From Freelancer to Founder

I started as one of those freelancers. A laptop, a Skype account, and a billing rate I made up on a Tuesday because someone in San Francisco needed a backend by Friday. By the time BearPlex was a real company — over 100 engineers across multiple offices — I had crossed two boundaries I didn't understand were boundaries when I was crossing them.

That's the part nobody warns you about.

The independent solo trajectory looks like a ladder. Freelancer to agency to product company. Step, step, step. The framing is comforting and almost entirely wrong. It's not a ladder. It's three structurally different businesses sitting next to each other with three different operating models, three different economic logics, and three different definitions of success. The freelancer who hires three subcontractors hasn't started climbing. They've just made the first business heavier.

Most freelancers who "scale" don't build a different business. They just clone themselves.

This memo is about what the second and third businesses actually are, why most agencies plateau before they touch the third, and the concrete architecture for crossing both gaps. I'll lean on real companies and real numbers because vague advice on this topic is worse than no advice. And I'll lean on what I saw walking it from Lahore — because the offshore version of this story has been written badly, and the people writing it are usually not the ones living it.

The economics of cloning yourself

Start with the numbers, because the numbers are the trap.

A US full-time freelancer in knowledge work reported a median income of $85,000 in 2024, per Upwork's own research. A senior engineer in Pakistan or the Philippines, working the same calendar week against the same Western clients, can earn a fraction of that. Pakistan's freelance remittances doubled to $710 million in two years on the back of more than 2.3 million active freelancers. The dispersion across geographies is real. The structural problem is identical.

Time is the product. There is no leverage. Every dollar of revenue is purchased with an hour of attention.

The freelancer who decides to "scale" almost always starts the same way. They subcontract. They take on more clients than they can deliver, then bring in two friends to absorb the overflow. Now there are three of them billing hours. Revenue triples. So does coordination cost, payroll risk, client management, quality variance, and the founder's email volume.

Nothing has changed structurally. The business is still selling time. There are just more clocks.

This is the moment most operators believe they have built an agency. They have not. They have built a slightly larger version of the first business — a freelance collective with a shared logo. The economics will tell on them inside two years.

What an agency actually is (and isn't)

A real agency is a coordination machine. Its core asset is not the founder's hours. It's a system that turns generic input — a project brief, a junior hire — into reliable output without the founder touching the work.

That is a different business than freelancing. It requires different muscles. And the data shows almost nobody develops them.

Promethean Research's 2024 industry data puts 87% of North American digital agencies under fifty employees. The North American digital agency count grew from 50,000 in 2024 to over 71,000 in 2026. The market isn't undersupplied. It's a long, fat tail of small shops, and that tail does not thin out into a healthy middle. It just stops.

Why do agencies plateau at 10–30 people? Because the operating model that gets you to ten — the founder reviews everything, sells everything, recovers from every screw-up personally — is the exact thing that breaks at twenty-five. Coordination cost grows faster than headcount. Senior people spend their day in Slack instead of in client work. The founder becomes the universal bottleneck for pricing approvals, hiring filters, and client escalations. Decisions queue behind one inbox.

The deeper problem is the billing model. When you sell hours, every minute spent building a reusable thing — a delivery template, an internal tool, a documented playbook — is a minute taken away from billable revenue. Jonathan Stark has been arguing for years that hourly billing is a tax on every minute you spend building something durable. He's right. The fee structure of the agency you started actively penalizes the architecture work that would let you grow into something else.

Most agencies don't fail. They just freeze at thirty heads, throw off cash, and define ambition down to "more of the same, please."

The productized middle

The first real architectural move is not to leap from agency to SaaS. It's to compress the variance in your service offering until the work begins to behave like a product.

This is what people call a "productized service." The phrase sounds like marketing fluff. The mechanics aren't.

DesignJoy, Brett Williams' one-person design subscription, charges flat monthly fees ($4,995–$7,995) for unlimited design requests with a 48-hour turnaround. One designer. No sales calls. No proposals. No scoping. Williams runs it solo and has reported run rates of $1.7M+ ARR. ManyPixels industrialized the same idea with dedicated designers per subscriber. Bench Accounting productized small-business bookkeeping into a fixed monthly fee with software-assisted humans behind it.

The pattern across all three is the same. They picked one service. They removed every variable they could. They priced flat. They published the price. They standardized delivery to the point where a new client looks essentially identical to the last one.

The economics shift sharply. Variance is the silent killer of services margins; when every project is bespoke, you're constantly absorbing scoping errors, rework, and edge cases. Productize hard enough and gross margin starts climbing from the typical agency 30–50% range toward something that begins to resemble software.

Pricing follows a ladder, and the ladder is real:

1. Hourly — billing is reactive, ceiling is your calendar.
2. Retainer — predictable revenue, but still service-bound.
3. Productized — fixed scope, fixed price, repeatable delivery.
4. SaaS — software does the delivery; you sell access, not labor.

Each step compresses variance and expands margin. Most agencies skip from rung one to rung two and stop there forever.

From service to software

This is where the third business begins, and the canon is full of evidence that it is genuinely a third business — not a bigger agency.

37signals is the textbook case. Jason Fried founded it in 1999 as a web design consultancy. The team kept losing track of client projects in email. They built an internal tool to fix it. That tool, shipped publicly in February 2004, was Basecamp. By 2005 the tool was earning more than the consultancy, and the partners closed the services business and went all-in on product. The IP they captured by living inside their own delivery problem became the wedge.

HashiCorp tells a quieter version of the same story. Mitchell Hashimoto built Vagrant in 2010 as a side project — specifically to solve a dev-environment problem that was unbillable time inside his consulting work. He and Armon Dadgar had no real business for four years. The first commercial product was a flop. They eventually unbundled into individual tools (Vault, Terraform, Consul) and IPO'd. Origin: a piece of internal infrastructure that solved a service-delivery pain.

Help Scout was founded in 2011 as a pivot from Brightwurks, a web design consultancy run by Nick Francis and his co-founders. WP Engine was Jason Cohen's third business after founding Sheer Genius (consulting) and bootstrapping Smart Bear. Tailwind CSS is the most recent loud example: Adam Wathan extracted Refactoring UI (over $2.5M in book sales by 2020) and Tailwind UI (seven figures in year one) from his own consulting and content work. Even Pixar, before Toy Story, was Lucasfilm's Graphics Group — a hardware-and-services unit Jobs bought in 1986 and ran selling $135,000 image computers to government and medical buyers while waiting for film technology to catch up.

What changes between phase two and phase three is gross margin.

Software services typically run 30–50% gross margin. SaaS Capital's 2024 data and OPEXEngine's benchmarks put SaaS gross margins around 71–77% across public and private companies. That's the wedge. That's why the third business, when it works, is worth so much more than the second.

The valuation gap is concrete. Marketing and digital agencies typically transact at 4x–8x EBITDA, with revenue multiples in the 0.5x–2.5x range. Private SaaS companies were trading in the 6–8x ARR range through 2024, with premium vertical SaaS hitting 7–9x or higher. A $10M-revenue agency and a $10M-ARR SaaS company are not in the same valuation universe. They are not in the same universe at all.

The architecture, concretely

If you take seriously that you are building a different business and not extending the current one, here is the architecture that actually works.

Productize one service first. Not a portfolio. One. Pick the wedge with the highest variance-to-value ratio in your current book — the engagement type where clients pay the most for the least bespoke work. Strip every option. Fix the price. Publish it. The first month of doing this honestly will reveal exactly which of your "custom" engagements were actually a fear of telling clients no.

Build a delivery operating system. Every recurring decision in the firm becomes a written rule. Onboarding checklists. Pricing trees. Scope-change scripts. Internal tooling for the steps you do over and over. The test is brutal and clarifying: the founder takes a two-week vacation. Does the system run? If the answer is "things would go to hell," the system doesn't exist yet.

Capture IP from every engagement. This is the single most underused move in the services world. Whatever you're building for a client this quarter is data about what should be a product. Document patterns. Extract templates. Build internal libraries. Notice when the third client in a row asks for the same module. The third occurrence is not a coincidence — it's a product spec written in client money.

Hire system-builders, not just service-deliverers. A senior engineer who only ships client work is not the same hire as a senior engineer who instinctively documents, extracts, and abstracts. The first scales linearly. The second scales the firm. Most agencies hire the first kind exclusively and then wonder why the org doesn't compound.

Walk the pricing ladder deliberately. Hourly to retainer to productized to SaaS is not just a billing change. It's an org design change at every step. Productized service requires marketing competence the retainer agency doesn't need. SaaS requires product, infra, and support competence the productized service doesn't need. Each rung is a new hiring profile.

The offshore studio question

This is where my own walk gets specific, and where the standard VC essay tends to get the story wrong.

If you build a services business in Lahore, Manila, Cairo, or Lagos, you have two structural advantages and one structural trap. The advantages are real: South and Southeast Asian developer rates run $15–$30/hour versus Western markets at three to ten times that, and the talent pool is enormous and growing.

The trap: those advantages are an arbitrage, not a moat. An offshore agency is a bet that the gap between what your market pays and what your delivery costs will stay open. That gap is closing every year as remote work normalizes global rates.

The honest thing to look at is what actually compounded out of the offshore world.

Toptal was founded in 2010 by Taso Du Val and Breanden Beneschott on the observation that engineers Du Val knew abroad were as good as his Silicon Valley peers and that nobody had built a serious filtering layer on top of that supply. Toptal didn't try to be the cheapest offshore agency. It built a vetting system, a matching engine, and a brand — three things an arbitrage trade can't build — and operated profitably without raising follow-on capital after its $1.4M seed.

Andela ran the more painful version. It started as an African engineering fellowship in 2014, scaled to over 1,000 fellows by 2018, then pivoted in 2019 to mid-and-senior global talent, laid off about 400 junior engineers, went fully remote in 2020, and opened up to Latin America and Eastern Europe. Brutal sequence. The pivot was correct. Junior-fellowship-as-product was a service shape; senior-talent-as-marketplace was a product shape. The honest version of the offshore story is that the companies that made it through were the ones that stopped being "the cheap version of a Western agency."

The lesson I took out of Lahore is simple and unromantic. Labor cost arbitrage gets you started. It does not get you a company. The agencies in our region that actually graduate to durable businesses do the same things the Boston and Austin agencies do, just with cheaper inputs at the bottom: they productize a wedge, document the delivery, build IP into every engagement, and hire system-builders. The geography is incidental to the architecture. People who confuse the two end up running shrinking arbitrage shops and blaming the macro.

When you should stay an agency

Worth saying out loud: not every agency should become SaaS, and the productized-service-to-product playbook is wrong for some firms.

Paul Jarvis made the case in Company of One for the small expert firm — high gross margin, low headcount, high owner cash flow, intentional ceiling. The math is real. A two-to-five-person specialist firm billing senior rates can throw off more cash to the founder than a thirty-person agency, with a fraction of the management load. The economics are good. The lifestyle is good. The exit is small but it exists.

Productized-service transitions also fail in three predictable ways. First, when the team picks the wrong wedge — productizing a service whose value actually lives in the bespoke variance you just stripped out. Second, when the founder confuses "I personally do this work fast" with "this is a repeatable delivery system." Third, in industries where customers genuinely buy expert judgment — heavy regulatory, complex litigation, deep custom-implementation work — productization narrows the offering past the point of usefulness.

SaaS is wrong, often. Industries where the buyer's problem is configuration, integration, and ongoing judgment — not a piece of software — punish founders who try to package judgment into a SKU. The SaaS pattern matches generic problems with generic solutions. Specialist firms exist because the problems aren't generic.

The honest test is simple. If the variance in your engagements comes from your client's industry, productize. If the variance comes from the unique judgment of the partners, stay an expert firm and price accordingly.

The third business

The walk from freelancer to founder is not a ladder you climb by working harder. It's a sequence of architectural decisions, each one designed to make the founder less central to the next phase.

The first business is your time. The second business is your team. The third business is your IP.

The companies that make it across both gaps — 37signals, HashiCorp, Help Scout, WP Engine, Tailwind, the Pixar arc — share a single move. They took something they built for clients, recognized it as a product, and did the painful work of separating it from the consultancy that birthed it. They didn't keep doing more of the second business. They stopped doing the second business.

The agencies that don't make it across don't usually fail. They freeze. The founder becomes the universal pricing approval, the universal escalation path, the universal hiring filter. Headcount stalls at twenty-five. Revenue plateaus. The shop becomes a personal annuity that pays well and grows slowly forever.

The architecture of the third business has nothing to do with the architecture of the first two. Anyone telling you otherwise is selling you the comforting ladder metaphor.

The harder, more honest framing: pick which business you actually want to build, and stop pretending the next phase is a continuation of the last one.