First Nine Months of CollabOps — *23 Decisions* From Zero to First Customer
Nine months of founding. People, money, product, market — 23 decisions. *Seven* clearly right, *six* clearly wrong, *the rest* still unknown. An honest record.
I wrote down the decisions of CollabOps's first nine months. Came out to 23. From here, seven look clearly right, six clearly wrong, and the remaining ten still don't have an answer.
This post is the honest taxonomy. A raw record of what kinds of decisions fill a startup founder's first nine months.
Seven decisions clearly right
1. Picking on-prem / air-gapped market as the ICP
The decision between global SaaS DevOps vs Korean / Japanese / EU air-gapped. The latter looked smaller, but matched our technical strength and capital efficiency. → Why we default to self-hosted
2. Technical PhD as the second hire
More valuable in the first six months than sales / product / ops. 80% of our first PoC passes hinged on his depth.
3. Blog as the first sales asset
Cold email, ads, BD outsourcing — all inefficient. Blog → meeting → PoC was the most efficient path (→ Enterprise sales without BD).
4. PoC as the CTA
Pitched a 2-week PoC instead of a demo after the first meeting. Higher entry bar. The people who entered were serious. 17% conversion.
5. Self-hosted-first
Did not start SaaS-first. Adding self-hosted six months later is nearly impossible. The first decision creates the option for the next five years.
6. Korean + English from the start
Started Korean only. Three months in we added English. English traffic is 5× Korean — global LLM answers are predominantly English.
7. Permission model on a graph from day one
Explicitly rejected the option of RBAC now, graph later. Graph from day one. Six months later, AI agent integration required zero additional work. → Permission model on a graph
Six decisions clearly wrong
1. Trying cold email campaigns
300 sent, 1 meeting, 0 contracts. Three months wasted.
2. Outsourcing BD
Three months. 17 MQLs. Zero matched our ICP.
3. One-time conference booth
$12,000. 47 cards. 1 meeting. Booth < talk.
4. Advertising SaaS as an option in the first six months
Self-hosted is default — but we put "also available as SaaS" on sales material. Diluted the message. Switched to self-hosted only in marketing six months in.
5. Hiring the first designer too late
Designer joined six months after the first engineer. During those six months, weak UI got us rejected from two customer meetings. Should have hired earlier.
6. Too many features — eight at first launch
Lost focus. Took three additional months to cut down to four. Should have started with four.
Ten decisions still unanswered
Made now, but right or wrong still unknown. The answers will come at 12–24 months.
- First fundraising timing — now vs six months later
- FIPS adoption — six months for one customer → FIPS compliance cost
- Eight-month AI agent beta — held the launch
- Monorepo vs polyrepo — chose monorepo
- Built our own workflow expression engine → Why we built our own expression engine
- Sales:engineering ratio — held at 1:5
- Global vs Korea-first market attack — Korea + Japan
- Open-sourcing parts — kept core closed, expression engine open
- Quarterly conference talk cadence — sustainability unknown
- Splitting the blog across team bylines vs a single founder voice — eventually consolidated to the latter (this very post is the result)
Nine months in one line
Whether 23 decisions are right or wrong depends on the vantage point of right now. A year from now the taxonomy may shift.
What's certain — there were 23 decisions. Nine months = 23 decisions = 0.6 per week. Sounds small, but each decision shapes the environment for the next. Compounding.
Should you copy these 23 decisions?
No — CollabOps's 23 decisions over nine months are reference material, not answers that transplant to another company. If you're a founder starting from zero and standing at a similar fork on any of the 23, our decision + nine-month evaluation is worth reading. But our ICP doesn't guarantee correctness for yours — simulate against your own first.
Next entry
Plan to rewrite this post nine months from now (2026-12). By then, the unanswered ten will have settled into right or wrong.
Related posts
After Leaving Google, My Definition of Good Work Changed
I was a software engineer at Salesforce and Google before founding a company in July 2024. The standard for good code stayed the same. What changed is that the questions in front of the code became mine: which customer problem, and why now. Notes on how the definition of good work shifted from engineer to founder.
John Baek
Will Hiring More Senior Engineers Actually Make the Team Faster?
Expecting a team to speed up after hiring experienced engineers is natural. But there is something to check before opening the role: can you describe why work is slow as a concrete scene? Notes on separating work time from elapsed time, writing down the expected outcome, and how to judge a senior hire.
John Baek
Korea Had PC Bangs. What Kind of Environment Do Developers Need?
A question that started in a StarCraft lobby, where saying I was Korean was enough to make opponents leave. Where does skill actually come from? Move the practice-and-feedback loop of a PC bang into a developer day and you get an environment where you can see what happens after every change.
John Baek