What ARPU measures, and why three versions of it exist
ARPU is period revenue divided by the number of users who could have produced it. It answers one question: how much is an average user worth per month? That makes it the natural bridge between a user-growth story and a revenue story — revenue is simply users multiplied by ARPU, so any revenue change decomposes into a volume effect and a price-and-mix effect.
The three variants differ only in the denominator, and each answers a different question.
ARPU uses all active users, including those on a free tier. It is the figure that tells you whether your whole audience is being monetised, and it is the one telecoms and consumer apps report. ARPPU uses only paying users, so it isolates pricing and packaging from conversion; a rising ARPPU with a flat ARPU means you are extracting more from the same small paying minority. ARPA uses billing accounts, which is the right denominator for business software where one contract covers many seats — a 40-seat account is one account, forty users, and one invoice.
None of the three is a defined accounting measure. They are operating metrics, and the denominator is a judgement you make. That is precisely why the SEC's disclosure guidance on key performance indicators asks registrants who present metrics like these to define them clearly, state how they are calculated, and flag any change in the method — because moving from “monthly active” to “registered” users can change ARPU several-fold without anything happening in the business.
The formula, the denominator problem, and the identity that matters
The numerator is the easy half: recognised revenue for the period, from the same population you are counting in the denominator. Strip out revenue from other product lines, one-off professional services, and anything not attributable to the users you are measuring.
The denominator is where every real disagreement lives. Your user count moves every day, so you need a single number for the period. The common convention is the average of the opening and closing counts, which this calculator uses: a business that grew from 92,000 to 98,000 users is credited with 95,000. That two-point average is exact when growth is linear and biased when it is not — a base that spiked mid-quarter and fell back will have its ARPU overstated. If you have monthly counts, average all of them instead.
The identity worth internalising is this: ARPU = ARPPU × paid conversion share. Multiply revenue per paying user by the fraction of your users who pay, and you recover ARPU exactly. On the default figures, $63.16 × 8.00% = $5.05. This is not a rearrangement for its own sake; it tells you that ARPU can only move for two reasons, and they call for entirely different responses. If ARPPU is flat and ARPU fell, your conversion funnel broke — look at the conversion rate. If conversion is flat and ARPU fell, your pricing or mix moved, and the plan-level detail is where to look.
Annualising is a straight scaling: multiply by 12 and divide by the number of months in the period. Do this only to make periods comparable, never to forecast. A monthly ARPU multiplied by 12 assumes no churn and no price change for a year, which is why ARR is built from committed recurring revenue rather than from an annualised ARPU.
Worked example: $480,000 of monthly revenue across a growing user base
A freemium app earns $480,000 in a month. It opened the month with 92,000 active users and closed with 98,000. Of those, 7,600 paid something during the month, spread across 5,200 billing accounts.
- Average active users. (92,000 + 98,000) ÷ 2 = 95,000.
- ARPU. $480,000 ÷ 95,000 = $5.0526 per active user per month.
- ARPPU. $480,000 ÷ 7,600 = $63.16 per paying user per month.
- ARPA. $480,000 ÷ 5,200 = $92.31 per account per month.
- Paid conversion share. 7,600 ÷ 95,000 = 8.00%.
- Check the identity. $63.16 × 0.08 = $5.05 — the ARPU from step 2, as it must be.
- Annualised ARPU. $5.0526 × 12 = $60.63 per active user per year.
- Seats per account. 7,600 ÷ 5,200 = 1.46 paying users per account, which is why ARPA exceeds ARPPU by that same factor: $63.16 × 1.46 = $92.31.
Now use the numbers to make a decision. Lifting the paid share from 8.00% to 10.00% — a quarter more conversion — raises ARPU to $6.32 and monthly revenue to $600,000 at unchanged pricing. Raising ARPPU by a quarter instead, from $63.16 to $78.95, produces exactly the same $600,000. Which lever is cheaper to pull is a business question; the arithmetic is indifferent.
How to read the result
There is no universal good ARPU, because the figure scales with your price point and your free-tier size. A $5 consumer app and a $2,000-a-month enterprise tool are both healthy at wildly different ARPUs. Judge it three ways instead.
Against your own trend. Rising ARPU with flat users means monetisation is improving. Falling ARPU with rising users usually means growth is arriving in a cheaper segment or a cheaper geography, which is not necessarily bad but must be deliberate.
Against acquisition cost. ARPU is only meaningful next to what a user costs. Divide gross-margin ARPU into your customer acquisition cost to get a crude payback in months; do it properly with the CAC payback period calculator. An ARPU that rises while CAC rises faster is not progress.
Against churn. ARPU is the numerator of the lifetime-value formula, and churn is its denominator. A 20% ARPU gain and a 20% churn increase roughly cancel. Feed both into the customer lifetime value calculator rather than judging either alone.
Watch the gap between ARPU and ARPPU, too. A very wide gap means a large free tier: cheap to serve if the marginal cost is near zero, expensive if free users consume support and infrastructure. A narrow gap means almost everyone pays, which caps how much conversion work can contribute to future growth.
The same revenue at different denominators
| Metric | Denominator | Count | Result | Must be disclosed with it |
|---|---|---|---|---|
| ARPU | Average active users | 95,000 | $5.0526 | The activity definition |
| ARPPU | Paying users | 7,600 | $63.16 | Whether trials count as paying |
| ARPA | Billing accounts | 5,200 | $92.31 | Seats per account (1.46) |
| Annualised ARPU | Average active users × 1 yr | 95,000 | $60.63 | That it assumes no churn |
| ARPU at period end | Closing users only | 98,000 | $4.8980 | That the base grew 6.5% |
The final row shows the size of the denominator choice: using closing users rather than the average cuts reported ARPU by 3.1% on this data, with no change in revenue.
ARPU is a non-GAAP operating metric, so define it in writing
ARPU appears in the earnings materials of most listed subscription and telecom businesses, but no accounting standard defines it. The SEC's 2020 interpretive release on key performance indicators in Management's Discussion and Analysis sets the expectations for that situation: disclose a clear definition of the metric and how it is calculated, explain why it is useful to investors, and if you change the method of calculation, say so and give the reason and the effect.
The same discipline is worth applying internally. Write down your activity window, whether trials and internal accounts are counted, how refunds and chargebacks are handled, and which revenue lines are in scope. A metric whose definition drifts quietly between quarters is worse than no metric.
Mistakes that make ARPU misleading
- Changing the activity definition without saying so. Switching from registered users to 30-day actives can multiply reported ARPU without a dollar of extra revenue.
- Using closing users instead of an average. On a base growing 6.5% in the period this understates ARPU by about 3%; on a fast-growing base the distortion is far larger.
- Comparing ARPU across geographies without saying so. Mix shifts toward lower-priced markets look identical to a pricing failure in a blended number.
- Mixing ARPU and ARPPU between periods. The two differ by the paid conversion share, which on the worked example is a factor of 12.5. Label every chart.
- Including one-off revenue. Setup fees, professional services and hardware sales inflate a metric that is supposed to describe recurring behaviour.
- Annualising a monthly figure and calling it ARR. Annualised ARPU assumes zero churn for twelve months. Committed recurring revenue is a different measure.
- Counting seats as accounts. ARPA computed on seats is just ARPPU under a different name, and it will not reconcile with your invoice count.
Where ARPU sits among the other revenue metrics
ARPU is a per-user figure for a period. Three neighbouring metrics answer adjacent questions and are frequently confused with it.
MRR is the total recurring revenue for a month, so ARPA is simply MRR divided by accounts. ARR is the annualised committed version of the same total, built from contract values rather than from an annualised average. Average order value is the retail analogue of ARPPU — revenue per transaction rather than per user per period — and the two only coincide when customers buy exactly once per period.
Downstream, ARPU is an input rather than an answer. Multiply it by gross margin and divide by net revenue churn and you have lifetime value; combine it with churn and expansion and you have net revenue retention, which is the metric that decides whether your existing base can grow without new customers. If you can only track two numbers, track ARPU and churn: between them they determine almost everything else about a subscription business's economics.
Key terms
- ARPU
- Average revenue per user: period revenue divided by average active users, paying or free.
- ARPPU
- Average revenue per paying user: the same revenue divided by users who paid during the period. At least as large as ARPU whenever paying users are a subset of your active users, which they should be.
- ARPA
- Average revenue per account: revenue divided by billing entities. The right denominator for multi-seat business software.
- Paid conversion share
- Paying users divided by average active users. Multiplied by ARPPU it gives ARPU exactly.
- Active user
- A user who met your activity test in the window — typically a login or a qualifying action in the last 30 days. The definition is yours to set and to disclose.
- Annualisation
- Scaling a period figure to a year by multiplying by 12 divided by the number of months. A comparability device, not a forecast.