Five metrics cover most of what you need. Each has a definitional trap that makes the number wrong in a way that flatters you.
MRR
Monthly recurring revenue: the normalised monthly value of active subscriptions.
The traps:
- Annual plans. A $1,200 annual plan is $100 of MRR, not $1,200 in the month it was paid. Booking it as a spike makes every subsequent month look like a collapse.
- One-off charges. Setup fees, overages and consulting are not recurring. Keep them out.
- Discounts. Book the amount actually charged, not list price.
- Failed payments. A subscription in dunning is not revenue until it collects.
MRR is only useful when you break it into movements:
Starting MRR
+ New (first-time customers)
+ Expansion (upgrades from existing)
− Contraction (downgrades)
− Churn (cancellations)
= Ending MRR
Two businesses can both show 10% growth: one with all new customers and no churn, one adding 30% and losing 20%. They are entirely different companies and only the breakdown shows it.
ARR is just MRR × 12. It is not a separate metric, and quoting ARR for a business with monthly plans and 8% churn is misleading.
Churn
Two kinds, and conflating them hides the truth.
Customer churn = customers lost ÷ customers at start of period.
Revenue churn = MRR lost ÷ MRR at start of period.
They diverge sharply when customers are different sizes. Losing ten $10 customers and keeping one $500 customer is 91% customer churn and 17% revenue churn.
Net revenue churn subtracts expansion from churned revenue. If expansion exceeds churn you have negative net churn, meaning revenue grows without new customers. That is the strongest signal in SaaS and rare below a few hundred customers.
Rough guidance: 3–5% monthly customer churn is normal for small self-serve SMB products. Above 8% you have a retention problem no amount of acquisition fixes.
LTV
Lifetime value. The simple formula:
LTV = ARPU × gross margin % ÷ monthly churn rate
The traps:
- Forgetting gross margin. LTV on revenue rather than margin overstates by whatever your hosting and support cost.
- Early churn estimates. With three months of data, dividing by a churn rate computed from a handful of cancellations produces a number with enormous error bars.
- Assuming churn is constant. It is not - it is highest in the first weeks and falls with tenure. The simple formula assumes a constant rate and overstates LTV for products with a bad first month.
Below a year of data, treat LTV as an order-of-magnitude estimate.
CAC
Customer acquisition cost = total sales and marketing spend ÷ new customers acquired, in the same period.
The traps:
- Excluding salaries. If you pay someone to do marketing, that is CAC.
- Mixing periods. Spend in month one produces customers in month three. Use cohorts or accept the smearing.
- Blending paid and organic. Blended CAC hides that paid costs $300 and organic costs $20. Report both.
- Counting free signups. CAC is per paying customer.
The ratios
LTV:CAC. Above 3:1 is the conventional healthy threshold. Below 1:1 you lose money on every customer. Above 5:1 usually means you are under-investing in growth, not that you are winning.
CAC payback period = CAC ÷ (ARPU × gross margin). How many months until a customer pays back their acquisition cost. Under 12 months is healthy for SMB; this matters more than LTV:CAC when cash is tight, because it is about survival rather than theoretical returns.
Below roughly $10k MRR, most of these are too noisy to act on. Track MRR movements and customer churn. LTV and CAC ratios need enough customers for the averages to mean anything, and acting on a ratio computed from twelve customers is worse than not computing it.
Launch it where the numbers are checked
RankCert ranks products on domain control we verify ourselves. Listing is free and the link stays dofollow whether or not you display the badge.