5 SaaS KPIs to track, with exact formulas
MRR, net revenue retention, CAC, LTV and gross revenue churn, each with its formula, one worked example that ties them together, and the mistakes that make the numbers look better than they are.
A SaaS company can track dozens of metrics. Five of them answer most of the questions a founder, a board or a buyer will ask: how much recurring revenue you have, whether existing customers are worth more or less over time, what a new customer costs, what that customer is worth, and how fast revenue leaks out.
This post gives each one an exact formula and runs all five through one small example company, so you can see how they connect. If you want the wider set (quick ratio, burn multiple, Rule of 40, cohorts), the complete SaaS metrics guide covers them.
The example company
Numbers are round on purpose. A B2B tool at the start of March:
| Item | Value |
|---|---|
| MRR on 1 March | $50,000 |
| Customers on 1 March | 500 |
| New MRR from new customers in March | $4,000 |
| Expansion MRR (upgrades, extra seats) | $3,000 |
| Contraction MRR (downgrades) | $500 |
| Churned MRR (cancellations) | $1,500 |
| Customers who cancelled | 15 |
| New customers in March | 60 |
| Sales and marketing spend in March | $30,000 |
| Gross margin | 80% |
1. Monthly recurring revenue (MRR)
Formula: MRR = the sum of every active subscription, normalized to one month.
- Monthly plans count at their monthly price.
- Annual plans count at the annual price divided by 12. A $12,000 annual contract is $1,000 of MRR every month, not $12,000 in the month it was paid.
- Setup fees, one-off services and non-contracted usage overages are not MRR.
Total MRR hides what moved it, so split the change into components:
Net new MRR = new + expansion + reactivation - contraction - churned
For the example: 4,000 + 3,000 + 0 - 500 - 1,500 = $5,000. MRR on 1 April is $55,000, a 10% month-over-month increase.
Mistakes to avoid
- Counting a returning customer as new. Reactivation is its own line, otherwise acquisition looks better than it is.
- Counting a downgrade on the day it is announced instead of the day it takes effect. Track committed downgrades separately as a warning sign, and book them when they happen.
The MRR calculator does the component arithmetic if you want to check your own month.
2. Net revenue retention (NRR)
Formula: NRR = (starting MRR + expansion - contraction - churned) / starting MRR x 100
New customers are left out on purpose. NRR asks one question: is the revenue you already had on day one growing or shrinking?
Example: (50,000 + 3,000 - 500 - 1,500) / 50,000 = 51,000 / 50,000 = 102% for the month.
Above 100% means expansion from existing customers more than covers what you lose. Below 100% means every dollar of growth has to come from new customers first.
Mistakes to avoid
- Excluding "unhealthy" customers, trials or promotional accounts to make the number nicer. Report NRR for everyone, then segment if you want.
- Annualizing a single good month. Monthly NRR moves a lot; look at a trailing 12-month figure for anything you report upward.
- Confusing it with logo retention. Losing ten small customers while one large one doubles its plan gives poor logo retention and strong NRR. Both are true, and both are worth tracking.
3. Customer acquisition cost (CAC)
Formula: CAC = total sales and marketing spend in a period / new customers acquired in that period
The spend side includes ad spend, salaries and commissions for sales and marketing staff, tools, agencies, events and content. Leaving out salaries is the most common way CAC ends up looking half its real size.
Example: $30,000 / 60 = $500 per customer.
Two refinements matter more than any benchmark:
- Paid CAC versus blended CAC. Paid CAC divides only paid-channel spend by customers from paid channels. Blended CAC divides everything by all new customers. Blended looks better when organic signups are strong, and it can hide a paid channel that is getting more expensive.
- Time lag. Spend in January often produces customers in February or March. If your sales cycle is long, compare this month's customers with spend from one or two months earlier.
CAC payback (months) = CAC / (ARPA x gross margin)
ARPA is average revenue per account per month: $50,000 / 500 = $100. Payback = 500 / (100 x 0.80) = 6.25 months. After that, each customer's gross profit is contribution rather than recovering its own cost. The CAC calculator handles the blended and paid versions.
4. Customer lifetime value (LTV)
Formula: LTV = ARPA x gross margin / monthly customer churn rate
Monthly customer (logo) churn in the example is 15 / 500 = 3%. One divided by the churn rate gives an average lifetime of about 33 months.
LTV = 100 x 0.80 / 0.03 = about $2,667.
LTV:CAC = 2,667 / 500 = about 5.3
A ratio around 3 is the rule of thumb most investors quote as healthy. Well below that, acquisition costs too much for what a customer returns. Far above it can mean you are spending too little on growth.
Mistakes to avoid
- Using revenue instead of gross profit. At 80% margin, using revenue overstates LTV by 25%.
- Trusting the simple formula when churn is not steady. Most products lose more customers in the first few months than later. If you have a year or more of data, a cohort retention table gives a better estimate than one average churn rate.
5. Gross revenue churn
Formula: gross revenue churn = (churned MRR + contraction MRR) / starting MRR x 100
Example: (1,500 + 500) / 50,000 = 4% for the month.
"Gross" means expansion is not netted off. It shows how fast revenue leaks before upsells patch the hole, which is why it belongs next to NRR rather than instead of it. In the example, revenue churn (4%) is higher than logo churn (3%), which means the customers who left or downgraded were larger than average. That is worth a look.
Annualizing correctly. Monthly churn does not multiply by 12. The formula is:
annual churn = 1 - (1 - monthly churn)^12
At 3% a month: 1 - 0.97^12 = 1 - 0.694 = 30.6% a year, not 36%.
Split voluntary from involuntary churn. Failed card payments are fixable with retries and reminder emails. Cancellations are a product or fit problem. Mixing them hides which fix you need.
How the five connect
Lower churn raises LTV. A higher LTV justifies a higher CAC. Expansion lifts NRR and net new MRR at the same time. When one number moves, the others usually tell you why, which is the reason to look at all five together every month rather than one headline number.
Tracking them without a spreadsheet
The work is mostly in getting the inputs from the right systems: subscriptions from billing, spend from the ad accounts, customer counts from billing or the CRM. In clariBI you can connect Stripe as an MCP app, add Google Ads or Meta Ads for spend, and ask in plain English: "What was MRR at the end of each month this year?", "How many new customers did we get each month?" or "What was total ad spend by month, Google Ads versus Meta Ads?" The language model plans the calculation, the arithmetic runs as code on your synced data, and any sentence with a figure the chart data cannot back is removed before you see it. The example questions list has more phrasing that works.
Before you trust any of the five, check the inputs: annual plans stored as annual amounts, cancellations dated when they take effect, and ad spend tagged by channel. A clean formula on messy data still gives the wrong answer.
A concrete next step: connect Stripe in a trial and ask for net new MRR by component for the last six months. If the components do not add up to the change in total MRR, you have found your first data problem, which is a useful thing to find early.