Magic Number

What is a good magic number?

The published thresholds cluster tightly. Lars Leckie's original post drew the line at 0.75: below it, step back; above it, pour on the gas. Scale Venture Partners calls 0.7x a fairly healthy efficiency baseline, treats a result above 1x as a compelling business investment, and one below 0.5x as a company that has not figured out its model.

What the magic number measures

It answers one question: how much recurring revenue does a dollar of sales and marketing spend buy. Rory O'Driscoll of Scale Venture Partners states it in words: take the change in subscription revenue between two quarters, annualize it by multiplying by four, and divide the result by the sales and marketing spend for the earlier of the two quarters.

Written as the original post wrote it:

Magic Number = (QRev[X] - QRev[X-1]) * 4 / ExpSM[X-1]

where QRev[X] is quarterly recurring revenue for period X, QRev[X-1] is the preceding period, and ExpSM[X-1] is total sales and marketing expense for the preceding period.

The lag is deliberate. The prior quarter's spending is assumed to produce the current quarter's growth, so the numerator and the denominator are pulled from different quarters on purpose.

The two versions, and why both exist

There are two formulas in circulation and they are not interchangeable.

The GAAP version is the one above: the sequential change in reported revenue, annualized by multiplying by four, over prior-period sales and marketing expense. Scale Venture Partners explains why it looks like that: the firm developed the magic number as a workaround, replacing annual recurring revenue with GAAP revenue. O'Driscoll's argument for the awkwardness is that because it is a GAAP-based number it is freely available for all public companies and comparable between them.

The ARR version drops the multiplication by four because the numerator is already annual. Scale describes net sales efficiency as taking net new ARR in the numerator, which accounts for both new sales and lost business, over the same total sales and marketing spending for the period. Scale notes that the magic number replaces net new ARR with the difference in GAAP revenue between the most recent periods and then annualizes that figure to bring it closer to an ARR value, which is by definition annual.

The practical consequence: never compare a magic number computed one way to a benchmark computed the other. And when the numerator is net new ARR rather than gross new ARR, churn is already netted out, which changes what the same number means.

Where the metric came from

Two accounts exist, and they do not fully agree, which is worth knowing before attributing it to anyone.

Scale Venture Partners claims it. Its own history says that while evaluating Omniture for an investment, Scale's Rory O'Driscoll was analyzing the company's revenue traction relative to its investment in sales and marketing, saw the company generating more than $2 in first-year revenue for every $1 invested in its go-to-market engine, and exclaimed that it was magic.

The earliest citable published formula is in a Hummer Winblad Venture Partners post by Lars Leckie dated March 17, 2008, which also ties the metric to Omniture: it says the key metric Omniture used to decide how much gas to pour on the fire was the magic number. That post is where the specific formula, the thresholds and the first worked example appear in print.

Both accounts route through Omniture. The safest attribution is that the metric came out of the analysis of that one company and was published by Leckie with Scale claiming its origin.

A worked example, from the original post

The first published example is worth reproducing because the arithmetic is checkable. The post's hypothetical company reports recurring revenue of $1,000,000 in the first quarter, $1,200,000 in the second and $1,500,000 in the third, with sales and marketing expense of $800,000 in the first quarter and $900,000 in the second. The post states the magic number is 1.0 for the end of the second quarter and 1.33 for the third.

Verify both:

• Second quarter. Revenue change: $1,200,000 - $1,000,000 = $200,000. Annualized: $200,000 x 4 = $800,000. Divided by the first quarter's sales and marketing spend: $800,000 / $800,000 = 1.0.
• Third quarter. Revenue change: $1,500,000 - $1,200,000 = $300,000. Annualized: $300,000 x 4 = $1,200,000. Divided by the second quarter's spend: $1,200,000 / $900,000 = 1.33.

Both match the published figures. Note what the second calculation shows: the company grew faster in absolute dollars and spent more, and the ratio still improved, because revenue growth outpaced the spending increase. That is the entire signal the metric carries.

What the benchmarks say now

Scale's long-term median is the most-cited figure. Across a dataset it describes as more than 1,000 growth-stage software and cloud businesses, Scale reports that the long-term median sales efficiency has hovered around 0.7, and specifies that the 0.7 figure is the median magic number for the dataset.

Benchmarkit's 2025 report, covering 2024 data, gives the reciprocal, which is worth translating. It defines the new customer acquisition cost ratio as total sales and marketing expenses divided by new customer ARR, and reports that the ratio increased by 14 percent in 2024 to a median of $2.00 of sales and marketing expense to acquire $1.00 of new customer ARR. Invert that: 1 / 2.00 = 0.5. On a new-business-only basis, the 2024 median was half a dollar of new ARR per dollar spent, materially worse than Scale's long-term 0.7 median on a broader basis.

Benchmarkit also reports that the expansion ratio sat at a $1.00 median against the new ratio's $2.00, which is the single most useful fact in the whole benchmark set: expansion revenue cost half as much to acquire as new revenue.

What it deliberately ignores

O'Driscoll is explicit that the metric is too simple, and lists what it leaves out: churn, gross margins, and the difference between new sales and upsells, all of which matter and add nuance.

That last point is load-bearing for anyone tempted to adjust the metric. The gross-margin adjustment that appears throughout the software metrics literature belongs to CAC payback period, not here. Benchmarkit defines CAC payback period as the number of months to pay back the sales and marketing expenses for new customers on a gross-margin adjusted basis. The magic number has no gross-margin adjustment in either of its published forms.

Common mistakes

• Mixing the formulas. Annualizing an ARR-based numerator by multiplying by four double-counts and produces a number roughly four times too large.
• Using the same quarter for numerator and denominator. The denominator is the prior period's spend.
• Adding a gross-margin adjustment and then comparing to a published benchmark. The benchmarks are not margin-adjusted; CAC payback is.
• Reading a single quarter. One quarter of revenue change over one quarter of spend is a noisy ratio, especially for a company with lumpy enterprise deals.
• Treating a high number as unambiguously good. A very high magic number often means a company is underinvesting in sales relative to demand, which is the reason the original post's advice above 0.75 was to spend more rather than to celebrate.
• Comparing new-business efficiency to blended efficiency. Benchmarkit's own figures show expansion ARR costing half what new ARR costs, so the mix drives the result.

How it relates to adjacent terms

CAC payback period answers the same question on a different axis, in months rather than as a ratio, and it is the version that carries a gross-margin adjustment. If you want to know whether the go-to-market engine works, the magic number is faster to compute from public filings; if you want to know when the cash comes back, payback is the right tool.

ARR is the input the metric was built to avoid. The whole reason the magic number uses sequential GAAP revenue times four is that private ARR figures are unaudited and inconsistently defined, while reported revenue is comparable across companies.

CAC is the denominator's close cousin and the source of most confusion. The magic number uses the whole sales and marketing line rather than an allocated per-customer cost, which is what makes it cheap to compute and also what makes it blunt.