How to choose the SaaS metrics that actually run a business

What SaaS metrics are for

SaaS metrics exist to answer one question: what should we do next? A metric that does not change a decision is not measurement — it is decoration that costs attention.

That test — does this number change a decision — is the whole discipline. Most dashboards fail it. They accumulate numbers because the numbers are available, not because anyone acts on them, and the result is forty figures nobody can hold in their head and none of which anyone can tie to a choice.

A useful metrics system is small. It has one north star — the single number that captures delivered value — and one metric per funnel stage to explain how the north star moves. Five to seven numbers, each earning its place. Everything else comes off the board.

What a north star metric is

A north star metric is the single number that best captures the value your product delivers to customers. It is the one you would optimize if you were allowed to pick only one, and its purpose is to point the entire company at delivered value rather than at activity that merely resembles progress.

Good north stars sit close to the moment value is delivered:

• A messaging product: messages delivered.
• A marketplace: transactions completed.
• A collaboration tool: documents actively co-edited.

Each rises only when a customer actually got something. That is the defining property. A north star you can inflate without delivering value is not a north star — it is a vanity metric with a promotion.

Input metrics — the levers you can actually pull

The north star is an output. You do not move it directly. You move the small set of input metrics that feed it — the controllable actions that, done more or better, make the output rise.

This is the distinction Amazon is often associated with: manage the inputs you control, not the output you only observe. A north star tells you whether value is being delivered. The input metrics tell you what to go do about it on a Monday morning.

Take the messaging product whose north star is messages delivered. That number is downstream of things you can act on directly:

Output (the north star)	Inputs (the levers)
Messages delivered	Accounts that connect a channel
	Teams that invite a second user
	Templates published per account

You cannot schedule a meeting to "increase messages delivered." You can schedule one to increase the share of new accounts that connect a channel in week one. The output is the scoreboard; the inputs are the game. A metric you cannot move by choosing to is a result to watch, not a lever to pull — and a team steering only by outputs is steering by things it does not control.

Why revenue makes a poor north star

Revenue is the most tempting north star and one of the weakest, because it is a lagging number — the result of value delivered, not the delivery itself.

A lagging metric tells you what already happened. It can stay healthy for months after the engine that produces it has broken, because churn and stalled activation take time to reach the income statement. By the time revenue falls, the decision that could have prevented it is a quarter behind you.

A better north star sits one step upstream, on the delivered value that revenue follows. When delivered value dips, you see it now — and revenue is still fine, which is exactly when you have time to act. Track revenue, always. Just do not steer by it, because steering by a lagging indicator means steering by the past.

Leading and lagging indicators, and why you need both

Revenue is one instance of a wider split that every metrics system has to get right: leading versus lagging.

• A lagging indicator reports the past. Revenue, realized churn, closed-won deals — they are accurate and they are late. By the time they move, the cause is already behind you.
• A leading indicator moves first. Activation rate, usage depth, product-qualified signups — they change before the lagging number they predict, which is what makes them steerable.

You steer by leading indicators and you are graded on lagging ones. A system built only from lagging numbers is a rear-view mirror: honest about where you have been, useless for the turn ahead.

	Leading indicator	Lagging indicator
Answers	What is about to happen	What already happened
Timing	Moves first	Moves last
Use	To act	To grade
Example	Week-one activation	Quarterly churn

The catch: a leading indicator is only leading if the evidence says so. The link between the early number and the later one is a claim you have to earn, not assume. Watch whether activation actually predicts retention in your own cohorts before you bet the roadmap on it. An unproven leading indicator is just an early number you happen to like — and a metrics system full of those is more dangerous than one with too few, because it feels predictive while pointing nowhere.

How to avoid the vanity-metric trap

A vanity metric is a number that rises reliably and predicts nothing. Total registered users, cumulative downloads, all-time page views — they feel good because they only ever climb, and they mislead for exactly that reason.

The tell is directionality. A number that can only go up is almost always measuring the wrong thing:

Vanity metric	Why it misleads	Better metric
Total registered users	Never falls, even as users leave	Weekly active users
Cumulative downloads	Counts intent, not use	Activated accounts
Total page views	Rewards traffic, not value	Return visits per user

The pattern in the right column: each can fall. A metric that can drop is a metric that tells you the truth when things go wrong, and a metric that tells you the truth is the only kind worth putting on a dashboard. If a number has gone up every month regardless of how the business is actually doing, it is decoration.

How to map metrics to the funnel

The north star tells you whether value is being delivered. To know why it moved, map one metric to each stage of the funnel. The standard five stages — often remembered as acquisition, activation, retention, revenue, referral — each get exactly one explanatory metric.

• Acquisition — are people arriving? One metric, e.g. qualified signups.
• Activation — do they reach first value? E.g. accounts that complete the core action.
• Retention — do they stay? E.g. cohort retention at week four.
• Revenue — do they pay and expand? E.g. net revenue retention.
• Referral — do they bring others? E.g. accounts that refer.

One metric per stage is the constraint that keeps the system usable. When the north star drops, you walk the five stage metrics and find the one that moved — that is your diagnosis, and it takes seconds because there are only five places to look. A stage with five competing metrics gives you an argument, not an answer.

This is also the difference between a north star and a KPI. You have one north star; the stage metrics are your KPIs, each owned by the team that moves it. The north star says whether you are winning; the KPIs say which part of the funnel is helping or hurting.

Velocity — the metric hiding between the stages

The five stage metrics tell you whether people convert. They do not tell you how long it takes. That gap — the time between one stage and the next — is where a business is often stuck without the dashboard noticing.

Velocity is the metric of the space between stages: signup to activation, trial to paid, first value to expansion. A conversion rate can hold perfectly steady while the time to convert quietly doubles. That is a business slowing down, and a funnel that reports only conversion rates will call it healthy the whole way.

Track at least one velocity metric wherever cash or momentum is at stake — usually time-to-first-value and time-to-paid. A deal that closes is worth less if it takes twice as long to close, because the cash arrives later and the pipeline moves slower. Rates tell you the funnel converts; velocity tells you whether it is speeding up or grinding down.

The product-led funnel has seven states, not five

The acquisition-to-referral five stages are a useful shorthand, but a product-led business usually tracks a more granular lifecycle — seven states a customer moves through, each a place the funnel can leak:

Visitor → Signup → PQL → Customer → Active customer → Renewed → Churned.

The two states in the middle are the ones that matter most in a self-serve business. A PQL (product-qualified lead) is a signup whose in-product behavior signals readiness to buy — the activation the whole top of funnel exists to produce. An active customer is one whose ongoing usage predicts renewal, as opposed to one who has paid but gone quiet. Naming both separates "signed up" from "getting value," and "paid" from "still here" — distinctions the five-stage funnel blurs.

Each transition carries two kinds of metric, and keeping them apart is its own discipline:

• Effectiveness — the conversion rate between two states (visitor→signup, signup→PQL, PQL→customer).
• Velocity — the time between them (days from signup to PQL, from PQL to customer, from customer to break-even).

A funnel that reports only effectiveness can look healthy while every transition quietly slows. Track both, at every state boundary, and the diagnosis is never more than one number away.

Measure health, not just activity

Conversion and revenue tell you what a customer did; they do not tell you whether the customer is healthy — likely to stay, expand, and refer. Two signals, combined, do.

The first is a Customer Behavior Index (CBI) — a single composite score of in-product engagement, built from the behaviors that actually predict advancing to the next lifecycle state: login frequency, core actions completed, features touched, depth of use. Weight each input by how strongly it correlates with retention in your own data, and one number tells you which accounts are thriving and which are drifting toward churn — before the renewal date arrives.

The second is NPS — how likely a customer is to recommend you. Together they define customer health = behavior + sentiment: the CBI says what the customer does, NPS says how they feel, and the gap between them is often the most interesting signal you have. When the two disagree, trust the behavior — what customers do predicts renewal more reliably than what they say in a survey, because people misreport their own habits.

The acquisition-economics set — CAC, LTV, and payback

Beyond the funnel sits the question of whether the funnel pays for itself. Four metrics answer it, and they belong together because none of them means much alone.

Metric	What it measures	How it is usually built
CAC	Cost to acquire one customer	Fully loaded sales + marketing spend ÷ new customers won
LTV	Value of a customer over their life	Gross-margin revenue per account ÷ churn rate
LTV:CAC	Whether a customer is worth more than they cost	LTV divided by CAC
CAC payback	How long until you earn CAC back	Months of gross margin to recover the acquisition cost

Two rules of thumb circulate widely here, and both are conventions, not laws: an LTV of roughly 3× CAC, and a CAC payback period under about a year, are numbers many SaaS operators aim for. Treat them as a starting sanity check for your stage and model, not a benchmark you have failed by missing. The right ratio depends on your margins, your churn, and how much capital you are willing to put at risk ahead of revenue.

These numbers are only as honest as their inputs, and the inputs are where they are usually gamed:

• Use gross margin, not revenue, in LTV. A dollar of revenue that costs 40 cents to serve is 60 cents of value, not a dollar. LTV built on revenue overstates every customer.
• Load CAC fully. Count salaries, tooling, and overhead, not just ad spend. A CAC that omits the sales team is a smaller, prettier, wrong number.
• Segment by channel and cohort. A blended CAC hides the channel that is quietly unprofitable behind the one that is subsidising it. The average is where bad channels go to hide.

The recurring-revenue set — MRR, retention, and expansion

In a subscription business, revenue is a stock that leaks and refills every month, so a second set of metrics tracks the shape of the base rather than the flow of new deals.

MRR and ARR — monthly and annual recurring revenue — are the headline. But the number that decides whether a SaaS business compounds is retention, and retention has to be measured in two different ways that can tell opposite stories.

Metric	What it counts	What to know
Logo churn	Customers lost	Can look mild while big accounts leave
Revenue churn	Dollars lost	The one that hits the P&L
Gross revenue retention (GRR)	Revenue kept, before expansion	Cannot exceed 100% — pure leakage
Net revenue retention (NRR)	Revenue kept, after expansion	Can exceed 100% — the compounding number
Expansion revenue	Upsell and cross-sell in the base	The engine that lifts NRR above GRR

Logo churn and revenue churn diverge whenever your customers are different sizes. Lose ten small accounts and keep one large one and your logo churn looks alarming while your revenue barely moves — or the reverse, which is worse and quieter.

The relationship between the two retention numbers is where the real signal lives. GRR is the floor — it can never rise above 100%, and it measures how much you leak. NRR includes expansion, so it can exceed 100%, which is the property operators and investors watch most closely: an NRR above 100% means the business grows even if it never wins another customer.

But NRR can lie in the direction of comfort. Strong expansion from a handful of accounts can mask heavy churn underneath — a healthy net number sitting on top of a base that is quietly emptying. Read GRR and NRR together. NRR tells you whether the base compounds; GRR tells you the truth NRR is capable of hiding.

The right metrics change with the company's stage

There is no fixed set of "the SaaS metrics." The metric that runs the company changes with what the company is trying to prove.

Stage	The question	The metrics that matter
Early	Does the value land?	Activation, cohort retention, qualitative evidence
Growth	Does it scale efficiently?	CAC payback, LTV:CAC, channel-level economics
Scale	Does it compound and last?	NRR, gross margin, cohort quality over time

Early on, you are learning. Efficiency metrics are premature and often actively misleading — optimising CAC before you know the product retains is optimising the cost of filling a leaking bucket. What you need is evidence that people reach value and stay. Retention is the honest early signal; almost everything else is noise dressed as data.

In the growth stage, the question flips to efficiency. Now CAC payback and LTV:CAC earn their place, because you have proven the value lands and are asking whether you can acquire more of it profitably and repeatably.

At scale, the question becomes durability — whether the base compounds and the margins hold as you grow. The metric that told you the truth at seed will mislead you at Series C, and a founder still steering by early-stage signals two stages later is a common, expensive mistake.

Tie your metric set to the current constraint

A business has one binding constraint at a time — the single thing most limiting growth right now. The most useful metrics system points straight at it, and demotes everything else.

If activation is the constraint, your LTV:CAC ratio is a distraction this quarter — improving it changes nothing while the bucket leaks at the top. If the constraint is a channel that will not scale, cohort retention is real but not what your attention should be on today.

So run the set like this:

1. Name the one constraint currently limiting growth.
2. Elevate the two or three metrics that measure it. These are what you review weekly and steer by.
3. Demote the rest to a reference tab — still tracked, no longer driving decisions.
4. Re-pick when the constraint moves, because it will.

This is the answer to "which of these dozen metrics actually matters." It is the small set attached to the thing currently limiting the business — not the longest dashboard, and not the same set you used last year. Measurement follows the constraint, and the constraint moves.

How to make the metrics stick

A metric without a target is a readout, and a metric without a review cadence is forgotten. Finishing the system means three things per metric:

1. A target — the number you are trying to reach, so movement means something.
2. A cadence — how often it is reviewed, so it stays in front of the people who move it.
3. An owner — one person accountable, because a metric everyone owns is a metric no one owns.

Then do the hard part: cut everything else. Every number not on the map comes off the dashboard. This feels like losing information and is actually gaining attention — a team looking at seven numbers acts on them, while a team looking at forty scrolls past all of them.

The point of measurement is not to know everything. It is to know the few things a decision turns on, and to notice quickly when one of them moves. A north star plus a stage map does exactly that, and a wall of vanity metrics does the opposite: it feels like insight and delivers none.