest Take 8 min

Honest Take — Before You Begin

Honest Take — Module 7: Metrics & Outcomes — One Number Per Product #

You will resist picking ONE number. The resistance will be predictable and reasonable-sounding: each candidate metric "feels important," ignoring any of them feels like flying half-blind, and the engineer in you will want a dashboard with 5-7 metrics because dashboards are easy to build and decisions are hard to make. I'm naming this before you start because the resistance is the entire reason most engineer-founders end up with beautiful dashboards and no decisions. A seven-metric dashboard is a hedge against the commitment of saying "this is the number that, if it moves, I do something — and if it doesn't move, I do something else." Tracking is billable activity inside a 50-person product org; in your context it is procrastination dressed as productivity, and the one-number constraint is what forces the decisions the tracking was supposed to inform.

The harder half of the deliverable is not the number; it's the kill threshold. Most founders can name a metric they'd like to grow. Far fewer can name the value below which they would retire the product. Both thresholds go on the one-page document or the document is theatre — and the kill threshold is what M13 will eventually execute against, which is exactly why you'll want to leave it "to be determined." Write a number even if it feels arbitrary, and revise it quarterly. An arbitrary threshold honestly revised beats a deferred threshold every time, because the deferred threshold is a kill decision you've pre-committed to never making.

On the Sean Ellis 40% rule, I am carrying the disclaimer at full strength because the disclaimer is the point: the "how disappointed would you be if this product disappeared?" survey, with ≥40% "very disappointed" as the product/market-fit signal, is a heuristic Ellis developed circa 2009-2010 through his consulting practice at Dropbox, LogMeIn, and Eventbrite. The 40% threshold is his observed pattern across 100+ startups — not a peer-reviewed effect size; directionally useful, not a statistical bright line. Use it as one input, never as an oracle. For most early indie products, simpler numbers — paid conversion from waitlist, weekly returning users, MRR — are the right one number, because they actually move at your stage and produce signal you can act on.

The two earlier drafts of this module pulled in different directions and the merge is worth making explicit. One was absolutist: one number per product, full stop, everything else is decoration and decoration is theater. The other spent its energy on the OKR question and allowed 2-4 written outcomes per quarter — while being blunt that there are two OKR traditions: Grove's original discipline (the leadership names the few outcomes that matter, in numbers, and reviews honestly at quarter-end) and the cargo-cult descendant most companies practice (cascading individual OKRs, elaborate ceremonies, decisions made elsewhere). The two drafts compose rather than conflict: 2-4 outcomes per quarter for your operation as a whole, one number per product, and no cascade — the cascade is the cargo cult, and you are one person; you'd be cascading to yourself.

As for Measure What Matters, skim it in a bookstore for thirty minutes, confirm the hagiography, and don't buy it. Being willing to dismiss a book everyone cites approvingly, when the substance doesn't earn the reverence, is itself part of this curriculum's stance.

Two more things the module owes you. First, Perri: Escaping the Build Trap is the most important read here, and if you have spent years employed inside product orgs, her description of teams that ship features without changing user behavior will produce a shock of recognition — you have worked inside that pattern, you knew something was off, and she gives you the language for what it was. The uncomfortable second recognition is finding the same pattern in your own products. Second, counter-metrics: name the things that must not break while you optimize the one number — lesson quality while you chase completions, time-to-task while you chase engagement — because the goodharting failure is invisible until it's expensive. And when you build the instrumentation, resist the pipeline. You're an engineer; a Plausible-events-plus-webhooks-plus-SQL-views rig will feel like the professional move. A counter in the database and a weekly cron that emails you the number is enough. The over-engineered pipeline is procrastination on the harder problem, which is deciding things based on the count.

Conclusion #

The discipline is the constraint, not the dashboard: one number per product, both thresholds written, 2-4 quarterly outcomes for the whole operation, counter-metrics named, simplest possible instrumentation. The Ellis 40% survey is one input with its heuristic disclaimer attached, never an oracle. Resistance to picking one number is the signal that you haven't yet committed to the keep-or-kill decision the metric exists to inform — and the kill threshold you write here is the tripwire M13 fires on.

Predictions #

  • Your first draft of the metrics document will list 4-6 numbers. The second, after rereading the module, will list 1-2. The 1 is the discipline.
  • Picking the kill threshold will take longer than picking the number, and you will try to leave it "TBD" at least once.
  • You will over-engineer the instrumentation in week one. Notice the impulse and stop at the database counter plus the weekly email cron.
  • Within 30 days, the one number will surface a question you've been avoiding — usually whether the product's current scope can plausibly ever hit the keep threshold. That question is M4 territory returning with data behind it.
  • Perri will produce a moment of recognition about an organization you've worked in, followed by a less comfortable recognition about one of your own products.
  • You'll be tempted to add vanity metrics — page views, signups, GitHub stars — to the document "for color." Refuse. One number, two thresholds, counter-metrics. Color is decoration; decoration is theater.
  • If you run the Ellis survey, you will be tempted to treat 40% as a pass/fail oracle in either direction. Re-read the disclaimer when it happens.