Seat Exposure: How Much Salesforce, Workday, HubSpot and Atlassian Bill Per User

Atlassian earns 95% of its revenue from products licensed by user count. Workday earns 92.5% from subscriptions priced mainly on its customers' headcount. Only 45.4% of Salesforce revenue sits in the two clouds it plainly sells per user, and HubSpot does not split its 97.8% subscription line at all. Per-seat revenue share is the number every SaaS bear case assumes, and none of these four companies publishes it.

The short answer. How much of Salesforce, Workday, HubSpot and Atlassian revenue is still per user? Most of it. Atlassian's user-licensed share is 95.0% and Workday's headcount-priced share is 92.5%. Salesforce is between 45.4% and 81.8%. HubSpot's 97.8% subscription line mixes seats, contacts and tiers without a split, so its seat share cannot be read from the filing.

The seat exposure scorecard for four SaaS incumbents

The table below sorts each company's most recent annual revenue into three buckets. The first is revenue sold on a per-user or per-employee meter. The second is revenue where the filing mixes seat and non-seat pricing without a split. The third is revenue that is plainly not seat-priced, such as services and marketplace fees.

The ranking runs from most exposed to least. Exposure here means one narrow thing: if customers need fewer human users, how much of this revenue line shrinks mechanically at renewal. It does not mean the company is in trouble. Two of the most exposed businesses on this list are also among the fastest growing.

Percentages are Zan Digital calculations from each company's revenue tables. The classification of each line into a bucket is our judgement, based on the filing text and the vendor's own pricing page. Periods differ by company and are not aligned.

How the per-seat revenue share was built

Every figure starts in a 10-K revenue table. For each company I took the most recent full fiscal year and the line items the company itself reports, whether by product, by deployment or by revenue type. Nothing here is modelled from analyst estimates.

Each line then went into a bucket using two tests. First, does the filing describe how the line is priced? Second, does the vendor's public pricing page sell that product per user, per employee or by some other unit? A line went into the seat bucket only when both pointed the same way.

Where the two tests disagreed, or where one line bundles products with different meters, it went into the mixed bucket. That bucket is deliberately large. Forcing a split the company has not published would mean inventing a number, and the whole point of the exercise is to separate what is disclosed from what is assumed.

This differs from the broader market studies behind the autopsy of per-seat pricing across the SaaS market. Those estimate how many vendors price by seat. This asks a narrower question about four large incumbents: how many dollars still ride on a seat count.

Atlassian: the most per-user revenue in the group

Atlassian reported fiscal 2026 revenue of $6,572.3 million for the year to 30 June 2026. Cloud contributed $4,410.6 million and Data Center $1,830.9 million. Marketplace and other, mostly fees on third-party apps, contributed $330.7 million.

Atlassian's own billing documentation says cloud pricing depends on how many users can access a product and on the plan. Users count toward the bill once added, even if they never log in. Data Center licences are also sold in user tiers. So 95.0% of revenue rides on a user count, the highest share in this group.

Growth that came from seats

The filing is unusually direct about where growth came from. Subscription revenue rose 27%, and the 10-K attributes that increase "primarily" to "paid seat expansion from our existing customers and price increases". That is a seat business working exactly as designed, with customers adding people.

The same filing names the reverse case. It warns that customers "may in the future slow their rate of expansion or edition upgrades or reduce their number of licenses." The risk is written down. It has not yet shown up in the numbers.

Rovo is bundled, not metered

Atlassian has chosen not to sell its AI as a separate meter for most customers. The 10-K describes Rovo as included at no extra cost in paid cloud editions, with pooled Rovo credit allowances in its Teamwork Collection bundle. That keeps the AI tied to the seat rather than replacing it.

I think that is the right call for Atlassian today and the riskiest one for 2028. If an agent closes tickets that a human engineer would have closed, the seat never gets bought. Bundling AI into the seat protects the price per seat, not the seat count.

Workday: priced per employee, not per user

Workday reported fiscal 2026 revenue of $9,552 million for the year to 31 January 2026. Subscription services contributed $8,833 million, or 92.5%. Professional services contributed $719 million and shrank slightly on the prior year.

The meter is what makes Workday different. Its 10-K describes the subscription model as largely based on the size of customers' employee headcount. A company with 20,000 employees pays for 20,000 workers in its HR system, whether 300 HR staff log in or 3,000.

Exposed to layoffs, not to logins

That changes the shape of the risk. Atlassian loses revenue when fewer people need the tool. Workday loses revenue when its customers employ fewer people at all. The 10-K says this plainly, warning that prolonged decreases in customer headcounts and significant reductions in force can hurt its results at renewal.

My view is that Workday's exposure is the more serious of the two over a 5-year horizon, because it is tied to the macro labour market rather than to tool choice. It is also slower. Workday recognises subscription revenue over contracts that typically run 3 years or longer, so a headcount cut reaches revenue only at renewal.

Flex Credits sit inside the subscription line

Workday now sells some AI products through Flex Credits, a pre-purchased credit model. The 10-K names it as a new pricing approach with limited pricing experience behind it. It does not break Flex revenue out of the subscription line, so the 92.5% is an upper bound on headcount-priced revenue. How few customers actually buy those credits is covered in the analysis of Workday's Flex Credits as a seat hedge.

Salesforce: under half of revenue is clearly per user

Salesforce reported fiscal 2026 revenue of $41,525 million for the year to 31 January 2026. Subscription and support was $39,388 million, split across 5 service offerings that were renamed under the Agentforce brand during the year. Professional services and other was $2,137 million.

Salesforce sells its core clouds per user. Its public Sales Cloud pricing page lists every edition in dollars per user per month. The 10-K adds that the number of paying subscriptions at each customer "ranges from one to hundreds of thousands." That range is the seat exposure in one sentence.

Revenue lines from the Salesforce 10-K for the year ended 31 January 2026. Shares calculated by Zan Digital against total revenue of $41,525M. Bucket assignments are judgement, not company disclosure.

The floor and the ceiling

The floor is 45.4%: Sales and Service, the two clouds sold per user with no ambiguity. The ceiling adds Platform, Slack and Other plus Integration and Analytics, which brings the seat-linked share to 81.8%. Marketing and Commerce stays out of both, because the filing gives no basis for putting it in either.

The working is short. $9,028M plus $9,818M is $18,846M, or 45.4% of $41,525M. Adding $8,882M and $6,232M gives $33,960M, or 81.8%. The real figure is somewhere between, and only Salesforce knows where.

Consumption is real but small

The 10-K says Salesforce sells some products, including Agentforce and Data 360, through a consumption-based model, and flags limited experience in pricing them. It does not size that revenue. The fourth-quarter results release put Agentforce annual recurring revenue at about $800 million, which is roughly 1.9% of fiscal 2026 revenue on a run-rate basis.

That comparison mixes a run rate with recognised revenue, so treat it as directional. It still makes the point. Salesforce's metered AI business is growing fast and remains a rounding error next to its seat base. The mechanics of how those credits are bundled into editions are set out in the breakdown of Agentforce editions and bundled credits.

HubSpot: seats, contacts and a meter it will not size

HubSpot reported 2025 revenue of $3,131.3 million. Subscription revenue was $3,063.9 million, or 97.8%. Professional services and other, which includes Payments, was $67.3 million. HubSpot ended the year with 288,706 customers and net revenue retention of 103.5%.

The 10-K names its own recent "seats-based and consumption-based pricing model changes" and warns that customers may renew "for fewer contacts or seats." That sentence lists two meters, which is exactly why HubSpot cannot be ranked on a clean number. The filing does not say what share of subscription revenue follows each one.

A worked example from the pricing page

The public pricing page shows how mixed the meter is. Marketing Hub Professional starts at $800 a month, includes 3 Core Seats and 2,000 marketing contacts, and charges from $45 a month for each extra seat. Sales Hub, by contrast, is priced per seat outright, from $90 to $150 a seat per month on paid tiers.

Take a 10-person marketing team on Marketing Hub Professional at list price. The bill is $800 plus 7 extra seats at $45, or $1,115 a month. The separately billed seats are $315 of that, about 28%. Remove 5 people and the bill falls to $890, a cut of 20% for a 50% cut in headcount. This is an illustration of list pricing, not a measurement of HubSpot's revenue mix.

The credit layer is newer still. Paid Sales Hub tiers include 500 to 5,000 HubSpot Credits, with packs and pay-as-you-go for more. The 10-K calls its consumption pricing "novel and evolving" and does not size it. How that credit layer changes renewals is covered in the analysis of HubSpot's Breeze credits repricing.

What the 10-K filings do not say about seat revenue

The most useful output of this exercise is a list of what is missing. None of the four companies discloses its paid seat or user count in its annual report. None states what share of revenue is consumption-based. Each describes new credit or usage pricing as a risk without telling you how much revenue already depends on it.

That silence is a choice. Microsoft reports Microsoft 365 commercial seat growth every quarter, which shows the number can be published by a company of any size. The seat counts that do get published are compared in the review of five vendors' productivity software seat disclosures.

The gap matters to buyers as well as investors. A vendor that does not publish its seat mix will not volunteer how a pricing change affects your bill either. The general mechanics of that trap are set out in the piece on seat compression and SaaS pricing.

Where this seat exposure scorecard is weakest

The scorecard is only as good as the bucketing, and the bucketing is mine. A different analyst could reasonably put all of Salesforce's Platform line in the seat bucket, or move part of HubSpot's subscription line into it. The direction of the ranking would probably hold. The exact percentages would not.

Seat-linked is not the same as at risk

The strongest objection is that seat exposure does not equal seat compression. Atlassian's seat count grew enough in fiscal 2026 to drive 27% subscription growth. If AI makes software teams larger or more numerous, a high seat share is an asset, not a liability. The scorecard measures sensitivity, not direction.

Pricing power cuts the same way. A vendor can lose 10% of seats and raise list prices by more than that. Salesforce, Atlassian and HubSpot all cite price increases in their filings. Exposure to seat counts only becomes revenue loss if the price per seat cannot rise to cover it.

There are mechanical gaps too. The 4 fiscal years end on 3 different dates, from December 2025 to June 2026. The Salesforce consumption figure is a run rate compared with recognised revenue. And I read the filings for specific language; a risk factor I marked "not found" may exist in a section I did not search closely.

Frequently asked questions

How to use this scorecard this week

If you buy from any of these vendors, pull your last renewal quote and write down which meter each line rides on: users, employees, contacts or credits. Then model your bill at 20% fewer people. The lines that do not move are the ones the vendor has already protected.

If you hold or follow any of these stocks, read the next 10-Q for one thing only. Look for the first company to publish its consumption revenue share or its paid seat count. Whoever discloses first is telling you their mix is moving in a direction they are happy to show.

Last verified 25 October 2026. Figures will be refreshed after each company's next annual report.

Related reading