+150 XP

The media funnel, from impression to subscriber

Three people look at the same campaign and report three different subscriber numbers. The performance marketer counts accounts created. The finance analyst counts first payments that cleared. The product manager counts people still active on day 30. Nobody is lying; they are each measuring a different gate. Every calculation in this module (acquisition cost, lifetime value, churn prediction, sector comparison) is built on top of the definitions set here, so the definitions have to be exact before anything is divided by anything.

Take one path through the funnel. Maria sees a 15-second ad during a football match. Three days later she clicks a retargeted social ad. She creates an account with a trial, watches one show, forgets about it, gets charged when the trial converts, and cancels six months later. That is five distinct events, five different populations, and five chances to count the same person twice.

Why the funnel matters more in media than most sectors

Streaming, gaming and publishing sell recurring access, so the funnel does not end at "sale". It continues into retention, because a subscriber acquired in January who churns (cancels or lapses) in March was a wasted acquisition cost. The second consequence is a currency mismatch: the top of the funnel is bought in impressions, the bottom is reported in paying accounts, and nothing in between converts one into the other automatically.

The measurement object has five gates:

  1. Impression to awareness: a real human is exposed to the brand.
  2. Awareness to visit: that person arrives on a landing page, store listing or article.
  3. Visit to signup: an account exists, with or without a payment method attached.
  4. Signup to paid: the first payment clears.
  5. Paid to retained: the second payment clears.

A gate is defined by two things: the event that counts as passage, and the population that had the chance to pass. Get either wrong and the rate is meaningless, however precise the decimal.

Gate 1-2: awareness and click-through

An impression is a delivered ad, not a seen ad. The IAB and MRC viewable impression standard sets the floor at 50% of pixels in view for at least one second (two seconds for video), which is why platform-reported impressions and third-party verified impressions rarely match. CPM (cost per mille), the cost of 1,000 impressions, is the buying currency at this gate; linear TV and connected TV (CTV, streaming device advertising) CPMs sit in the tens of dollars in the US, with CTV usually the higher of the two, per industry estimates that vary widely by campaign.

Reach and frequency matter more than impressions at this gate, because 10 million impressions can mean 10 million people seen once or 1 million people seen ten times. Netflix's ad-supported tier, launched in late 2022, reached tens of millions of monthly active users within two years, which makes the same company both a buyer and a seller of this inventory.

Click-through rate (CTR) is clicks divided by impressions. Media CTRs are typically under 1%, and awareness ads are not designed to convert on the click, so a low CTR is not automatically a broken gate. The trap is treating CTR as a quality score for creative that was bought for recall.

Gate 3: the signup gate

What counts as a signup differs by business model, and this is the single most common source of arguing over numbers.

For a publisher, there is a free registration step before any payment. The New York Times separates registered users (email captured, metered access) from paying subscribers, and reports subscriptions as well as subscribers, because one person can hold the news, Cooking, Games and The Athletic products at once. Counting subscriptions as people inflates the funnel; counting people as subscriptions understates revenue.

For a service with no free trial, gates 3 and 4 collapse. Netflix removed free trials in the US in 2020, so an account is created and billing begins in the same session: there is no trial population to measure. DAZN sells rolling monthly plans alongside annual commitments billed monthly, plus one-off pay-per-view events in some markets, so a "signup" can mean a 30-day relationship or a 12-month one at different prices.

Two counting rules to fix before reporting anything:

  • Gross adds versus net adds. Gross adds are new paying accounts in the period. Net adds subtract cancellations. Marketing is accountable for gross adds; the board usually sees net adds. Netflix stopped reporting quarterly membership numbers from 2025, which is a reminder that even the industry's counting anchor is a choice, not a constant.
  • Reactivations. A former subscriber who returns is a win-back, not an acquisition. Media businesses have large lapsed bases and seasonal returners (a boxing card on DAZN, an election on The New York Times), so folding reactivations into new adds makes acquisition look cheaper than it is.

Gate 4: signup to paid, the conversion gate

The conversion event here is a cleared first payment, not a submitted card. Failed and declined payments (involuntary non-conversion) can account for a meaningful slice of the gap, and they belong in a different bucket from people who deliberately cancelled the trial.

The denominator is where most reported trial conversion rates fall apart. Trials started in March and trials expiring in March are different populations, and if signups are growing, dividing March conversions by March starts understates conversion badly. State the rule explicitly: conversion is measured on the cohort that started the trial, tracked until its trial window closes.

Card-upfront trials convert at higher rates than card-free ones, partly because inertia works in the service's favour. That design choice is now a regulatory question as much as a marketing one: the US FTC pushed to codify click-to-cancel requirements under its negative option rulemaking, which then ran into legal challenge, and the EU has treated manipulative interface design (dark patterns) as an enforcement priority. Conversion built on friction is measurable and also contestable.

Gate 5: the retention gate

Retention is passage through the second payment. Everything after that is a survival curve, not a gate.

Monthly churn rate = Subscribers lost in month / Subscribers at start of month

Three distinctions to hold: voluntary churn (the user cancels) against involuntary churn (payment fails); gross churn against net churn, which offsets cancellations with reactivations and can hide a leaking base behind a growing headline; and cancellation date against end-of-access date, since someone who cancels on day 3 of a paid month is usually still a subscriber for 27 more days. Bundled subscriptions show lower churn because cancelling means unwinding several products at once.

Measurement providers such as Parks Associates and Antenna publish churn benchmarking, and they define the numerator differently from each other and from internal billing systems. Treat any external churn figure as directional until you know which of the distinctions above it uses.

Engagement: the state between the gates

Engagement is not a sixth gate. It is a state variable measured on people who have already passed gate 5, which is why it is defined here and modelled elsewhere in the module.

  • Hours watched per subscriber per month, the standard consumption measure for video services.
  • DAU/MAU ratio, daily actives over monthly actives, common in gaming and news apps; above 20% is usually read as strong stickiness, under 10% means the app is opened rarely.
  • Content completion rate, the share of starters who finish an episode or film, used to judge content-market fit.

The rule that keeps the whole funnel honest is the denominator chain: each rate divides by the output of the gate before it, never by the top of the funnel.

python
def gate_conversion(cohort):
    """Cohort-based funnel: every rate uses the prior gate's output."""
    visits = len(cohort)
    signed_up = sum(1 for u in cohort if u["account_created"])
    paid = sum(1 for u in cohort if u["first_payment_cleared"])
    renewed = sum(1 for u in cohort if u["second_payment_cleared"])
    return {
        "visit_to_signup": signed_up / visits,
        "signup_to_paid": paid / signed_up if signed_up else 0,
        "paid_to_retained": renewed / paid if paid else 0,
    }

Knowledge check

1. Why does the media funnel matter more for streaming and subscription businesses than for one-off purchase businesses?

2. A campaign has a high CPM but a very low CTR. What does this combination suggest about where the funnel is leaking?

3. Maria signs up for a free trial, forgets about it, and gets charged automatically when the trial converts. Which funnel gate does this scenario primarily illustrate a risk at?

MULTIPLE CHOICE

4. Select ALL correct answers about the five gates of the media funnel described in the lesson.

Select all the correct answers.

MULTIPLE CHOICE

5. Select ALL correct answers about why CPM alone is an insufficient measure of media campaign success.

Select all the correct answers.

Putting the whole funnel together

GateWhat counts as passagePopulation measuredDefinitional trap
Impression → visitClick or measured site arrivalVerified viewable impressionsPlatform impressions counted as people reached
Visit → signupAccount createdUnique visitors, deduplicatedSessions counted as visitors
Signup → paidFirst payment clearedCohort that signed upTrials expiring in period used as denominator
Paid → retainedSecond payment clearedCohort that paid onceCancellation date read as end of access
Lapsed → returningPayment after a gapFormer subscribersReactivations reported as new adds

The largest leakage is rarely at the top. It usually sits between visit and paid, which is why onboarding (recommendation quality, first-session experience, payment recovery) is treated as a marketing responsibility in streaming and not only a product one.

🎬 [VIDEO: "How Netflix Uses Data to Fight Churn" - youtube.com - search for recent talks or explainer videos from streaming analytics conferences covering churn prediction and engagement metrics, useful for seeing these formulas applied to real dashboards]

Key takeaways

  • The funnel has five gates: impression to awareness, awareness to visit, visit to signup, signup to paid, paid to retained. Each is defined by a passage event plus the population eligible to pass.
  • A conversion is the event, not the intent: a cleared payment rather than a submitted card, an account created rather than an email opened.
  • Business model changes where the gates sit. No free trial collapses gates 3 and 4 (Netflix in the US); a registration wall adds a free stage before payment (The New York Times); event-driven demand and mixed plan lengths reshape both (DAZN).
  • Separate gross from net at every gate: gross adds against net adds, voluntary against involuntary churn, new acquisitions against reactivations. Blended figures hide the leak.
  • Cohort denominators, not period denominators. Measure the group that entered a gate until its window closes, and divide each rate by the gate before it.