ARPU: What Average Revenue Per User Really Tells You (and When It Lies)
Average revenue per user is the fastest way to understand what kind of company you're looking at — and one of the easiest metrics to read wrong. The formula, the real numbers, and the one movement that fools people.
On this page
ARPU — average revenue per user — is revenue for a period divided by the users active in that period. Dropbox closed fiscal 2025 with $2.521B of revenue from 18.08 million paying users: an ARPU of $138.91. That one division tells you what kind of machine you’re looking at — a high-volume consumer utility — before you’ve read a single strategy slide.
ARPU = total revenue in a period ÷ active users in that periodMonthly or annual — decide once, label always.
ARPU, ARPA, ACV — three sizes of “per”
The family gets muddled because the unit changes.
Divides by individual users — natural for consumer and per-seat products.
Divides by customers — the right unit when one account holds many users.
Measures the annualized contract, which for B2B is usually the number that matters — we’ve broken that one down separately.
When Outreach’s Manny Medina told Nathan in May 2019 that his average customer had gone from about $2,400 to $3,300 a month in eighteen months, that was ARPA — and the move told the whole story of the company’s march upmarket.
What real per-user numbers look like
Disclosed figures, dated, across the spectrum:
| Company | Per-user figure | When | What it implies |
|---|---|---|---|
| Expensify | ~$9/seat/month | Dec 2017 | Bottom-up land-and-expand; volume is everything |
| Webflow | ~$40/customer/month | Jul 2018 | Prosumer self-serve; CAC must stay near zero |
| SafetyWing | $45/policy/month | Apr 2023 | Consumer subscription with a B2B upsell layer |
| Dropbox | $138.91/payer/year | FY2025 | Mature freemium at maximum harvest |
| Workboard | $125K first-year ACV | Sep 2018 | Different metric entirely — contracts, not users |
The scale-defining fact:
Expensify needed hundreds of thousands of active seats and a growth motion with no salespeople at all.
Workboard needed fifty customers and a coaching team.
Neither number is “better.” Each dictates a company shape, and the mismatch between per-user revenue and go-to-market cost structure is one of the most reliable startup killers we see in the data.
What ARPU is actually for
- Detecting mix shift — rising ARPU with stable users means customers are upgrading, buying more seats, or the mix is moving upmarket. Outreach’s $2,400→$3,300 was exactly that, driven by enterprise deals with security and governance attached.
- Pricing power receipts — when SafetyWing moved nomad insurance from $35 to $45 over six years (“inflation adjustments,” per CEO Sondre Rasch), retention held, and the raise dropped straight through.
- Forecasting — Users × ARPU decomposes a revenue plan into two testable assumptions instead of one blended guess.
When ARPU lies
The classic deception is composition. ARPU can rise while the business shrinks: churn your cheapest users and the average of the survivors climbs. Dropbox’s fiscal 2025 is a live specimen of the reverse-benign version — ARPU slipped from $140.23 to $138.91 and paying users slipped from 18.22M to 18.08M, together confirming contraction rather than repricing; the full history is here.
Never read ARPU without its denominator. The pair moves in four combinations, and only two of them are good news.
Freemium adds one more trap: blended ARPU across free and paid users is a vanity number. Divide by payers (as Dropbox reports) or say explicitly that you didn’t.
Moving it
The honest levers are the ones that show up elsewhere in this series:
- Packaging — tiers that map to willingness to pay (the pricing breakdown).
- Seats and usage — expansion inside accounts.
- The move upmarket — with the caveat Expensify’s David Barrett put memorably in 2017, having refused it: “I’m much more focused on massive scale than on trying to squeeze harder.”
Squeezing and scaling are both legitimate strategies. ARPU is just the gauge that tells you which one you’re actually executing.
Compare per-customer economics across thousands of SaaS companies at getlatka.com/saas-companies.
SourcesNathan’s May 2019 Outreach interview; founder-disclosed per-user figures as dated in the table — Expensify (Dec 2017), Webflow (Jul 2018), Workboard (Sep 2018), SafetyWing (Apr 2023); Dropbox fiscal 2025 disclosures.