Part four of the monetization playbook. This one is a lookup, not a read-through: which metrics matter, how many customers and visitors a revenue target actually requires, conversion benchmarks by model, and what the whole thing is worth at exit. The choices that move these numbers are in part 3.

#Which numbers matter

MRR is the number you operate on. ARR is just MRR × 12 — the one you say in valuations and comparisons. Operationally, five numbers matter, and they’re the ones the playbook has been optimizing all along:

• MRR growth % — whether anything is working. Watch for celebrating gross adds while churn eats them.
• Churn, by cohort — the denominator under your entire LTV. Blended churn hides that recent cohorts are worse.
• NRR — whether the base compounds or decays on its own.
• Gross margin — what a revenue dollar is actually worth. This one bites when AI COGS get ignored until the bill arrives.
• CAC payback, per channel — which channels you can afford; the blended number hides the dead ones.

with acq as (      -- customers won per channel last month
  select channel, count(*) as new_customers, avg(mrr) as avg_mrr
  from customers
  where date_trunc('month', started_at) = date_trunc('month', current_date) - interval '1 month'
  group by channel
),
cost as (          -- what each channel spent that same month
  select channel, sum(amount) as spend
  from spend
  where month = date_trunc('month', current_date) - interval '1 month'
  group by channel
)
select
  acq.channel,
  round(cost.spend / acq.new_customers, 0) as cac,
  round(
    (cost.spend / acq.new_customers)     -- CAC
    / (acq.avg_mrr * 0.80), 1            -- ÷ (ARPU × gross margin) — replace 0.80 with your GM
  ) as payback_months
from acq join cost using (channel)
order by payback_months desc;            -- dead channels (longest payback) rise to the top

#How many customers

ARR ÷ (price × 12). Run this before building — it tells you what kind of machine you’re signing up to operate:

Read the rows as different lives, not different prices. $9/mo to $1M ARR means operating a nine-thousand-customer consumer machine — support volume, card failures, churn at consumer rates. $299/mo means finding 280 businesses. Same revenue, unrecognizable businesses.

#How many visitors

Customers ÷ conversion. Two rates separate a visitor from a paying customer — visitor → free signup, then signup → paid — and the end-to-end rate is their product (funnels multiply). Both rates are tabulated under conversion benchmarks below; multiplied out at the medians, the models nearly converge — the finding worth a chart:

Per 1,000 visitors, gated freemium’s 90 signups become 5 paying customers; the no-card trial’s 45 become 3.6; the card-required trial’s 35 become 10.5. The model with the worst signup rate produces the most customers per visitor — the card filters tire-kickers before the funnel instead of after, where they cost you COGS and support. Two cautions before you conclude “always require a card”: per-visitor efficiency ignores virality and expansion, which are freemium’s actual case (Decision 5); and these are self-reported survey medians, so treat them as priors and replace them with your own funnel data the moment you have any.

Now the arithmetic the heading promises, worked for $1M ARR at $49/mo: that’s ~1,700 customers (table above), and on a no-card trial at the medians, 1,700 ÷ 3.6 paying per 1,000 visitors ≈ half a million visitors (card-required: ~160k). And that ignores churn: at 3% monthly churn, holding 1,700 customers means replacing ~50 every month — 14k visitors a month forever, just to stand still. That’s the churn treadmill in visitor units, and it’s why the low-price rows of the customer table are traps for anyone without a distribution engine. Sanity-check every product idea this way before writing code: “where do half a million visitors come from?” kills bad ideas cheaper than building them does.

#Conversion benchmarks

Most conversion numbers circulating online are folklore — recycled, misattributed, or averaged across incomparable funnels. The credible public sources, in order of trust: platform-measured billing data (ChartMogul’s GTM report, ~2,500 companies; Recurly, 76M subscribers) beats structured surveys (ChartMogul × Growth Unhinged, Jan 2026, n=200; Lenny Rachitsky × Kyle Poyar, 1,000+ products) beats agency client samples (First Page Sage). Everything below is labeled with which it is.

Free signup → paying, by model (the 2026 ChartMogul survey, B2B software, six-month window):

(The older, larger Lenny/Poyar survey has trials blended at 8–12% good / 15–25% great and freemium at 3–5% / 6–8% — between the split numbers, as you’d expect. Benchmarks drifted down from 2023 to 2026.)

Visitor → signup runs opposite to conversion: freemium ~9–13% of visitors sign up, opt-in trial ~5–8.5%, card-required ~2.5% (survey and agency data; no platform-measured source for this stage exists — be suspicious of anyone quoting one).

with cohorts as (
  select
    date_trunc('month', signed_up_at) as cohort,
    signed_up_at,
    first_paid_at
  from users
)
select
  cohort,
  count(*) as signups,
  round(
    1.0 * count(*) filter (
      where first_paid_at <= signed_up_at + interval '6 months'   -- converted inside the window
    ) / count(*), 3
  ) as conv_6mo
from cohorts
where cohort + interval '6 months' <= current_date   -- CENSORING: only cohorts old enough to have had the full window
group by cohort
order by cohort;

Does price change the numbers? Yes — but nobody publishes that table. No public source segments trial→paid by price band, so anyone selling you “conversion at your price point” invented it. The published direction is consistent though: enterprise-targeting products convert self-serve at the lowest rates, developer tools at about half the median, and ChartMogul’s platform data shows self-serve stops carrying the sale around $100/mo — companies layer in sales there, and sub-$25 products close half their deals inside a week. Reasonable inference: sub-$25 self-serve converts at roughly 2–3× the rate of $100+/mo. This is the dead zone, seen from the funnel side.

B2B vs B2C look contradictory in the sources; both claims are true. Platform data: B2C converts inside week one (~16% by day 7) or never; B2B converts slowly (~2.5% by day 7) but ends higher over six months. And Recurly’s famous 50–66% trial conversions are card-on-file consumer trials, where “converting” mostly means not cancelling before the first charge — a different denominator entirely; never compare it to opt-in numbers.

#What sets the valuation

Two regimes, boundary around $1M ARR:

• Below it, buyers pay for profit. Small products trade at roughly 2.5–4× SDE (seller discretionary earnings — profit with your salary added back). Revenue a buyer can’t keep isn’t worth a multiple.
• Above it, growing, revenue multiples take over: roughly 3–5× ARR for private bootstrapped deals; more with fast growth (public SaaS medians float around 5–7× and swing with the market).

Either way, the multiple — the only part of the formula you still control at exit — is set by revenue quality: growth rate first, then NRR, churn, gross margin, founder-dependence, channel concentration. A dollar of 90%-margin ARR and a dollar of 40%-margin AI-wrapper ARR are not the same dollar, and buyers discount accordingly. Which closes the loop: the five dashboard numbers above, the multiple’s inputs, and this playbook’s targets are the same list.