Nobody's automation bill jumps because they added more automations.
It jumps because they added steps to workflows that were already running — a filter here, an enrichment lookup there, a second notification — and the billing model charges for every one of them, on every run.
That's the mechanic behind almost every "why did this suddenly cost so much" moment, and it's entirely predictable once you understand what you're actually buying.
The short version
Three billing models:
| Model | Charges for | Grows with |
|---|---|---|
| Per task (Zapier) | Each action step | Steps × runs |
| Per operation (Make) | Each module run | Modules × runs |
| Per execution (n8n) | Each workflow run | Runs only |
The consequence: under the first two, making a workflow more sophisticated increases its cost. Under the third, it doesn't.
And the workflows worth building are the sophisticated ones — which is why the gap widens in exactly the direction you'll naturally develop.
How tasks are counted
The detail that surprises people, and it's worth getting concrete.
A "task" is an action step, not a workflow run.
A workflow with a trigger and four actions, running 500 times a month, consumes 2,000 tasks — not 500.
What else counts that people don't expect:
Filters and conditions consume in most models, even when they reject the run. The evaluation happened.
Each branch path has its own steps, so a workflow with two paths that reconverge often duplicates the tail rather than sharing it.
Lookups and searches are steps. Checking whether a record exists before creating it doubles that portion.
Formatting and data transformation steps count. Splitting a name into first and last is a step.
The pattern: the things that make a workflow good — validation, conditional handling, error paths, enrichment — are all steps, and all billable.
The polling trap
Specific to operation-based billing, and worth its own section because it's invisible.
Polling triggers consume operations even when they find nothing.
A scenario checking a spreadsheet every fifteen minutes runs 96 times a day. If nothing changed, it still consumed 96 operations — around 2,900 a month, producing zero output.
How to avoid it: use webhooks wherever the source tool supports them. A webhook fires only when something actually happens, so consumption matches real events rather than clock time.
Where it bites hardest: monitoring workflows. "Check for new rows," "watch for status changes," "look for new files." These feel cheap and run constantly.
Worth auditing specifically. If your consumption is high and your workflow count is low, polling is the first place to look.
Why the bill jumps
Four causes, in rough order of frequency.
1 — You added steps to existing workflows. The most common by a distance. Workflow count unchanged, consumption up sharply, and nobody connects the two.
2 — Volume increased. More form submissions, more orders, more triggers. Genuine growth, and the cost is proportional.
3 — A loop. Two workflows triggering each other, or one that updates a record it also watches. This produces runaway consumption and it's usually obvious in the run history once you look.
4 — A polling trigger you forgot about. Running every fifteen minutes for eight months, producing nothing.
The diagnostic: open your platform's task or operation usage report and sort by consumption. It's rarely the workflow you'd guess. The full audit.
When self-hosting starts making sense
The threshold question, answered honestly rather than promotionally.
The saving is real. Per-execution billing with no execution limits removes volume as a cost variable entirely. At high volume, the difference between that and per-step billing is substantial.
The costs that offset it:
Setup. A container on a small server is straightforward if you've done it before and a genuine barrier if you haven't.
Ongoing maintenance. Updates, backups, security patches, monitoring. Recurring, not one-off.
Debugging without support. Community forums and documentation rather than a support desk.
Key-person risk. If one individual maintains it and leaves, you own infrastructure nobody understands. This is the cost nobody puts in the comparison and it's frequently the largest.
The honest threshold — two conditions, both required:
Your spend is high enough that the annual saving meaningfully exceeds the time you'd spend maintaining it.
Someone on the team is genuinely comfortable running a server, and ideally more than one person.
If either fails, the sticker-price comparison is misleading. A cheaper platform that stops working when your one technical person is on holiday isn't cheaper.
Reducing consumption without switching
Often the better answer, and it's overlooked because it's less interesting than a migration.
Use webhooks instead of polling. Frequently the single biggest reduction available.
Consolidate steps. Two formatting steps doing related work can often be one. Under per-step billing, that's a direct saving on every run.
Filter as early as possible. If a workflow only needs to act on 10% of triggers, filtering at step one means the remaining 90% consume one step rather than six.
Check for native integrations. If your CRM and email platform already talk to each other directly, you're paying a connector to duplicate something free. Worth checking first.
Delete what isn't earning. The audit usually finds workflows with zero runs, duplicates, and things built for a campaign that ended.
Move the highest-volume workflow only. If one workflow accounts for most of your consumption, that's the one to rebuild on a cheaper platform. Leave the rest — a partial migration captures most of the saving for a fraction of the effort.
The calculation that actually matters
Not cost per task. Cost against value.
Time saved per run × runs
− platform cost
− build time (amortised)
− maintenance time
− cost of errors
= actual net
Maintenance is the term people omit, and it's rarely small — APIs change, fields get renamed, tokens expire.
Which means a rising bill isn't automatically a problem. If your automation spend doubled because volume doubled and the workflows are saving proportionally more time, that's working as intended.
It's a problem when consumption grows faster than value — which usually means step-heavy workflows, polling you forgot about, or automations still running for processes that changed. Measuring it properly.
Frequently asked questions
Why did my Zapier bill jump?
Almost always because steps were added to workflows that were already running, rather than because you built more automations. Task billing charges per action step, so adding a filter, a lookup, and a notification to a workflow running 500 times a month adds 1,500 tasks without changing your workflow count. The other common causes are a genuine volume increase, a loop where two workflows trigger each other, and a polling trigger you'd forgotten about. Sort your usage report by consumption — it's rarely the workflow you'd guess.
How are tasks counted?
Per action step, per run — not per workflow. A trigger plus four actions running 500 times consumes 2,000 tasks. What surprises people is what else counts: filters and conditions consume even when they reject the run, each branch path has its own steps, lookups and searches are steps, and formatting operations are steps. The things that make a workflow reliable — validation, conditional handling, error paths — are all billable, which is why sophistication is directly taxed under this model.
When is self-hosting cheaper?
When two conditions both hold: your spend is high enough that the annual saving meaningfully exceeds the maintenance time, and someone on the team is genuinely comfortable running a server — ideally more than one person. Self-hosting removes per-step billing entirely, which is a substantial saving at volume. What it adds is setup, updates, backups, patching, monitoring, and debugging without a support desk, plus key-person risk if only one person can maintain it. That last cost is usually missing from comparisons and is frequently the largest.
What to do next
Open your platform's usage report and sort by consumption.
Two things to look for: a workflow consuming far more than its importance justifies, and any polling trigger running on a schedule. Both are common, both are invisible until you look, and both are usually fixable without changing platforms.
Most people find their bill is concentrated in one or two workflows they'd never have suspected.
Free: The automation planning checklist.
Related guides
- Zapier vs Make vs n8n — the billing models compared
- Marketing automation tools — the three tiers
- Is your automation actually saving time? — the full net calculation
- One workflow, three tools — consumption in practice
Join the Newsletter
Get practical marketing tactics delivered straight to your inbox.
Written by
Muhammad Basim
Related Articles
One Workflow, Three Tools: A Build Comparison
Feature tables don't tell you what a tool is like to use. The only thing that does is building the same thing twice and noticing where you got annoyed. So here's one realistic workflow — lead routing with enrichment and conditional assignment — specified once and mapped across Zapier, Make, and n8n. What changes between […]
GoHighLevel: 7 Automation Workflows to Build First
GoHighLevel gives you a hundred things you could automate, which is exactly why most accounts end up with forty half-built workflows and no measurable result. Seven are worth building first. They're the ones tied directly to revenue — catching leads before they cool, recovering appointments that would otherwise vanish, and asking for reviews at the […]
Zapier vs Make vs n8n: An Honest Comparison
A warning before the comparison, because you'll want it when reading anything else on this topic. A large share of the "Zapier vs n8n" content online is published by hosting companies and agencies that sell n8n hosting or n8n implementation. That doesn't make their analysis wrong — the cost gap they describe is real — […]