TL;DR. Audited filings show how differently hours and products behave. Across the latest annual reports of Salesforce, Workday, ServiceNow, HubSpot and Adobe, subscription lines produced a combined gross margin of 84.6%, while the same companies’ professional services lines produced -9.9%. Three of the five spent more on delivering services than they charged for them. Accenture, a company built on billable people, reported 92% utilization and a 31.9% gross margin in fiscal 2025. Each dollar of delivery cost returned $6.51 of subscription revenue against $0.91 of services revenue in the software group and $1.47 at Accenture. A solo freelancer cannot copy a software company, but can move along the same axis one rung at a time: custom service, standardized service, productized service, self-serve asset, software. The CEO move: decide which deliverable you will stop customizing, and treat the hours you put into building as capital with a payback target. The student move: audit the last 12 months of client work for repeated patterns, then test the smallest product that the pattern supports before building anything larger.
This article covers the transition itself: what to extract from client work and in which order. Once you have picked the deliverable, the companion piece on how to scope, package and charge a fixed price for a productized service covers the pricing step. If recurring client income is the nearer goal, the retainer playbook is the shorter route, and the digital product price benchmarks show what the later rungs are listed at.
Why do billable hours stop scaling?
The economics were described long before the creator economy. William Nordhaus, reviewing the work of William Baumol and co-authors in a National Bureau of Economic Research paper, summarises the argument this way: sectors whose productivity grows more slowly than the economy’s average tend to experience above-average cost increases, and the taste for labor-intensive services, where productivity growth is intrinsically limited, pulls resources toward them. Using US industry data for 1948-2001, Nordhaus found that technologically stagnant sectors clearly had rising relative prices and declining relative real outputs. Baumol’s own examples included the performing arts, handicrafts and custom clothing.
Custom client work belongs in that list. Much of each hour is spent understanding one client’s situation. The price of the hour can rise. The number of hours cannot.
Management research reaches the same point from the firm’s side. Mohanbir Sawhney, writing in Harvard Business Review in 2016, observes that a consulting or law firm that wants to double revenue has to double its professional staff, and that professional services firms struggle to move gross margins above 40% as they scale, against 60% to 90% for product companies. Only the public preview of that article was readable for this piece, so the figures are quoted as Sawhney’s framing, not as measured data. The next section tests them against audited numbers.
What do audited filings show about service margins and product margins?
Some software companies report revenue and cost of revenue separately for their subscription line and their professional services line. That gives a rare like-for-like comparison: same company, same customers, same year, two delivery models.
Table 1. Revenue and cost of revenue by line, as printed in each company’s latest annual report (US$ millions)
| Company (fiscal year end) | Product line | Revenue | Cost of revenue | Services line | Revenue | Cost of revenue |
|---|---|---|---|---|---|---|
| Salesforce (31 Jan 2026) | Subscription and support | 39,388 | 6,796 | Professional services and other | 2,137 | 2,474 |
| Workday (31 Jan 2026) | Subscription services | 8,833 | 1,531 | Professional services | 719 | 790 |
| ServiceNow (31 Dec 2025) | Subscription | 12,883 | 2,569 | Professional services and other | 395 | 414 |
| HubSpot (31 Dec 2025) | Subscription | 3,063.9 | 445.3 | Professional services and other | 67.3 | 63.2 |
| Adobe (28 Nov 2025) | Subscription | 22,904 | 2,027 | Services and other | 540 | 501 |
Source: Forms 10-K filed with the US Securities and Exchange Commission (sources 1-5). HubSpot reports in thousands (3,063,917; 445,336; 67,349; 63,151) and is shown here in millions. HubSpot’s services line includes payments revenue, and Adobe’s includes maintenance for some on-premise licences and advertising offerings, so neither is a pure consulting line.
Table 2. CEOtudent analysis of Table 1: gross margin by line and revenue per dollar of delivery cost
| Company | Product line gross margin | Services line gross margin | Gap (points) | Revenue per $1 of delivery cost: product | Revenue per $1 of delivery cost: services |
|---|---|---|---|---|---|
| Salesforce | 82.7% | -15.8% | 98.5 | $5.80 | $0.86 |
| Workday | 82.7% | -9.9% | 92.5 | $5.77 | $0.91 |
| ServiceNow | 80.1% | -4.8% | 84.9 | $5.01 | $0.95 |
| HubSpot | 85.5% | 6.2% | 79.2 | $6.88 | $1.07 |
| Adobe | 91.2% | 7.2% | 83.9 | $11.30 | $1.08 |
| Five companies combined | 84.6% | -9.9% | 94.6 | $6.51 | $0.91 |
CEOtudent analysis of sources 1-5. Gross margin = (revenue – cost of revenue) / revenue for each line; recomputed with a script. ServiceNow prints its own percentages for these lines, 80% and (5%), which match.
Three things stand out.
The gap is between 79 and 99 points inside the same company. Combined, the five subscription lines earned $87,072 million on $13,368 million of cost, and the five services lines earned $3,858 million on $4,242 million of cost: a gross loss of $384 million on services.
Services are sold at or below cost on purpose. Salesforce states in its filing that its professional services organization helps the adoption of its offerings and helps secure larger subscription contracts, and that the cost of professional services may exceed the revenue from them in future periods. Workday states that third parties provide the majority of deployment services for its customers. The companies with the highest product margins treat hours as a way to sell the product, and push the rest of the hours to partners.
A pure services company looks completely different. Accenture reported revenues of $69.7 billion for the fiscal year ended 31 August 2025, cost of services equal to 68.1% of revenues, and a gross margin of 31.9%. It also reported utilization of 92% and a workforce of approximately 779,000. Even with almost every available hour sold, each dollar of delivery cost returned $1.47 of revenue. That sits inside Sawhney’s “below 40%” description, and the subscription lines sit at the top of his 60% to 90% range. The combined subscription margin is 52.7 points above Accenture’s.
Table 3. Labour marketplaces, as printed in their latest annual reports
| Company | Metric | Latest year | Prior year |
|---|---|---|---|
| Upwork (2025) | Marketplace take rate | 18.7% | 18.0% |
| Upwork (2025) | Total gross margin | 78% | 77% |
| Upwork (2025) | Gross services volume | $4,028 million | $4,008 million |
| Fiverr (2025) | Marketplace take rate | 27.7% | 27.6% |
Source: Upwork Form 10-K and Fiverr Form 20-F (sources 7-8). Take rate is marketplace revenue divided by marketplace volume, as each company defines it. Upwork states that since May 2025 talent on new contracts pay a fee of 0% to 15% of earnings, with a flat 10% on older contracts.
The platforms that intermediate freelance hours do not sell hours themselves. They sell a standardized product (matching, payments, trust) and earn a software-like gross margin on it, while the people doing the hourly work carry the delivery cost.
What does productizing actually mean?
Two academic terms are useful here, and they point in opposite directions.
Servitization is the older one. A 1988 paper by Sandra Vandermerwe and Juan Rada in the European Management Journal carries the term in its title, “Servitization of business: Adding value by adding services”, and a 2017 overview in Industrial Marketing Management by Kowalkowski and co-authors notes that service growth strategies and retreats from them have been pursued for more than fifty years. That literature is about product companies adding services.
Productization is the reverse move, and it is the one that matters for a freelancer. A 2021 conceptual article in the Journal of Business Research by Wirtz, Fritze, Jaakkola, Gelbrich and Hartley describes service productization as a process that transforms variable, ad hoc services into well-defined service products. According to its abstract, a well-defined service product is three things at once:
- Specified: it has a formalized value proposition and is configured, standardized and systemized, often also modularized and bundled.
- Branded: it has a name, symbol or design.
- Priced: it has a clearly stated price.
The abstract also notes potential drawbacks of highly productized services. Only the abstract was readable for this piece.
The practical reading: productizing is not the same as building a course. It starts when a deliverable gets a fixed scope, a name and a public price, and it can stop there.
What is the Productization Ladder?
Table 4. The Productization Ladder (CEOtudent editorial framework)
| Rung | What the client buys | Your time per sale | Entry criteria (move up only when all are true) | What you stop customizing |
|---|---|---|---|---|
| 0. Custom service | Your hours on their problem | All of it | Starting point | Nothing |
| 1. Standardized service | The same outcome, delivered by your checklist | Slightly less, more predictable | Delivered at least 3 times; you can write the steps down; step order rarely changes | Process and tools |
| 2. Productized service | A named package with fixed scope and a stated price | Fixed and capped | Steps hold for 3 more deliveries without rewrites; delivery hours vary little; you can state what is excluded | Scope, price, timeline |
| 3. Self-serve asset | A template, toolkit or course built from your delivery materials | Near zero per sale, plus support | Clients ask for the material itself; the buyer can get the result without your judgment; you can reach buyers beyond your client list | Delivery itself |
| 4. Software or automated tool | Access to a tool that performs the repeated step | Near zero per sale, plus maintenance | The asset sells steadily; one step is mechanical and frequent; you can fund and maintain a build | The remaining manual step |
Three rules keep the ladder honest.
- Climb one rung at a time. Each rung produces the raw material for the next. The checklist from rung 1 becomes the scope document for rung 2. The intake form, examples and templates from rung 2 become the asset at rung 3.
- Every rung is a valid place to stop. The filings above show that large companies keep a services line because it sells the product. A freelancer at rung 2 with a waiting list has no obligation to reach rung 4.
- Earlier rungs keep paying while later ones are tested. The move reallocates hours; it does not end client work.
Which client deliverable should you productize first?
The best first candidate is usually the dullest one: the piece of work that has been delivered many times with few surprises. The Repeat Audit turns that judgment into a score.
How to run it. List every distinct deliverable invoiced in the last 12 months. For each one, score six criteria from 0 to 3, multiply by the weight and add up. Frequency and variance carry double weight because they decide whether there is a pattern to extract at all.
Table 5. The Repeat Audit (CEOtudent editorial framework)
| Criterion | Weight | Score 0 | Score 1 | Score 2 | Score 3 |
|---|---|---|---|---|---|
| Frequency: deliveries in the last 12 months | x2 | 1 | 2 | 3-5 | 6 or more |
| Low variance: how similar the deliveries were | x2 | Each one different | Same goal, different steps | Same steps, different content | Nearly identical |
| Low client input: what you need from the client | x1 | Weeks of access and meetings | Several working sessions | One call and some files | A form |
| Buyer reachability: can you reach more buyers like this | x1 | Only by referral | Through a few known contacts | Through a community or channel you are in | Through search or a marketplace |
| Outcome clarity: can the result be stated in one sentence | x1 | Depends on the client | Roughly | Yes, with caveats | Yes, and it can be checked |
| Shelf life: how long the method stays valid | x1 | Months | About a year | A few years | Stable |
Reading the score (maximum 24). 17 or more: productize first. 11 to 16: standardize it as a service (rung 1) and score it again after three more deliveries. 10 or less: keep it custom and price it accordingly.
Table 6. Worked example with three illustrative deliverables (CEOtudent analysis, hypothetical scores)
| Deliverable | Frequency | Low variance | Low client input | Reachability | Outcome clarity | Shelf life | Weighted score | Reading |
|---|---|---|---|---|---|---|---|---|
| Analytics dashboard setup | 3 | 3 | 2 | 2 | 3 | 2 | 21 / 24 | Productize first |
| Onboarding email sequence | 3 | 2 | 1 | 3 | 2 | 1 | 17 / 24 | Productize first |
| Brand strategy workshop | 2 | 1 | 0 | 2 | 1 | 3 | 12 / 24 | Standardize as a service |
The workshop is the worst first candidate: it needs the most client time and varies the most. It can still earn well as a custom service.
Two checks before trusting the score. First, shelf life deserves extra suspicion in categories that AI tools are compressing; our analysis of which template categories are growing and which are dying is a useful cross-check. Second, a high score says a pattern exists. It does not say strangers will pay for it. That is a separate test, covered below.
How many product sales replace a block of billable hours?
The figures below are illustrative assumptions chosen by CEOtudent, not observed market data: a freelancer billing $100 an hour who wants to replace 40 billable hours a month, selling through a direct checkout link on Gumroad. Gumroad’s published fee for direct sales is 10% plus $0.50 per transaction, with card processing of 2.9% plus $0.30 on top (company fee page, accessed 9 October 2026).
Table 7. Illustrative break-even units (CEOtudent analysis; assumptions are hypothetical)
| Product (rung 3) | Price | Net per sale after fees | Build time | Forgone billing at $100/hour | Sales to recover the build | Sales per month to replace 40 hours at $50 / $100 / $150 an hour |
|---|---|---|---|---|---|---|
| Template pack | $29 | $24.46 | 30 hours | $3,000 | 123 | 82 / 164 / 246 |
| Toolkit or mini-course | $99 | $85.43 | 60 hours | $6,000 | 71 | 24 / 47 / 71 |
| Flagship course | $299 | $259.63 | 120 hours | $12,000 | 47 | 8 / 16 / 24 |
CEOtudent analysis. Net per sale = price – (10% + $0.50) – (2.9% + $0.30). Units are rounded up. Taxes, refunds, support time, advertising and audience-building time are excluded, so real break-even points are higher.
Compare that with rung 2. Take a package sold at a fixed $1,500 that used to need 15 hours as custom work, which is $100 an hour. If standardizing brings delivery down to 9 hours, the same sale earns $166.67 an hour, an uplift of 66.7%, and frees 6 hours. Four such sales a month free 24 hours without a single new buyer.
Three conclusions follow.
- The productized service pays back first. It needs no new audience, only existing demand delivered more consistently. This is why it sits before the asset on the ladder.
- Low-priced assets are a volume business. Replacing 40 hours at $100 with a $29 template takes 164 sales every month. Someone without an audience should read the piece on the cold-start problem for digital products before choosing that rung.
- The higher your hourly rate, the higher the bar. At $150 an hour the same template has to sell 246 times a month. Well-paid specialists often do better stopping at rung 2 or building one higher-priced asset.
The freed hours from rung 2 are the capital for rung 3. In the example, 24 freed hours a month fund the 60-hour toolkit build in two and a half months without reducing income.
What does a 30/60/90 transition plan look like?
Days 1-30: audit and standardize (rung 0 to rung 1).
- Run the Repeat Audit on the last 12 months of invoices. Pick the top-scoring deliverable.
- Write the delivery checklist from memory, then correct it during the next real delivery.
- Track actual hours per step. You need this number for pricing and for Table 7.
- Decide what the package will exclude. Exclusions are what make a fixed price safe.
Days 31-60: package and sell the productized service (rung 2).
- Give the package a name, a one-sentence outcome, a fixed scope, a timeline and a stated price.
- Offer it to past clients and two or three new prospects before publishing it anywhere.
- Deliver it at least three times using only the checklist. Log every deviation.
- Record how many hours each delivery freed compared with the custom version.
Days 61-90: test the smallest asset (rung 3).
- Take the single most requested component of the package (often the template or the intake diagnostic) and make it usable without you.
- Sell it before building it out, using the 7-day pre-sale protocol.
- Set a stop rule in advance: a number of pre-sales and a date. If the number is missed, the component goes back into the service as a bonus and nothing more is built.
- If the number is met, spend only freed hours on the build, and write down the payback target from your own version of Table 7.
Rung 4 is outside the 90 days on purpose: software deserves its own decision after the asset has sold steadily.
How do you decide whether to climb the next rung?
The five-question decision test (CEOtudent editorial framework). Answer for the specific deliverable, not for your business in general.
- Repeat: Have you delivered it at least three times in the last 12 months with the same steps?
- Write-down: Could a competent peer deliver it from your checklist without calling you?
- Stranger: Can the buyer understand the outcome and the price without a sales call?
- Reach: Do you already have a way to put it in front of buyers who are not current clients?
- Payback: At your real hourly rate, is the number of sales needed to recover the build something you can reach with the reach from question 4?
Reading the result:
- Yes to 1 and 2 only: standardize (rung 1), keep selling it as a service.
- Yes to 1, 2 and 3: productize the service (rung 2).
- Yes to all five: test a self-serve asset (rung 3).
- No to question 1: there is nothing to productize yet. Sell it again first.
When should you NOT productize?
- When the value is the judgment. If clients pay for a decision made in their specific context, standardizing the work removes what they are buying. Such work scores low on variance and client input in the Repeat Audit for good reason.
- When there is no repetition yet. Fewer than three deliveries is an anecdote. A product built on it encodes guesses.
- When demand already exceeds capacity at a good rate. Raising the price or moving clients to a retainer is faster and carries less risk than building an asset.
- When you cannot reach buyers beyond referrals. Table 7 is a volume table. Without distribution, the build hours are spent and not recovered.
- When the method has a short shelf life. A course about a tool’s current interface can be out of date before it recovers its build cost.
- When the service line is what sells everything else. The filings in Table 1 show companies deliberately keeping a low-margin services line because it wins the higher-margin sale. A freelancer’s custom work can play the same role for a package or an asset.
A base rate helps keep expectations realistic. A product is a small new venture, and new ventures often close. US Bureau of Labor Statistics data on private sector establishments show that of those opened in the year ended March 2020, 51.4% were still operating in March 2025, and of those opened in the year ended March 2015, 34.7% survived to March 2025. Those figures describe employer establishments, not digital products, so they are context, not a forecast. They argue for the cheapest possible test at each rung.
What this does not mean
- A freelancer will not earn software margins. Tables 1-3 describe large public companies with sales teams, capital and years of product investment. They show the direction of the economics, not a result to expect.
- Gross margin is not profit. Software companies spend heavily on product development and marketing below the gross margin line. A solo product has the same hidden costs in the form of unpaid build and promotion hours, which is why Table 7 counts forgone billing.
- The Ladder and the Repeat Audit are editorial frameworks. The weights and thresholds are reasoned judgments, not statistically validated instruments. Adjust them and note why.
- The break-even figures are illustrative. Replace the rate, hours, prices and fees with your own before acting on them.
Frequently asked questions
What is the difference between a productized service and a digital product?
A productized service is still delivered by you, but with a fixed scope, a name and a stated price. A digital product (template, toolkit, course) is used by the buyer without your time in each sale. On the Productization Ladder these are rungs 2 and 3.
Which should come first, a productized service or a course?
The productized service. It sells to demand you already have and frees hours quickly: in the illustrative example, a $1,500 package cut from 15 to 9 delivery hours earns $166.67 an hour. A course needs new buyers and, at $299 with a 120-hour build, 47 sales just to recover the build time.
How many times should a service be delivered before productizing it?
The framework here uses three deliveries with the same steps as the minimum for standardizing, and three more without rewriting the checklist before fixing scope and price. That threshold is an editorial rule of thumb, not a research finding.
Do software companies really lose money on professional services?
On a gross basis, three of the five companies in Table 1 did in their latest fiscal year. Salesforce reported $2,137 million of professional services revenue against $2,474 million of cost, and states that these services help secure larger subscription contracts.
How much does a platform keep from a freelance contract?
It depends on the platform and is not a simple fee. Upwork reported a marketplace take rate of 18.7% for 2025 and Fiverr 27.7%, each measured as marketplace revenue over marketplace volume, which includes fees paid by both sides.
Does productizing mean giving up custom work?
No. Custom work funds the transition and often sells the package, as services lines support subscription sales in Table 1.
Sources
- Salesforce, Inc. Annual Report on Form 10-K for the fiscal year ended 31 January 2026, filed 2 March 2026, US Securities and Exchange Commission. Results of operations and consolidated statements of operations.
- Workday, Inc. Annual Report on Form 10-K for the fiscal year ended 31 January 2026, filed 6 March 2026, US Securities and Exchange Commission.
- ServiceNow, Inc. Annual Report on Form 10-K for the fiscal year ended 31 December 2025, filed 29 January 2026, US Securities and Exchange Commission.
- HubSpot, Inc. Annual Report on Form 10-K for the fiscal year ended 31 December 2025, filed 11 February 2026, US Securities and Exchange Commission.
- Adobe Inc. Annual Report on Form 10-K for the fiscal year ended 28 November 2025, filed 15 January 2026, US Securities and Exchange Commission.
- Accenture plc. Annual Report on Form 10-K for the fiscal year ended 31 August 2025, filed 10 October 2025, US Securities and Exchange Commission.
- Upwork Inc. Annual Report on Form 10-K for the fiscal year ended 31 December 2025, filed 13 February 2026, US Securities and Exchange Commission.
- Fiverr International Ltd. Annual Report on Form 20-F for the fiscal year ended 31 December 2025, filed 12 March 2026, US Securities and Exchange Commission.
- William D. Nordhaus. Baumol’s Diseases: A Macroeconomic Perspective. NBER Working Paper 12218, National Bureau of Economic Research, May 2006. Full text read. Baumol’s 1967 American Economic Review article is cited as discussed by Nordhaus and was not opened directly.
- Mohanbir Sawhney. Putting Products into Services. Harvard Business Review, September 2016. Public preview only was readable.
- Jochen Wirtz, Martin P. Fritze, Elina Jaakkola, Katja Gelbrich and Nicole Hartley. Service Products and Productization. Journal of Business Research, 137 (2021), 411-421. Abstract only, read from the university repository record.
- Christian Kowalkowski, Heiko Gebauer and co-authors. Servitization and deservitization: Overview, concepts, and definitions. Industrial Marketing Management, 60 (2017), 4-10. Abstract only.
- Sandra Vandermerwe and Juan Rada. Servitization of business: Adding value by adding services. European Management Journal, 6(4) (1988), 314-324. Bibliographic record only; cited for the term in its title.
- US Bureau of Labor Statistics. Business Employment Dynamics, Table 7: Survival of private sector establishments by opening year, total private. Accessed 10 October 2026.
- Gumroad Help Center. Gumroad’s fees. Company-published, accessed 9 October 2026.
Tables 2, 6 and 7 are CEOtudent analyses recomputed with a script from sources 1-8 and 15 and from stated illustrative assumptions; Tables 4 and 5 and the five-question decision test are CEOtudent editorial frameworks. CEOtudent ran no survey and interviewed no freelancers. Deliberately not used: US Census Bureau Nonemployer Statistics, because its data service required an access key that was not available; Jaakkola (2011) and Harkonen and co-authors (2015) on productization, because neither the full text nor the abstract could be opened; Shopify and Atlassian filings, because they do not report a comparable professional services cost line; and any general statistic on what share of freelancers sell products or how much creators earn, because no primary source for such a figure was read.
This content was compiled with the support of AI following in-depth research, then written and prepared for publication by the CEOtudent editorial team.
