Use this when: You're flying blind — no north star metric, no dashboards, or you have dashboards that nobody looks at and that don't drive decisions.
You're done when: You have a north star metric paired with a guardrail, a three-layer metric stack, and a review cadence that the team actually follows.
The Sequence
Template
Connect features to metric buckets
Once your measurement system is live, use it to pressure-test your roadmap. Sort every feature into one of three buckets:
- Metrics Movers — designed to move a specific KPI like growth, engagement, or revenue
- Customer Requests — frequently asked for in support tickets and feedback
- Customer Delighters — surprises that build emotional attachment but aren't requested and won't obviously move a number
Very few features are actually metrics movers, and you should know which ones they are before you ship. Customer requests build trust. Delighters build love. But if you can't label each feature in your sprint with one of these three buckets, your measurement system is disconnected from your roadmap.
Nash also makes a distinction most teams miss: there are two ways to improve engagement metrics.
- Lowering friction — fewer clicks, faster load times, smoother onboarding. Measured by funnel drop-offs.
- Increasing desire — making users actually want to come back. Measured by frequency and depth of use.
Software teams obsess over friction and rarely invest in desire. Your measurement system should track both.
Metrics Change Shape as You Scale
Elad Gil watched this pattern repeat across Google, Twitter, Stripe, and Airbnb: the metrics that matter at 20 people are not the metrics that matter at 2,000. Early on, you track product-level signals — activation, retention, engagement. Post-product-market-fit, distribution metrics become just as important. Gil argues that "outsized companies like Google, Facebook, and Uber were aggressive and calculating about growth from their earliest days" — but the metrics they tracked shifted as the company grew. His advice: define the right metrics, get agreement on them, and track them to enable alignment on product priorities. The emphasis is on agreement. A metric nobody looks at is worse than no metric, because it creates the illusion of measurement without the discipline.
Gil also pushes PMs to be deeply technical about data. "The more technical the product manager, the more likely they are able to analyze the data needed to make crucial trade-offs." If your PM can't query the data themselves, there's always a bottleneck between insight and action. Build your measurement system so the people making decisions can access the numbers directly, not through a data team queue.
Example
Hiten Shah's rule from building KISSmetrics: the PM owns the retention metric. Not the growth team, not marketing. If the product isn't delivering value, users churn, and that's a product problem. Make retention someone's explicit responsibility at the product layer, not a shared concern that nobody actually tracks.
A collaboration SaaS had 12 metrics on their dashboard and nobody checked it. They simplified:
- North star — percentage of teams with 3+ active members within 14 days
- Guardrail — support ticket volume doesn't spike
- Business layer — MRR, churn rate
- Product layer — activation rate, weekly active teams, feature adoption
- Input layer — onboarding funnel steps, error rates, load times
They set a Monday morning review for the product layer, daily Slack alerts for input metrics crossing thresholds. Within a month they spotted that mobile onboarding was 3x worse than desktop — something invisible in their old dashboard because it averaged everything together.
Written with ❤️ by a human (still)