Zapier vs Make vs n8n vs scripts
A decision guide for automation tools. Where each shines, where each costs more than it should, and when a 30-line script wins.
Most "best automation tool" articles compare features. Triggers, actions, integrations, pricing tiers, side by side in a grid. The actual deciding question is none of those. It's this: who's running this thing in 6 months when it breaks at 2am and the CEO is asking why a deal didn't sync to Salesforce?
Each of these four tools has a different "future you" they're optimized for. Pick the one that matches the person who'll inherit the mess.
The four, briefly
Zapier is the fastest to set up and the most expensive to scale. It's the default for "I just need this thing to talk to that thing." 7000+ integrations, a decent UI your non-technical cofounder can poke at, and a billing model that punishes growth. Pricing is per-task and adds up fast once a workflow gets popular.
Make (formerly Integromat) is more powerful, with a visual flow editor that handles complex branching better than Zapier. Cheaper at higher volumes. Steeper learning curve, especially around iterators, aggregators, and the 12 different ways data can flow between modules. Once you get it, you don't go back.
n8n is open-source, self-hostable, fair-code license. Most flexible, most setup. Free if you self-host on a $5 VPS, pay-per-execution if you let them host it. Has a real code node, so you can drop a JavaScript snippet into the middle of a flow when the visual blocks aren't doing what you need.
Custom scripts is the option people forget exists. A 50-line Vercel cron, a Cloudflare Worker, a tiny Python script on a Raspberry Pi. Free at any volume. Hardest to maintain in theory, easiest to debug in practice, because the whole thing is in one file you can read in 90 seconds.
The decision matrix
| If... | Pick |
|---|---|
| You've never written code | Zapier |
| You want flexibility but no code | Make |
| You want self-hosted, full control | n8n |
| The integration is complex, custom, or high-volume | A script |
| You're a team that'll outlive this | Whatever your most-technical team member can maintain |
The last row is the one most people miss. Every automation tool is a hidden hire. Whoever maintains it needs to actually want to.
When Zapier wins despite the price
Any time the alternative is you fighting with Make's UI for 2 hours to save $30 a month. Zapier costs $40 and saves you $400 in your time. If a Zap takes 15 minutes to build and never breaks, that's the right call even if a script would have been "free." Your hours have a real number attached to them.
When Make wins
Complex branching ("if X and Y but not Z, route to this handler"), iterators that loop over arrays of line items, scenarios with 10+ steps. Visual debugging is genuinely better than Zapier's, because you can see the data at every step and rerun from the failure point. If your automation has ever hit Zapier's "Paths" feature and felt like you were wrestling with a hose, Make is your answer.
When n8n wins
Privacy-sensitive data you don't want shipping through someone else's servers (you self-host). High volume where per-execution pricing would eat you alive (no per-execution cost on self-hosted). Teams with DevOps muscle who already run a few small services and one more is no big deal. Worst pick if no one on the team wants to maintain a server. The free tier is only free if your time is also free.
When a script wins
You only need this for ONE thing. You're already writing code anyway, so a Cloudflare Worker is just another file in a repo you push to weekly. You want full control over retries, logging, and error handling. The bar to writing a script is much lower than people think with Cursor or Claude Code in the loop. A "watch Stripe webhooks and add the customer to HubSpot" script is 40 lines of TypeScript you can spec in plain English and have running in 20 minutes. The hard part used to be writing it. That's gone.
The cost reality at 100k events/month
- Zapier: ~$700-1000/month
- Make: ~$50-150/month
- n8n cloud: ~$50-100/month
- n8n self-hosted: ~$10/month for a tiny VPS
- Custom script: cents on Vercel or Cloudflare
The spread is real. At 1k events/month nobody cares. At 100k+, the wrong pick is a line item your CFO will eventually notice.
The "you don't need this" test
Before picking any automation tool, run this check: is the manual process really 10 minutes a week? If yes, do it manually for 3 more months until you're sure the workflow isn't going to be deprecated, restructured, or replaced by a feature in one of the tools you already pay for.
Half of automation projects automate work that ends up being deprecated. The other half automate work that's actually 2 minutes a week and the automation breaks 4 minutes a week. Be honest about which one you're building.
The right move is often "leave it manual, set a calendar reminder, revisit in a quarter." The tool you don't deploy can't break.
Up next
Once you've picked a tool, the question becomes which automations are worth your weekend to build. Next guide, "Personal automations worth the setup time," covers the handful of flows that actually pay back the hours you spend setting them up, and the ones that look great on a Twitter thread but cost more than they save.
Next in this pillar
Personal automations worth the setup time