Build your automation or buy SaaS: three-year cost, worked out

"Can't you just wire it up in Zapier?"

"At your volume, Zapier will hit the ceiling."

Both are right, depending on the case. This isn't an argument about taste — it's arithmetic plus boundaries.

Short version: buy SaaS for simple rules, build for complex logic or sensitive data. The threshold is in three places.

---

1. Define the job

For comparability, assume this:

Daily: scrape 3 sources → clean and deduplicate → write to a sheet → notify on failure

That's the most common shape and the one where the two options diverge most.

---

2. What SaaS actually costs

Public pricing at time of writing (check the vendor — it changes):

Platform	Entry tier	Quota	Fits
Make	$9/mo	10,000 ops	Solo / small team, cheapest
Zapier	$19.99/mo	750 tasks	Largest ecosystem, highest unit cost
n8n Cloud	€20/mo	2,500 executions	Self-host friendly

Make       $9  × 36 = $324   ≈ ¥2,300
Zapier     $20 × 36 = $720   ≈ ¥5,200
n8n Cloud  €20 × 36 = €720   ≈ ¥5,600

None of that looks expensive. Three hidden costs:

1. Quotas fill fast. "Three sources a day" sounds small, but one source may be 5–10 operations (fetch, parse, clean, write). Three sources × 8 ops × 30 days = 720/month. Add two sources or retry logic and you're at the ceiling — and the next tier is 3–5× the price.

2. Complex logic hits platform limits. SaaS handles linear A → B → C well. Layered conditionals, custom cleaning (regex, dedup algorithms), failure compensation — the flow diagram fills with branches and becomes harder to maintain than code.

3. Data must leave your network. Orders, customer data, financial data through a third party fails many compliance reviews.

---

3. What building actually costs

Item	Effort	Price
Process mapping	1 day	included
Scraping	2–3 days
Cleaning / dedup	1–2 days
Scheduling + alerts	1 day
Testing + docs	1–2 days
Total	7–10 days	from ¥6,000

Ongoing: VPS ¥30–100/month, plus maintenance (2–3 adaptations a year, roughly ¥200/month amortised).

Year 1:      ¥6,000 + ¥600 + ¥2,400 = ¥9,000
Years 2–3:   ¥600 + ¥2,400          = ¥3,000/year
Three years: ¥15,000

Without maintenance, three years is ¥6,000 + ¥1,800 = ¥7,800.

---

4. Side by side

	Make	Zapier	Build (no maintenance)	Build (with maintenance)
Year 1	¥780	¥1,730	¥8,400	¥9,600
Three years	¥2,300	¥5,200	¥7,800	¥15,000
Data leaves network	Yes	Yes	No	No
Logic ceiling	Medium	Medium	None	None
Who you call when it breaks	Vendor	Vendor	Yourself / the developer	Yourself / the developer

Short term, SaaS is less than half the cost. Self-hosting only wins on no-maintenance plus long use.

---

5. Three thresholds

Quota

Check your actual 30-day consumption:

• Under 50% of the plan: room to grow, use SaaS
• Over 80%: the next tier is 3–5× the price — do the build maths
• Already over the limit once: build it

Branch complexity

Count the "ifs":

• 0–3: SaaS is fine
• 4–6: it works but maintenance hurts
• Over 6: write code

Rule of thumb: if you need comments in a flow diagram to understand what you built, you should be writing code.

Data sensitivity

If the words order, customer or financial appear anywhere — build, or at minimum verify the vendor's data processing terms.

---

6. There's a third option

SaaS for orchestration, code for logic.

Scripts do the custom cleaning; Make or Zapier handles the trigger and notification. You get testable code where it matters and don't maintain a scheduler.

That's what I run here: scraping and cleaning as scripts, triggering on a timer, notifications by email.

---

7. When not to automate at all

The most important section.

If the task:

• Takes under 15 minutes a day
• Happens a few times a month
• Changes its rules constantly

Don't automate it. The build cost covers years of doing it by hand.

Test: time saved per day × 250 working days > build cost + annual maintenance.

30 min/day × 250 = 125 hours/year
At ¥100/hour = ¥12,500/year
Build ¥6,000 + maintenance ¥2,400 = ¥8,400 year one
→ Worth it

5 min/day × 250 = 20.8 hours/year
At ¥100/hour = ¥2,080/year
Build ¥8,400
→ Not worth it; use SaaS or keep doing it manually

---

Bottom line

Validate the need with SaaS first; build only once it's a long-term constant. Building immediately either over-invests or produces a script nobody uses.

The first step of the automation service is process mapping — and if it shouldn't be automated, I'll say so.