Should an SMB Hire Its First In-House Engineer? A True-Cost Table for Three Scenarios

1. The scene: a NT$65,000 offer letter stuck on the GM's desk for three weeks

A 48-person industrial parts trading firm runs orders on Excel, complaints on LINE, and quotes hand-copied from Email. It spent roughly NT$960,000 on outsourced builds last year. HR just closed a full-stack engineer with three years' experience at NT$65,000/month. The GM's arithmetic: 65,000 × 14 months = NT$910,000, "cheaper than outsourcing, and he's here whenever I need him." That arithmetic is wrong, and it is wrong in a predictable way. This article breaks the true cost of a first engineer into numbers you can fill in, and gives you one reusable decision formula.

2. Five industry myths worth breaking first

• Myth 1: headcount cost equals salary × 14. Under Taiwan's Labor Standards Act and Bureau of Labor Insurance rules, the employer carries labor insurance, employment insurance, pension contributions and the employer share of National Health Insurance—statutory items alone run about 17–20% of salary. A NT$65,000 salary costs roughly NT$77,000.
• Myth 2: an in-house person produces at 100%. A first engineer has no team, no code review, no on-call rotation. Meetings, ad-hoc IT chores and teaching-mode communication consume 30–40% of hours. Real product hours land at 95–115 per month.
• Myth 3: hiring ends the risk. A solo technical role carries the highest turnover exposure. One departure costs recruiting plus gap plus ramp—2.5 to 3.5 months of salary equivalent at SMB scale.
• Myth 4: one hire covers everything. No single person is backend, frontend, mobile, design, security and DevOps at once. Hand a multi-surface roadmap to one person and delivery serializes: three things finish in a queue, not in parallel.
• Myth 5: outsourcing costs more because of margin. The margin is real, but you are buying amortized tooling, existing components and multi-person skill coverage. The comparable unit is not monthly price—it is the effective hour rate below.

3. The core formula: Effective Output Hour Rate (EOHR)

Put both options into the same unit before comparing.

• EOHR = annual total cost of ownership ÷ annual effective output hours
• In-house TCO = salary × 14 × 1.19 (statutory burden) + equipment and licenses + amortized recruiting + management hours + turnover reserve
• In-house effective hours = 12 × monthly effective hours (use 105 conservatively)
• Outsourced TCO = contract value + internal coordinator hours (about 6–10 hours/month)
• Outsourced effective hours = contracted hours, since meetings and management are already priced in

Decision rule: compute both, then apply three gates. If (a) outsourced EOHR ÷ in-house EOHR ≤ 1.3, and (b) stable annual demand < 1,000 hours, and (c) technical surfaces ≥ 3 → outsource. If (a) > 1.6, annual demand > 1,400 hours, and surfaces ≤ 2 → hire. Anything in between → hybrid: one core hire plus outsourced coverage.

4. Three scenarios compared

The deciding variable is not company size. It is the intersection of stable annual demand hours and number of technical surfaces. Company B should not rush to hire precisely because its demand spans four surfaces, and one person would serialize every deadline.

5. The full hidden-cost list (baseline: NT$65,000/month)

• Statutory employer burden: ~19% of salary, about NT$173,000/year
• Recruiting: job-board and headhunting fees. Per market rates on 104 Job Bank, technical headhunting runs 15–20% of annual salary, roughly NT$140,000–180,000 per placement; in-house recruiting still burns 35–50 HR and manager hours
• Equipment and licenses: laptop NT$60,000 amortized over three years, plus IDE and cloud services around NT$35,000/year
• Management hours: three hours a week of manager alignment, 156 hours/year at NT$900/hour ≈ NT$140,000
• Ramp-up loss: ~50% output in the first three months, roughly NT$115,000 of value
• Single-point-of-failure reserve: replacement cost of 2.5–3.5 months' salary, amortized at about NT$60,000/year
• Opportunity cost: serialized delivery pushes the second and third initiatives out by 4–6 months
• Total: nominal NT$910,000 → true annual cost of ownership about NT$1,510,000, a 66% gap

6. A 10-dimension scorecard for evaluating an outsourcing partner (0–5 each)

Total 50. Score ≥ 40: safe for a long-term relationship. 30–39: single projects only. Below 30: keep looking.

7. How ScriptWalker's four engagement models map—and who we are wrong for

• Project: defined scope, one-time build. Fits scenario A. From NT$180,000.
• Retainer: steady but scattered demand, a fixed monthly hour pool. Fits scenario B. NT$45,000/month for 40 hours.
• Advisory: you have decided to hire, and need someone to set up specs, architecture and acceptance criteria. NT$20,000/month for 8 hours.
• Full outsourcing: the whole digital workflow, including operations and on-call.

Please do not hire us if:

• You need physical IT support on site within four hours—swapping drives, pulling cable. We have no deskside staff.
• You want a single cheapest lump sum and will not accept modular quotes or change orders.
• Your decision chain runs more than three layers deep with no single point of contact. Meeting overhead will eat the hours you paid for.
• Requirements are still being invented and you expect three or more pivots within three months. Start with Advisory, not a project.
• You want us to operate a ten-year-old undocumented system with no original author and no permission to refactor.

8. The 90-day startup playbook

• Days 1–14: inventory current tools and data flows, produce a system map and a demand-hour estimate, and lock both EOHR figures.
• Days 15–30: ship the smallest, most painful workflow first—automated quote generation, for example—and establish the acceptance cadence and change-order format.
• Days 31–60: extend to the second and third workflows while building the documentation library and the account-ownership register. Domain, cloud and database all sit in the client's name.
• Days 61–90: compare actual hours against the estimate and recompute EOHR. If outsourced hours exceed 110/month for three consecutive months, open the hiring evaluation.
• Day 90 review gate: three metrics—on-time delivery ≥ 85%, average change-order response < 2 business days, internal coordinator time < 10 hours/month.

9. Decision checklist: check 9 or more before you hire

• ☐ Stable annual technical demand of 1,400 hours or more
• ☐ Technical surfaces concentrated in two or fewer areas
• ☐ Someone internally can write an acceptable specification
• ☐ A manager can give three hours a week to technical alignment
• ☐ The 19% statutory burden is budgeted
• ☐ Equipment and cloud licensing of ~NT$95,000/year is budgeted
• ☐ You accept roughly 50% output in the first three months
• ☐ You have a fallback if the person leaves (docs or external support)
• ☐ You can offer a technical career path
• ☐ The salary band is at market, not below it
• ☐ A 4–6 month delay on secondary projects will not miss revenue targets
• ☐ No headcount reduction is planned in the next 18 months
• ☐ Internal policy exists for source code and account custody

10. FAQ

Q1: We already have a candidate. Is it too late to recompute?

No, and you should. Filling in both sides of EOHR takes about 90 minutes. The usual outcome is "hire anyway, but rewrite the job description"—narrow from full-stack to backend and data, and outsource the remaining surfaces.

Q2: Won't outsourcing lock us in?

That depends entirely on contract terms and account ownership. The source repository, domain and cloud root account should be registered to the client on day one, with the vendor joining as an invited collaborator. Do that and switching vendors costs only a ramp-up curve.

Q3: In a hybrid setup, how do we split work without collisions?

Split by system boundary, never by task. In-house owns the core modules that evolve for years; the vendor owns well-bounded integrations, frontend work and cyclical projects. The seam between them is a single API contract document.

Q4: We are very small. Should we invest at all?

Small companies should invest in data consolidation rather than headcount. Pull the data scattered across LINE, Excel and Email into one source of truth first; a one-time project of NT$180,000–350,000 usually lowers the cost of every decision that follows.

Q5: How do we know outsourced hours are not padded?

Price by acceptable increment rather than raw hours, and put a weekly deliverable list in the contract. If a timesheet cannot be mapped to something demonstrable, that is your warning sign.

11. Next step

We offer a free 30-minute diagnostic: bring your annual demand list and the salary band you are considering, and we will compute both EOHR figures with you on the call. If the answer is "hire," we will say so plainly, and can help you write the interview questions and spec templates for the role.

• Email: [email protected]
• Phone: 0916-224-047
• LINE: @ufv9089p