Choosing Your North Star Metric: A Framework for Product Analytics

Published on: March 12, 2025

Every product team tracks dozens of metrics. The North Star Metric (NSM) is the one that matters most — the single number that best captures the value your product delivers to users, and that most reliably predicts long-term revenue growth. Getting it right is one of the highest-leverage decisions in product analytics.

What Makes a Good North Star?

A good NSM passes four tests simultaneously:

• Measures user value, not business value. Revenue and DAU are lagging business metrics, not value metrics. Airbnb's NSM is "nights booked" — that's when a host earns money and a guest has an experience. The revenue follows.
• Is leading, not lagging. It should move before revenue moves, so you have time to react. If it's correlated with your 6-month revenue but you can observe it today, it's a good north star.
• Is actionable by your product team. "Brand NPS" is a fine metric but your product team can't move it directly week-over-week. The NSM should be something your PMs and engineers can build toward.
• Is unambiguous. Everyone in the company should understand and agree on what it means. A metric that requires a footnote to explain is not a north star.

Classic North Star Examples

Looking at how successful companies chose their NSMs reveals the pattern:

• Airbnb: Nights booked. The moment of value exchange — the host earns, the guest travels.
• Slack: Messages sent within a team (specifically, teams that hit 2,000 lifetime messages). Below this threshold, teams churn. Above it, almost no one does.
• Facebook: Daily Active Users (DAU/MAU ratio). Tracks habit formation, not just scale.
• Spotify: Time spent listening. More listening time = more value delivered = better ad/subscription revenue.
• HubSpot: Weekly active teams. Not individual users — teams, because team adoption drives contract renewal.

Notice the pattern: each NSM is tied to a specific "moment of value" unique to that business model.

How to Find Your NSM

There's a reliable analytical process for this. Start by mapping the value moment — the specific action a user takes that represents them getting the core value of your product. Then validate it statistically:

import pandas as pd
from scipy import stats

# Hypothesise candidate NSM actions (e.g., 'payment_completed', 'report_shared')
candidate_nsm = 'payment_completed_d7'  # Did user complete a payment in first 7 days?

# Split users by whether they hit the candidate NSM
hit_nsm    = df[df[candidate_nsm] == 1]['retained_d90']
missed_nsm = df[df[candidate_nsm] == 0]['retained_d90']

# Test: does hitting the NSM predict 90-day retention?
t_stat, p_value = stats.ttest_ind(hit_nsm, missed_nsm)
lift = hit_nsm.mean() / missed_nsm.mean()

print(f"D90 retention lift: {lift:.2f}x  (p={p_value:.4f})")
# If lift > 2x and p < 0.01 → strong NSM candidate

Run this for 5–8 candidate actions. The one with the strongest retention lift, highest statistical significance, and most obvious connection to user value is your NSM.

Leading vs Lagging Indicators

The NSM itself is often lagging relative to the inputs that drive it. You need a tree of leading indicators beneath it:

• NSM (e.g., weekly active payments per user)
• Level 1 drivers: activation rate, payment success rate, feature discovery rate
• Level 2 inputs: onboarding completion, D1 retention, app load time, support resolution time

When your NSM drops, the drill-down through this tree shows which driver moved and where to focus. Without this structure, a drop in the NSM is just an alarm with no actionable signal.

Signs Your NSM Is Wrong

NSMs need to be revisited as products mature. Watch for these warning signs:

• The NSM is growing but revenue isn't. The metric has decoupled from value — usually because the product changed and the old NSM no longer captures the real value moment.
• Teams are gaming it. If teams are hitting the metric without delivering user value (e.g., artificially inflating sessions with notifications), the metric has become a proxy that's being exploited.
• It no longer differentiates active from inactive users. If churned users have a similar NSM to retained users in their last month, the metric has lost its predictive power.

Conclusion

The North Star Metric is less a measurement choice and more a strategic commitment. When the whole company — product, engineering, marketing, finance — aligns around a single number that represents user value, decision-making gets faster and clearer. The fight about which metric matters dissipates. The work becomes: how do we move this number in a way that's sustainable?

Getting it right is worth spending two weeks on. Getting it wrong costs you years of optimising for the wrong thing.