The Funnel: Awareness to Retention, Measured Honestly | Distribution and Audience Building Tutorial

The Funnel as a Mental Model

A funnel is a simple mental model for how users become customers:

Awareness      they've heard of you
Interest       they looked into you
Trial          they tried
Conversion     they paid, or signed up, or committed
Retention      they came back
Referral       they brought others

Real user paths are messier. Someone might hear about you, forget, see you again six months later, trial, abandon, come back three months after that, convert, churn, and then be won back by a drip email. The funnel is not a precise description of reality; it is a scaffold for measuring and improving.

Different models carve this differently:

AIDA: Attention → Interest → Desire → Action (oldest, marketing-focused)
AARRR (Pirate metrics): Acquisition → Activation → Retention → Referral → Revenue (Dave McClure, startup-focused)
The bowtie: Prospect → Lead → Trial → Customer → Advocate (B2B SaaS)

Pick one and use its vocabulary consistently. Mixing them confuses everyone.

What to Measure at Each Step

The value of the funnel is that it tells you where to look when things aren't working. A generic funnel plus a list of metrics per step:

Awareness

Reach: how many people encountered your brand this month?
Impressions: how many ad views, organic views, etc.?
Brand search volume: how many people Googled your name?

Awareness is hardest to measure directly. Use proxy metrics and trend lines.

Interest

Site visits: how many people visited after being aware?
Click-through rate: what fraction of impressions click?
Time on site / page depth: did they engage beyond landing?

Trial

Sign-ups: how many created accounts or started free trials?
Activation rate: how many reached the first real value moment (more below)?

Conversion

Trial-to-paid: of those who activated, how many paid?
Average sale price / ACV: how much per customer?

Retention

D1/D7/D30 retention: classic mobile metric; adapts to many products
Monthly active users / weekly active users: are they actually using it?
Churn: what percentage are leaving each period?
Cohort retention curves: how well does a given month's cohort retain over time?

Referral

Referral rate: what fraction of new users came from existing ones?
Viral coefficient (K-factor): chapter 9 covers this
NPS or proxy: how likely are users to recommend?

Finding the Leak

A healthy funnel looks like this:

1,000  awareness (site visits)
  200  trial (signed up)
   50  conversion (paid)
   30  retained at 30 days
   10  referred someone

Those are illustrative numbers. The pattern to look for: where is the biggest drop-off?

If awareness is fine but interest is tiny, the messaging or the offer is wrong. If interest is fine but trial is tiny, the landing page or sign-up flow is broken. If trial is fine but conversion is tiny, the trial experience isn't demonstrating value, or the pricing is wrong. If conversion is fine but retention is tiny, the product isn't delivering on its promise. If retention is fine but referral is tiny, the product isn't shareable or the social proof isn't built.

The biggest leak gets the most attention. Fixing a 10% → 20% gain on the worst step beats fixing a 55% → 65% gain on the best step almost every time.

Activation: The Most Important Step

Most founders over-focus on top-of-funnel and under-focus on activation. Activation is the moment when a user experiences the product's core value for the first time. It is the single biggest predictor of retention.

Examples:

Slack: a team has sent 2,000 messages. Teams that cross this threshold retain dramatically better
Facebook (historically): 7 friends in 10 days. Users who hit this retain; users who don't usually leave
Dropbox: first file stored and accessed from a second device
Spotify: X songs saved to library

If you don't know your activation metric, find it. Look at retained users vs churned users across their first 30 days. What action predicted the retention? That's your activation event.

Once you know it, the whole funnel becomes clearer. Your job is to get users to that moment as fast as possible, then to do it repeatedly.

The North Star Metric

Most startups benefit from picking one top-level metric that aligns the team. Not every metric that looks like a dashboard.

Good north stars:

Weekly active users (for most B2C)
Monthly recurring revenue (for most B2B SaaS)
GMV (for marketplaces)
Nights booked (Airbnb, historically)
Minutes watched (streaming)
Number of core actions completed (messaging apps, productivity tools)

A north star has two properties:

It captures value to the user (usage, not vanity)
Growth in the north star predicts long-term business growth

"Sign-ups" is a bad north star (you can game it with free trials and discount codes that don't lead to value). "Weekly active users with at least 5 actions taken" is better.

Once you have a north star, the funnel's job becomes: how do we move more users toward this?

Vanity Metrics

Metrics that feel good but don't predict anything:

Total registrations (without an active filter): includes dead accounts
Page views (without engagement): includes accidental visits
Social followers (without engagement rate): most are inactive
GMV (for a marketplace, without liquidity): can grow while the business weakens
Downloads (for an app, without opens): downloads aren't use

Vanity metrics show up in pitch decks because they trend up reliably. They don't belong in operational dashboards. Teach yourself to feel suspicious of numbers that only move in one direction.

Cohort Analysis

A cohort is a group defined by when they arrived. Track them over time. The curves reveal trends a single monthly snapshot hides.

The three shapes from the marketplace tutorial apply here:

Smiling curves: later cohorts retain better than earlier ones. The product is improving faster than the market is hardening
Frowning curves: later cohorts retain worse than earlier ones. Either the product is getting worse, or you're reaching less qualified users
Flat curves: retention stabilises after an initial drop. Stable, not compounding

If you only look at current-month retention, you'll miss these trends. If you look at cohort curves, you see the direction.

Leading vs Lagging Metrics

A critical distinction:

Lagging metrics: revenue, churn, LTV. They tell you what happened
Leading metrics: activation rate, NPS, product-usage patterns in the first week. They predict what will happen

Operators work on leading metrics, not lagging ones. You can't directly change revenue this quarter; you can only change the inputs that produce it. Those inputs are leading metrics.

Example: if activation rate drops from 45% to 35%, revenue will drop in three to six months. By the time the revenue number registers, the problem has been accumulating for a quarter. The early warning is in activation.

Measuring When Volume Is Low

Early-stage startups can't run clean funnel analyses. With 200 users per month, every metric is noisy. Two responses:

Run longer windows: measure quarterly instead of weekly
Use qualitative data: 20 well-conducted customer interviews beat a terrible dashboard at 200 users per month. Chapter 6 of the customer-discovery canon (Blank, Constable) is all about this

At low volume, you have two jobs: make the data you do have more meaningful (longer windows, smaller confidence intervals), and supplement it with qualitative depth. Numbers fail to tell you what users actually think. Interviews do.

Attribution

"Where did users come from?" is harder than it looks. Touch attribution is especially tricky:

Last-touch attribution: credit the final channel before sign-up. Simple, but overweights search (users often search to "find the thing they heard about on a podcast")
First-touch attribution: credit the first channel the user encountered. Better for brand channels, worse for conversion channels
Multi-touch attribution: assign fractional credit to all channels along the path. More accurate, much harder to implement
Self-reported attribution: ask the user at sign-up "how did you hear about us?"

Self-reported attribution is underrated. It is not scientifically clean, but the aggregate patterns are often more useful than what your analytics tool tells you.

Privacy and iOS 14

Attribution is becoming legally and technically harder. iOS 14+ broke a lot of Meta's targeting precision. Cookie deprecation in Chrome is coming. GDPR has been teaching the industry to live with less data. This hurts channels that relied on precise attribution (paid social) more than channels that didn't (content, direct).

Adapt by:

Collecting more first-party data (email captures)
Relying less on pixel-based attribution
Investing in self-reported signals
Measuring at the cohort level rather than the individual

The old ad-attribution era (where you could track every user across every site) is ending. Plan for distribution patterns that survive the change.

Common Pitfalls

"Our conversion rate is 30%." Conversion from what to what? "Conversion" alone is not a metric; it's a category. Always specify the two steps

"MAU is growing." Are they engaging? If MAU is up but minutes-per-user is down, the platform is growing less engaged users faster than engaged ones. Wrong direction

"We have 100k users." Registered or active? The gap is huge. Founders often hide behind the bigger number

"We'll measure later." Without a baseline, you can't improve. Instrument the funnel before you start the growth work, not after

"Attribution is too noisy to act on." Noisy does not mean useless. Aggregate patterns (which channel is trending up, which is trending down) are usable even when individual attribution is fuzzy

Next Steps

Continue to 07-compounding.md for the part of distribution that grows on its own.