Average Deal Size: What the Published Figures Assume

A board deck arrives with one line reading "average deal size" and a figure beside it, and no other context. Nobody in the room can say whether that number came from closed-won revenue divided by closed-won deals, from booked contract value divided by every opportunity that reached the pipeline, or from a first-year figure on multi-year contracts. Three defensible calculations produce three different numbers from the same quarter, and the deck shows one of them.

Average deal size looks like the least ambiguous metric on a revenue dashboard and is one of the easiest to compute two ways by accident. The published benchmarks make that worse rather than better, because a figure quoted without its population is a number about somebody else's company.

This page sets out the formula, the figures that are actually published with a dated source and a stated sample, and the segmentation that has to travel with any of them before they mean anything.

The formula, and the two choices inside it

Salesforce's own blog post on the metric, published on 26 April 2024 under the byline Samuel Holzman, gives the calculation plainly. It states the formula as "Average deal size = Total revenue generated / Total number of closed-won deals" and works it through with an example: a company generating $500,000 in revenue across 50 closed deals has an average order size of $10,000.

Two decisions sit inside that arithmetic and neither is visible in the result.

The first is what goes in the numerator. Total contract value, first-year value, annual recurring value and recognised revenue all produce different averages on the same set of deals, and the gap widens with contract length. A three-year contract counted at total value makes the average roughly triple what an annualised count produces.

The second is what qualifies for the denominator. Closed-won deals is the common choice and the one Salesforce's formula names. Teams that include renewals, expansions or downgrades in the same count are measuring something else, usually something flatter, because expansion deals cluster smaller than new business. The same page recommends reading the metric beside win rate, sales cycle length and customer lifetime value rather than alone, which is the honest instruction and the one that gets skipped.

That denominator problem is the same one win rate and closing ratio each carry, and it is worth settling once for all three rather than three times separately.

Total contract valueEverything the customer committed to

• Largest of the three on any multi-year book
• Moves with contract length rather than with selling
• Useful for sizing a commission plan against booked value
• Misleading in any comparison against an annualised benchmark

Annual contract valueOne year of the same contract

• The unit every published SaaS survey uses
• Comparable across companies with different contract lengths
• The right numerator for a benchmark comparison
• Understates what the seller actually brought in

Recognised revenueWhat the period actually earned

• Matches the finance view and the income statement
• Lags the sale by the length of the delivery schedule
• Wrong unit for judging a seller inside a quarter
• Right unit for judging a cohort after one

Annual contract value is the unit worth standardising on if any external comparison is coming, because it is the unit the published surveys use.

What is actually published, with a source and a sample

Very little of what circulates as an average deal size benchmark names its population. One recurring exception is worth reading in full.

SaaS Capital publishes an annual survey of private B2B SaaS companies. Its post "What is the Average Deal Size for Private SaaS Companies?", published on 14 August 2026, describes the source as its 15th annual report with data from over 1,000 respondents, and gives its headline figure: across all companies in the 2026 survey, the median annual contract value is $24,266, down from $26,265 the previous year. The post is explicit about why it exists, stating that public company metrics "tend to skew higher due to scale, funding, and market position" and that the survey is focused solely on private, B2B SaaS companies.

Three cuts of that same survey do more work than the headline.

By company size, the same post reports that companies with $10 to $20 million in annual recurring revenue reported a median annual contract value of $46,788, which it puts at 85% higher than the median of $25,278 reported by companies with $3 to $5 million in ARR. The same SaaS Capital post states that the pattern of rising deal size with company size begins to break down above $20 million in ARR.

By retention, the same post publishes a median annual contract value of $61,802 for companies reporting net revenue retention of at least 120%, against $26,269 for companies reporting NRR below 120%.

By funding type, for private companies above $1 million in ARR in 2025, the post reports bootstrapped companies at a median ACV of $18,643 on a median ARR of $4,400,000, and equity-backed companies at a median ACV of $39,880 on a median ARR of $9,600,000.

Its own closing position is the one to carry away: the post states there is no universal benchmark for deal size that applies to every SaaS company, and offers the data as a reference point rather than a target.

The same figures, as that publisher reports them, with the population attached to each:

Those figures belong to that survey and that population. A services business, an on-premise vendor or a company selling to consumers sits outside it, and nothing in the numbers themselves marks the boundary once they are pasted into a deck.

Why a blended average is the least useful cut

An average across a book with two segments in it describes neither segment.

A company selling a self-serve tier at low value and an enterprise tier at high value has a blended average sitting between the two, at a value no customer has ever paid. Watching that blended number move tells you the mix changed. It does not tell you whether either segment got better or worse, and those are the two questions a sales leader is actually asking.

The fix is not sophisticated. Cut the average by the axes that genuinely change the number, and read each cut on its own: segment, product line, new business against expansion, and channel. Where a cut has too few deals to be stable, say so and read it as a range rather than as a point. Deal counts inside a quarter are usually small enough that one unusual contract moves the mean noticeably, which is why a median beside the mean is worth carrying.

The same reasoning appears in the surveys themselves, which report medians rather than means for exactly this reason.

Before quoting the number

• ✓Yes: Which numerator: total contract value, annual value, or recognised revenue
• ✓Yes: Which deals count: new business only, or renewals and expansions too
• ✓Yes: Mean or median, and how many deals the figure rests on
• ✓Yes: Cut by segment and product line, since a blended figure describes nobody
• ✓Yes: For an external benchmark: the publisher, the date, and the population surveyed
• ✕No: Treating one quarter's movement as evidence about selling rather than about mix

What the number actually decides

Deal size is a design input rather than a scoreboard, and it decides more downstream than a dashboard line suggests.

It decides how much acquisition effort a deal can carry. A motion whose average deal is worth a few thousand dollars a year cannot support a long multi-threaded enterprise process, and one worth six figures cannot be served by a purely self-serve funnel. That constraint is the whole content of transactional selling, and it is arithmetic rather than philosophy.

It decides the shape of the commission plan. The rate people argue about is downstream of contract value and margin, which is why a commission plan that survives a bad quarter derives the rate from deal economics rather than from a published average.

It decides which outbound motion is affordable at all. The threshold that matters is the contract value at which a sale can support human selling time, and B2B SaaS lead generation on either side of that threshold works through what changes on each side of it.

And it decides what a pipeline coverage number means. Coverage is a multiple of a target, and a target built on an average that is drifting produces a coverage figure that looks stable while the underlying book changes shape.

1. Step 1Fix the numerator

   Pick total, annual or recognised value and write the choice down beside the metric

2. Step 2Fix the denominator

   New business closed-won only, unless expansion is deliberately included and labelled

3. Step 3Cut before averaging

   Segment, product line and channel, each reported separately with its deal count

4. Step 4Read it beside its companions

   Win rate, cycle length and retention, since deal size alone explains none of them

Where it belongs on a dashboard

Average deal size is a lagging figure with a long tail, and reading it weekly produces noise and a standing invitation to explain variance that is not there. A quarterly read, cut by segment, against the same cuts from the previous four quarters, is what makes a trend visible.

It also belongs beside a leading figure rather than alone. Which numbers change a decision sets out the workable set and the pairing rule: every lagging number carries a companion that says whether the change came from selling or from mix.

For an outbound team the practical consequence is narrow. If the average is falling and the segment mix has not changed, the targeting has drifted down-market, and the fix sits in the list rather than in the copy. If the average is falling because the mix changed, that was a decision somebody made, and the metric is reporting it correctly.

Where the constraint turns out to be the number of qualified conversations at the top rather than the size of the ones that close, no amount of deal-size analysis repairs it. We are paid on attended meetings that meet criteria agreed in writing before launch. See what a first campaign produces before rebuilding a forecast around a metric that may not be the binding one.

The short version

The formula is total revenue divided by closed-won deals, and the two definitional choices inside it decide the answer more than the selling does. Published benchmarks are usable only with their publisher, their date and their surveyed population attached, and the one survey worth quoting here says plainly that no universal benchmark exists.

Cut the number before averaging it, carry a median beside the mean when deal counts are small, and read it beside win rate and cycle length rather than on its own. Then use it for what it is good for, which is deciding what a sale can afford to cost.

Figures above are quoted as published by SaaS Capital on 14 August 2026 and by Salesforce's blog on 26 April 2024, from pages fetched on 2 September 2026 with dated snapshots retained. Verify current figures with each publisher before relying on them.

Sources: SaaS Capital, What is the Average Deal Size for Private SaaS Companies?, Salesforce, Average Deal Size: The Secret to Forecasting with Confidence