Metrics That Matter

Most dashboards measure what's easy, not what's true. Signups, pageviews, and downloads feel like progress and predict almost nothing. The numbers that actually matter are the ones that tell you whether people came back, whether they paid, and whether either is improving, and there are fewer of them than you think.

The concept

A metric is only worth tracking if a change in it would change what you do. Everything else is decoration. The founder's job early on is not to build a beautiful dashboard, it's to know the handful of numbers that reveal whether the business is working, and to be brutally honest about what they say.

Vanity vs actionable

The single most important distinction:

A vanity metric makes you feel good; an actionable metric tells you what to do next. If a number can only go up and never forces a decision, it's decoration.

📐 If a metric can't change a decision, stop tracking it. The test for every number on your dashboard is simple: would a change in it change what you do tomorrow? Signups, pageviews, and followers usually fail that test; activation, retention, and conversion pass it.

AARRR: the funnel in five stages

A durable framework ("pirate metrics") for organising what to measure:

 ACQUISITION   how do people find you?          (traffic, sources, CAC)
 ACTIVATION    do they reach first value fast?  (onboarding, aha-moment)
 RETENTION     do they come back?               (the one that matters most)
 REFERRAL      do they tell others?             (virality, word of mouth)
 REVENUE       do they pay?                      (conversion, LTV, MRR)

Most founders obsess over acquisition (the top) and neglect retention (the middle) and revenue (the bottom), which is backwards. A product that acquires well but retains poorly is a leaking bucket; pouring more in doesn't help (Chapter 48). Fix the middle before you widen the top.

The one metric that matters

At any given stage, pick a single north-star metric that best captures real value delivered, and focus the team on it. It changes over time (activation early, retention next, revenue later), but at each moment there's usually one number that matters more than all the others. A north star keeps you from optimising ten things badly. Choose the one that, if it went up, would most clearly mean the product is genuinely working.

Leading vs lagging

• Lagging indicators confirm what already happened (revenue, churn). True, but too late to act on.
• Leading indicators predict what's coming (activation rate, engagement depth, week-1 retention). Actionable, because you can still change them.

Watch both, but steer by leading indicators, they're the steering wheel; lagging ones are the rear-view mirror.

Cohort analysis: the honest view of retention

Aggregate numbers hide the truth. "10,000 active users" can mean a healthy, growing base or a leaking bucket masked by fresh acquisition. Cohort analysis, grouping users by when they joined and tracking each group over time, reveals which. If week-4 retention is flat across cohorts, you're leaking; if it's rising, the product is genuinely improving. You cannot see product-market fit in aggregate numbers; you see it in the cohort retention curve flattening above zero.

The six numbers before product-market fit

Before fit, you don't need a metrics department. You need to honestly know:

 1. ACTIVATION RATE   % of new users who reach first value
 2. RETENTION CURVE   % still active at week 1 / 4 / 8, by cohort
 3. CONVERSION RATE   % who take the key action (pay, if you charge)
 4. CHURN             % who leave each period
 5. CAC (if spending) what it costs to acquire one customer (Chapter 48)
 6. THE NORTH STAR    the one number that best means "it's working"

 The most important of these is #2. If retention doesn't flatten, nothing else matters.

Instrument before you launch, not after

You can't recover data you never collected. The events that reveal your funnel, signup, activation, key action, return, must be firing before the traffic arrives (Chapters 38, 45). A launch with no analytics is a launch you can't learn from, the one moment of real usage, unmeasured, gone.

📐 Best practice

Track actionable metrics, not vanity ones, if it can't change a decision, drop it.

Organise your funnel with AARRR, and prioritise the middle (retention) over the top.

Pick one north-star metric per stage and focus on it.

Steer by leading indicators; confirm with lagging ones.

Use cohort analysis to see retention honestly, aggregates lie.

Know the six pre-fit numbers cold, especially the retention curve.

Instrument before launch, not after, you can't recover uncollected data.

Define each metric precisely, what counts as "active," "activated," "churned."

Look at the curve, not the snapshot, trend beats point-in-time.

Be honest about what the numbers say, especially when they're bad.

💀 Common mistakes

⭐ Optimising acquisition while retention leaks. Pouring water into a bucket with holes.

Celebrating vanity metrics, signups, downloads, followers that never convert.

No analytics at launch, the one moment of real data, unmeasured.

Reading aggregates instead of cohorts, a leak hidden by fresh signups.

Tracking everything, deciding on nothing, a dashboard nobody acts on.

Undefined metrics, "active" means three different things to three people.

Steering by lagging indicators, reacting to churn you could have predicted.

Snapshot thinking, a number today with no trend behind it.

Ignoring the retention curve, the single clearest signal of fit, unwatched.

Lying to yourself, reframing a bad number until it sounds good.

The professional workflow

 1. INSTRUMENT BEFORE LAUNCH
    signup · activation · key action · return · payment (Chapter 38)

 2. DEFINE EACH METRIC precisely (what counts as active/activated/churned)

 3. MAP YOUR AARRR FUNNEL and find the weakest stage

 4. PICK A NORTH STAR for the current stage

 5. BUILD COHORTS — group by join date, track retention over time

 6. WATCH THE SIX PRE-FIT NUMBERS — especially the retention curve

 7. STEER BY LEADING INDICATORS; confirm with lagging

 8. REVIEW on a regular cadence; act on what the numbers say

 9. RE-PICK the north star as the stage changes

Tools, websites & costs

Product analytics is free at early scale. PostHog, Amplitude, and Mixpanel all have free tiers well beyond what a pre-fit product needs. The cost is discipline, not dollars.

Alternatives & trade-offs

Few metrics vs many. Few (a north star plus the six) keeps focus and risks blind spots; many gives coverage and drowns you in noise you never act on. Early, ruthlessly few, you can't act on more than a handful anyway.

Quantitative vs qualitative. Numbers tell you what is happening; user conversations and session replays tell you why (Chapter 4). Metrics without qualitative context lead to optimising the wrong thing. Pair them, the number flags the problem, the conversation explains it.

Build vs buy analytics. Building your own gives control and costs engineering time you don't have; buying (PostHog et al.) is faster, free at your scale, and standard. Buy, there's no early advantage in a homegrown analytics stack.

Real-time vs periodic review. Real-time dashboards feel urgent and invite overreaction to noise; periodic review (weekly) sees trends and resists twitchiness. Early-stage signal is noisy, review on a cadence, don't stare at a live counter.

Product analytics vs web analytics. Web analytics (Plausible) tells you about traffic and sources; product analytics (PostHog) tells you about behaviour, funnels, and cohorts. You need product analytics to see retention, the metric that matters most. Web analytics alone can't show you fit.

Checklist

• Analytics are instrumented before launch
• Every metric I track is actionable, not vanity
• Each metric is precisely defined
• My AARRR funnel is mapped and the weakest stage identified
• I have one north-star metric for this stage
• I use cohort analysis to read retention honestly
• I know my six pre-fit numbers, especially the retention curve
• I steer by leading indicators
• I review on a cadence and act on the results
• I'm honest about what the numbers say

📓 Case Study: the framework without the data

Project: SOLIS. This is a case study about a metrics plan that was never fed a single real number, because the product never launched (Chapter 45).

⭐ The framework was right, and chosen before launch. The plan explicitly centred on retention as the north star, for a habit-forming self-improvement app, the whole thesis was visible proof of change over 90 days, which is a retention bet by definition. The product was designed around day-streaks, a progress system, and a completion arc, all mechanisms whose entire purpose is to move the retention curve (Chapter 50). Picking retention as the metric that matters, for this product, was exactly correct. The instinct to measure whether people come back rather than how many signed up was sound.

⚠️ But not one of the six pre-fit numbers has a value. Activation rate, retention curve, conversion, churn, CAC, north star, every one is undefined, because:

• Analytics were never wired (Chapter 38). No event fired for signup, lesson completion, streak, or return.
• The product never launched (Chapter 45), so no user ever activated, retained, converted, or churned.
• There is no cohort to analyse, because there is no user who joined on any date.

The dashboard was a blank the size of the whole business. The retention curve, the single clearest signal of product-market fit, and the exact thing this product's design was a bet on, was never plotted, because it had no data points.

⚠️ The instrumentation gap was known and deferred. Wiring analytics was on the launch-readiness list and consistently lost to more building (more lessons, more art). This is the chapter's hardest warning made concrete: a launch with no analytics is a launch you can't learn from, and SOLIS was heading toward exactly that, an unmeasured launch of a product whose core hypothesis (people will retain because they see change) could only ever be confirmed or refuted by the retention curve it wasn't set up to measure.

🚩 Everything in this chapter is untested here, with a bitter edge. SOLIS is a product specifically designed to win on retention that never measured retention. The framework choice was excellent; the execution collected zero data. The most important number in the handbook, for the product most dependent on it, has no value.

What generalises:

1. Choosing the right north star is necessary but not remotely sufficient. SOLIS chose retention correctly and learned nothing, because it never instrumented or launched to feed it.
2. Instrument before you launch, always. The one thing you cannot do after the fact is recover data you never collected, and the retention curve is precisely the data a habit product lives or dies by.
3. A metrics plan is not metrics. A framework with no numbers is a hypothesis, and this one stayed a hypothesis.

Lessons

1. Track actionable metrics, not vanity ones. If a number can't change a decision, it's decoration.
2. ⭐ Retention is usually the metric that matters most. Acquisition without retention is a leaking bucket.
3. Use cohort analysis, aggregates lie. You see product-market fit in the retention curve flattening above zero, never in a total.
4. Choosing the right north star is necessary but not sufficient. You still have to instrument it and launch to feed it.
5. Instrument before you launch. You cannot recover data you never collected.
6. Know your six pre-fit numbers cold, especially the retention curve.
7. Steer by leading indicators, confirm with lagging ones.
8. Pair numbers with conversations, the metric flags the problem, the user explains it.
9. A metrics plan is not metrics. A framework with zero data is a hypothesis, not a measurement.

Next: Chapter 50: Retention and Lifecycle →