Rexz Bot logo Rexz BotBots and automation, explained plainly
Automation Tools

Automation Tool Pricing: Tasks vs Operations Explained

A reader wrote to me in September because she had signed up for a plan with 750 tasks a month, built two small workflows, and received a limit warning on the eleventh day. Nothing had gone wrong.

A laptop showing an automation canvas of five connected steps beside a usage dashboard with a monthly limit bar

A reader wrote to me in September because she had signed up for a plan with 750 tasks a month, built two small workflows, and received a limit warning on the eleventh day. Nothing had gone wrong. Her workflows were doing exactly what she designed them to do. The plan simply was not counting what she assumed it was counting.

This is the single most common way beginners overspend on automation, and it has nothing to do with choosing a bad platform. Every vendor meters usage differently, the words they use overlap without meaning the same thing, and the pricing page rarely spells out which parts of your workflow are billable. Once you can translate between the three or four counting systems in use, estimating your real cost takes about ten minutes.

What each vendor is actually counting

There are four metering models in common use, and the names are not standardised across vendors.

A task, on the platforms that use the word, usually means one successful action that changes something outside the workflow. Sending an email counts. The trigger that started the run does not. Filters and formatting steps generally do not. This is the most forgiving model and the one people assume everybody uses.

An operation is a step. Any step. The trigger counts, the filter counts, the step that reformats a date counts, and if a step runs inside a loop over twelve rows, that is twelve operations. A workflow with five boxes on screen costs five operations every time it runs, whether or not anything happened at the end.

A run or execution is one firing of the whole workflow regardless of how many steps are inside it. Good news for complex workflows, and the model where polling triggers quietly hurt you.

A credit is a weighted unit, where ordinary steps cost one and expensive ones cost more. Anything involving a language model, document parsing or image handling tends to cost somewhere between five and fifty. Credits are the hardest model to forecast and the most common on the newer tools.

Why the same workflow costs three times as much elsewhere

Take a concrete workflow. A form submission arrives, a filter drops anything without a phone number, a formatting step tidies the name, a row is added to a spreadsheet, and a notification goes to a chat channel. Five boxes on the canvas. Say 400 submissions a month, of which the filter discards 250 because they are incomplete.

Metering modelWhat gets countedMonthly usage
Tasks, actions onlyThe 2 actions, only on the 150 that pass the filter300
Operations, every stepAll 5 steps on 150, plus 2 steps on the 250 filtered out1,250
Runs or executionsOne per submission, steps irrelevant400
Credits, weightedAs operations, but add 10 per run if a model summarises the note5,250

Same workflow, same business, and the usage number swings between 300 and 5,250. If you sized your plan from the first row and bought the platform in the second, you will be out of allowance in the second week, which is precisely what happened to the reader who wrote to me.

The row that catches almost everybody is the filtered traffic. Under an operations model you pay for the trigger and the filter on all 400 submissions even though 250 of them are thrown away immediately. A filter does not save you money unless it sits in front of the expensive part and the vendor bills by action rather than by step.

The trap in cheap per task pricing

Here I disagree with most of the comparison articles, which line up the plans by cost per task and declare a winner. That number is close to meaningless on its own, for three reasons.

The first is the one above: a cheap operation is not comparable to an expensive task, and the ratio between them for a typical five step workflow is around four to one. A platform charging a third as much per unit can still cost you more.

The second is polling. Where a tool has no webhook available it checks for new data on a schedule, and on several platforms every check consumes a unit whether or not there is anything new. A workflow polling every five minutes burns roughly 8,640 checks a month doing nothing at all. Switching that one trigger to a webhook, if the source supports it, can cut a workflow's cost by ninety percent. If you have not met the term, a plain explanation of webhooks is worth ten minutes before you pick a plan.

The third is retries. A failed step that retries three times may bill three times, and errors are not rare. A flaky integration on a nightly job can produce a bill that looks like a usage spike with no change in your business.

So the useful comparison is not cost per unit. It is cost for your actual workflows at your actual volume, which you can only work out by counting your own steps.

How to estimate before you commit

This takes about ten minutes per workflow and it is the whole job.

  • Draw the workflow and count every box, including the trigger, filters and formatting steps. That number is your steps per run.
  • Find your real monthly volume from the source system, not from memory. Count the last three months and take the busiest one, because the busiest month is the one that breaks the plan.
  • Note where the filter sits. Steps before the filter are paid on all traffic. Steps after it are paid only on what survives.
  • Look up which model the vendor uses, in their documentation rather than on the pricing page. Search their help centre for the word they bill in and read the page that defines it.
  • Add a third for retries, loops and testing. Building a workflow consumes allowance too, and a busy afternoon of testing can take a few hundred units.

Then run both numbers. Steps multiplied by volume gives your operations figure, actions multiplied by surviving volume gives your task figure, and the plan you need is whichever is larger under the model your vendor uses. Most beginners find the honest answer is one tier above where they were about to start, and that is cheaper than finding out on day eleven.

One more habit worth forming: check your usage dashboard weekly for the first month. Every platform has one, almost nobody looks at it, and it shows which single workflow is consuming most of your allowance. Mine was a status check that nobody read. If you have several tools running at once, the broader cost problem is covered in the hidden cost of too many tools, and the specific differences between the two best known platforms are compared in Zapier against Make.

FAQ

Does a workflow that fails still cost me?

Usually yes, up to the point where it failed, and again on each retry. Steps after the failure are not billed. This is why a broken integration can look like a sudden usage spike.

Do filters and conditions count as billable steps?

On operations and credits pricing, yes. On action based pricing, generally no. This single difference accounts for most of the surprise in a first invoice.

Is the free tier enough to learn on?

For learning, yes. Free tiers commonly give 100 to 1,000 units a month, which is plenty for two or three light workflows and enough to find out how your chosen platform counts. It is not enough for anything a customer depends on.

Should I switch platforms to save on pricing?

Rarely worth it below about 50 dollars a month, because rebuilding and retesting your workflows costs more time than the saving is worth. Above that, run the two numbers properly first. More guidance on choosing tools sits in our automation tools section.

OP
Oskar Petrov

Oskar explains automation tools for beginners, helping people start simple and avoid overcomplicating their setup.

More posts by Oskar

More in Automation Tools

01doorganico 084goad 3hot8 4465456 after5internprogram ahandfulofblessings ahmadpc ahomcybersecurity allodialresources annodex antiquesandperiodcollectibles aotc asianworldkashmir beedlectrics bloggertemplatefree bluefeatheragency bodyweightsolution campaign08blog carfixadvisor catstreetcollective cbstvsports cnfanspurchasing committeeof500yearsofdignityandresistance covertopsdjs cs-kaiyuangame cuestacomunicaciontotal cxoutlet dancebetgame deskgearhub destinyvillastay detectmagazine di33 diamondsteelbuildings donmo dyanadropship emarsysfinance emimattl escapechan eskitarih evautoplay everythingweddings-orlando farmzone feedviewerapp foxstudiosscratch gamevui123 gleasongrade greetingsday hacklabasu haskell iw-kaiyun jagoanproperti jilidskripsi joltbusinessfinance juegodepokergratis juegospucca kaeru-festival latestmehndidesigns lestanzediottaviano litemagazines localcommunityhero lustrsazan mariashebets menswheels newwebstechs1 nikkit ootheday oxigenvpn plurfitness prefetial probudgetcarpetcleaning psychicmrsmartin pyaten qvantor rejoignezlogis riselayouts ruchin rwxqsgd5 sakaltime shibiaowang shop-bot skrtips spotlesshaven strcoding tetapatrium topselller uneplanteartificielle vipertechcr webtech1223411 wo-kygaming yamaha-mekar yeknava yoybuysheets zenunwa yamaha-mekar jilidskripsi ankertoto ankertoto ankertoto