Agency or SaaS: the same million, opposite games

Agencies get to revenue fast and stall on hiring, SaaS gets there slowly and then compounds — what the data says about time to $1M, agency margins, and which hard problem you are choosing

A post went around recently comparing the two paths to $1M a year, written by someone who had done both. The summary: agencies are much easier to start and brutal to scale past low seven figures, where success comes down to sales skill and hiring; SaaS is much harder to start, needs more luck, involves far fewer meetings, and gives you a better life if you can pull it off.

Most of that holds up. One part of it does not, and the part that survives is more interesting than the summary suggests.

The slow start is real, and it is slower than people admit

ChartMogul's Against the Odds: The 2025 SaaS Growth Report looked at 6,525 software companies with ten or more years of history. Only 3.3% reached $1M ARR within a year of first revenue. About 13% got there within three years, and 25% within five. Roughly half made it to $1M at all within a decade, one in ten reached $10M, and one in fifty reached $25M.

Read that as a distribution rather than a scoreboard. The median founder building software is not in a slow patch that better execution would fix. They are living the normal case. Two to five years is the shape of the curve, and the plans that assume otherwise are usually built from the 3.3%.

Meanwhile an agency can invoice in week two. That asymmetry is the whole argument for consulting, and it is a good one: you can sell work you have not built yet, and the buyer of an outcome does not need a product to exist before they pay for it.

Agency revenue is bought with headcount

Where the agency path gets expensive is the second million. Promethean Research's 2026 State of Digital Services surveyed 119 agency leaders and found an average agency of 31 employees, average after-tax net margin of 13%, and average industry growth of 7.5% in 2025. The same data puts revenue per full-time employee at around $163K.

Those three numbers together describe the ceiling precisely. If a head produces roughly $163K, then $1M of agency revenue is about six or seven people, and $3M is about twenty. Growth is not a strategy question at that point, it is a recruiting and management question, which is exactly the claim in the original post: past low seven figures it comes down to hiring. The 13% net margin is the other half of the story. A million in agency revenue is not a million in your pocket; it is a payroll with a modest remainder, and the remainder is what has to absorb a client leaving.

Software has the opposite shape. Revenue per employee is not bounded by hours, which is why the same $1M can be one person and why the second million costs so much less than the first.

The retention claim is the weak one

The original post puts average agency retention at 6 to 12 months. That is far below what agency benchmarks report. Focus Digital's 2026 churn analysis gives average client lifespans of 56 months for retainer-based agencies and 24 months for project-based ones, with hybrid and performance models in between. Treat those numbers as indicative rather than precise — the report does not disclose its sample size or methodology — but the ordering is the point, and it is consistent across sources.

Six to twelve months is not what agency retention looks like. It is what project work looks like when it is sold as a retainer. If your clients leave inside a year, the problem is usually that you sold a deliverable rather than an ongoing outcome, and no amount of extra sales skill fixes that — it just refills a leaking bucket faster.

This matters for the comparison, because short retention is what makes agency work feel like a treadmill. Fix retention and the agency path looks considerably more like a business and less like a job with better numbers.

What you are actually choosing

Strip away the lifestyle talk and the two models differ on one axis: where distribution lives.

In an agency, distribution is a conversation. You can generate demand by talking to people, one at a time, and each conversation has a large enough contract value to justify the hour. That is why sales skill dominates: the offer is rarely unique, so the win goes to whoever is better at the meeting.

In SaaS, no single conversation is worth the hour. At $30 a month, you cannot sell your way to a million one call at a time, so distribution has to happen without you present — through search, community presence, content, word of mouth, a channel that keeps working while you sleep. The thing people call "luck" in software outcomes is mostly this: whether one of those unattended channels caught.

So the honest framing is not easy versus hard. It is: do you want the problem that is solved by showing up, or the problem that is solved by building something that works when you don't?

The lifestyle claim, with one correction

Fewer meetings is true and it is a real quality-of-life difference. Lower stress is not automatic. Agency stress is acute and legible: a client is unhappy, a deadline is Friday, someone quit. Software stress is diffuse — the numbers are flat, nobody is angry, and you have no idea whether the last two months of work mattered. Plenty of people find the second kind harder, and it is worth knowing that about yourself before choosing on the basis of a meeting count.

The path most people actually take

The two are not exclusive, and the sequence that works is well worn: run services to fund the runway, but pick clients in one narrow niche and build the product out of the work you keep repeating. You get paid to do the customer research, you learn where the job is genuinely hard, and you start with users who already trust you.

The trap is the split-attention version — a bit of consulting, a bit of product, neither getting the years it needs. The version that works treats the agency as deliberately temporary and time-boxed, with the niche chosen for what it teaches rather than what it pays.

Whichever way you go, the constraint is the same one: the software path needs distribution that runs without you in the room. That is the part that takes the two to five years, and the part most builders leave until the product is "ready". If you want a concrete plan for where your first users actually are, OctoLoops does that research and drafts the outreach, and you keep the judgment about what to send.