Every SaaS company has the same instinct about services revenue, and it's mostly right.
One-time work doesn't recur, so it doesn't compound. It's people-delivered, so it doesn't scale. It gets a worse multiple in a valuation, and a business with too much of it stops being a software business. All true, all worth taking seriously.
Which is why a 25% services attach rate reads as a problem rather than as a result.
But some of that one-time work isn't revenue. It's an instrument, and it's being priced as an output when it's functioning as an input.
Run the Numbers
A customer on $10,000 a year buys $2,500 of training. Non-recurring, people-delivered, everything the instinct dislikes. On the report it's 25% of that account's revenue in the wrong category.
What the training actually does is move them through something faster. Their team gets competent at the thing sooner, so they hit the volume, the output, the operational state that makes the next purchase genuinely useful to them.
They reach it in six months instead of eighteen. And there they buy the add-on, which is $5,000 a year.
That $2,500 didn't earn $2,500. It bought $5,000 a year of recurring revenue, which is the exact revenue the instinct says you should be protecting.
The ratio improves too, since $2,500 against $15,000 reads better than $2,500 against $10,000. Ignore that. It's a timing artifact and it isn't the argument.
Two Ways It Pays, and Nobody Models Either
Acceleration. If the customer would have reached that milestone anyway, you pulled twelve months of $5,000 forward. That's a year of revenue that would otherwise sit in the future, plus a year of compounding on top of it.
Probability. If they wouldn't have reached it, the entire annuity exists because of the training. That's not $5,000 pulled forward, that's $5,000 a year that was never going to happen.
The second case is the larger one and it's the one nobody counts, because a milestone the customer never reached leaves no record. There's no line item for the expansion you didn't get from the customer who stalled.
What That Costs Per Dollar
Put it beside acquisition, which is the only fair comparison.
$2,500 producing $5,000 of ARR is 50 cents per dollar of revenue in the first year. It recurs, so by year two the same $2,500 has produced $10,000 and the cost is 25 cents. By year three it's under 17 cents.
Buy that same $5,000 of ARR through acquisition and you'll pay something close to a full year of it, and then pay again next year for the next one.
Revenue acquisition cost is the number that makes this visible, and almost nobody runs it on services. The cheapest ARR most companies ever buy is sitting in a category they've been trying to shrink.
Why the Reporting Can't See It
Three reasons, and they compound.
The two numbers live in different rows. Services revenue and subscription revenue are reported separately by design, so nothing connects the $2,500 to the $5,000 it produced.
They live in different periods. The cost lands in one quarter and the return lands two or three quarters later, by which point nobody is looking for a cause.
And the metric everybody watches is a ratio. Services as a percentage of revenue is a mix number, and mix numbers can only ever tell you there's too much of something. There's no version of that metric that can say this line bought that one.
Which Services Actually Qualify
Not all of them, and the distinction matters more than the argument.
Plenty of one-time work is just work. Implementation you should have automated. Configuration the product ought to handle. Onboarding labour you're giving away and calling it goodwill. None of that is an accelerant, and pricing it doesn't make it one.
The test is a single question.
Does this move the customer toward a milestone they'd otherwise reach later, or not at all?
Training that makes a team competent enough to hit volume qualifies. Migration that gets their real data in, so the thing they bought finally has something to work on, qualifies. A workshop that produces the internal decision they've been stuck on for two quarters qualifies.
Fixing what should have worked out of the box does not. That's a debt, not an offer, and charging for it is a different conversation with a worse ending.
What to Do With This
Take your non-recurring line items and sort them into two piles: work that fixes something, and work that moves somebody forward.
For the second pile, name the milestone each one accelerates and the recurring item waiting on the far side of it. Then measure what it produced rather than what it earned.
Some of what you've been apologising for in board meetings is the highest-return money in the business. You just haven't been allowed to report it that way.
Lincoln Murphy formally named and popularized Customer Success starting in 2010 and has spent 15 years connecting it to expansion revenue and commercial outcomes. Read The Premise.