Reference
Expansion Metrics,
Defined
ARPA, ACV, ASP, ARPU, NRR, GRR, LTV, CAC, RAC. What each one measures, what actually moves it, and the part almost nobody puts side by side: how each responds to new-logo acquisition versus expansion.
The Short Version
Most of these metrics barely respond to new customers. Several of them do not respond at all, and one of them mathematically cannot. That is not an argument against acquisition, which every company needs. It is an observation about which lever moves which number, and it is worth having in front of you before the next board meeting.
| Metric | New logos move it? | Expansion moves it? |
|---|---|---|
| ARPA | Usually dilutes it | Directly increases |
| ACV | Sets a new one | Raises existing ones |
| ASP | Yes, this is its lever | Largely unaffected |
| ARPU | Usually dilutes it | Increases |
| NRR | No, excluded entirely | The only unbounded input |
| GRR | No | No, excluded by definition |
| LTV | No, adds customers at current LTV | Raises both factors |
| CAC | Yes, this is what it measures | Not measured by it |
| RAC | 35 to 50 cents per dollar | Low teens per dollar |
Read the middle column down. If your only growth lever is new logos, most of the numbers your board asks about are being addressed indirectly or not at all.
The Metrics
MRR
Contracted recurring revenue normalized to a month. The base almost everything else on this page is derived from.
Watch for one-time fees, services, and overages folded in. Anything non-recurring inside MRR overstates it, and every metric calculated downstream inherits the error.
ARR
Annual Recurring Revenue is contracted recurring revenue annualized. Annual Run Rate is a recent period extrapolated to a year, and it frequently includes revenue that will not recur.
Same three letters, two different figures, quoted interchangeably in the same meeting. Run rate is usually the larger of the two and always the softer.
Never build a calculation on a number labelled ARR without establishing which one it is. The gap between them is precisely the non-recurring revenue nobody wants you to ask about.
ARPA
Total recurring revenue divided by the number of accounts. The cleanest single read on whether your customer relationships are getting more valuable over time.
New logos usually dilute it, because new accounts typically start smaller than the average of a base that has had time to grow. Expansion raises it directly, which makes ARPA the metric most honestly tied to expansion performance.
The distinction that matters: ARPA can be inflated fragilely by stuffing the initial sale, which shows up as growth and then bills back as contraction at renewal. Milestone-triggered expansion raises the same number durably, because nothing was sold before the customer could use it.
ACV
The annualized value of a customer contract. Almost everyone treats it as a number set at signing, reports it as a single figure, and optimizes it at the point of sale.
That framing is the problem. ACV is a curve, not a point. The interesting question is not what a cohort's ACV was on day one but what it is eighteen months later, and the only thing that bends that curve upward is expansion.
New logos set a new ACV. Expansion raises the ACV of contracts you already hold, which is why a cohort's ACV eighteen months in tells you more about your operating model than any signing-day average ever will.
ASP
The average value of deals closed in a period. A sales-efficiency metric, and the one acquisition motions are usually tuned to optimize.
ASP is also where an honest objection to strategic unbundling lives: if you take things out of the initial sale, does your ASP not fall? Sometimes, slightly. Here is the full accounting.
A leaner initial sale closes faster, because the buyer is not negotiating over things they cannot use yet. The account then ends up above where the bundled deal would have landed, because an item presented at the milestone that earns it prices higher than the same item buried in a list on day one, while renewal contraction never arrives. Optimizing ASP in isolation optimizes one moment in a relationship that has many.
ARPU
Total revenue divided by users. A consumer and product-led metric, appropriate when the user is the buying unit.
For most business-to-business companies, ARPA is the better instrument, because the account is what expands and the account is what churns. If you are tracking ARPU on a base of multi-seat business accounts, you are measuring seat growth and calling it revenue growth.
NRR
What this year's customers are worth a year later: starting revenue, minus churn, minus contraction, plus expansion, on the same cohort.
New logos are excluded from the calculation entirely. You can have the best acquisition year in company history and your NRR will not move by a decimal, which surprises people every time.
Churn and contraction are bounded components; played perfectly, their best possible contribution is zero lost. Expansion is the only unbounded input, which is why you cannot defend your way above 100 percent. You can only expand your way there.
GRR
Retention with expansion excluded. It measures only what you kept, which means it mathematically cannot exceed 100 percent.
GRR is a useful hygiene number and a hard ceiling. It tells you how well you hold what you have, and it will never tell you anything about growth. A company reporting strong GRR and weak NRR is holding its base perfectly and expanding none of it.
LTV
How long a customer stays, multiplied by what they spend while they are with you. Two factors, which means exactly two levers.
Acquiring more customers does not change LTV. It adds customers at whatever LTV your current operating model produces, which makes LTV a property of the machine rather than of the pipeline.
Orchestrated expansion pulls both levers at once: more bought per customer, and longer lifetimes because a customer who can see their next milestone has a reason to stay. That is why LTV compounds rather than adds.
Gross Margin
On this page it matters for one reason: it decides which LTV you are actually holding.
The easiest LTV calculation is revenue across the lifetime. The more accurate one is margin across the lifetime, because that is what the customer contributes to the bottom line. A customer paying a thousand a month who costs nine hundred a month to serve produces thirty-six thousand in revenue LTV over three years and thirty-six hundred in margin. Same customer. Two very different numbers.
Most software businesses are nowhere near that spread, and for most decisions revenue-based LTV is good enough. The full accounting, and when the difference starts to matter, is here.
CAC
What it costs to acquire one new customer. Well understood, universally tracked, and genuinely necessary.
CAC stops at the first close. It says nothing about what it costs to acquire the second, third, or tenth dollar from that same customer, which is the entire blind spot the next metric exists to fill. Related and more useful than CAC alone: how fast it pays back, because payback speed is what funds aggression.
RAC
The total cost to acquire one dollar of revenue, expressed as a percentage. CAC counts customers and treats them as interchangeable; RAC follows the dollars, from every source.
Split by source, the spread is the whole argument. New-logo revenue carries the full load: marketing, sales development, commission. Typically 35 to 50 cents per dollar acquired. Expansion carries almost none of it, because the customer is already qualified and the trust already exists. Low teens. Renewals, single digits.
Same dollar of revenue, a third of the cost or less, depending entirely on where it came from. And the cheap dollars are usually the ones nobody is paid to go get.
Terms People Use Loosely
None of this changes how you operate. It changes whether the number you are quoting means what you think it means.
"Churn"
At least three things. Logo churn is a count of customers who left, which weights a small account the same as a large one. Revenue churn is the dollars that left, which is what reaches the profit and loss. Contraction is an account that stayed and shrank, which is not churn at all and is a purchase-time decision arriving late.
People who say "gross churn" usually mean logos. The number they are shown is often dollars. Losing five small accounts and losing one large one are identical logo churn and very different revenue churn.
"Retention"
Gross or net, and the two answer different questions. If the number can exceed one hundred percent, it is net.
Blended, cohort, or per customer
LTV, ARPA, ACV and ARPU all mean something different depending on what they are averaged across, and the level is almost never stated. A blended figure is diluted by every new customer and by segments that behave nothing alike. A cohort figure is the only one that shows whether the number is improving.
A blended figure and a cohort figure are frequently different by a multiple. Ask which one you are looking at before you act on it.
"Unit economics"
A category, not a metric. Different people put different things inside it. Ask which specific numbers they mean.
What to Do With This
The reason to put these side by side is not to demote acquisition. Every company needs new customers, and nothing here argues otherwise.
It is to notice that the five numbers a board asks about mostly respond to a lever that most companies have never built. NRR excludes new logos by definition. LTV is a property of the machine. ARPA dilutes as you acquire. And the cheapest dollars in the business, by a factor of three, are the ones sitting in accounts you already serve.
The number underneath all of it is what your existing base should be producing and is not. That is latent revenue, and it takes ninety seconds to find out roughly what yours is.
The companion specification is Revenue Under Management: how a book is sized and measured, the seven customer events that change it, and what it cannot see.