Freelance n8n developer or automation agency: how to know whether they can keep it running

Wednesday, eleven in the morning, month fourteen. The workflow that loads orders into the system stopped working on Monday, you wrote on Monday afternoon and again yesterday, and for three days orders have been entered by hand. On the other side there is nobody absent or angry: there is a person who started a new project and is billing someone else this week.

Which of the two versions of that Wednesday you get is decided months earlier, and not on price: it is decided by the unit the quote is written in. By the hour or by the project means the work was paid for in full on delivery day, and Wednesday's fix is time that does not bill. By the month means you bought a slot in someone's calendar, and your urgency competes against other urgencies worth the same. That line is not moved by price or by years of experience.

The economic comparison against hiring someone in-house is a different discussion and it is in the 12-month math. Here the question is smaller and more actionable: how to know, on the first call, whether the freelancer who sent you that n8n, Make or Zapier quote will still be answering in month fourteen. It is three questions about the past —promises about the future are free and everyone answers them well— and one piece of math.

Look at how they bill you before you read how much they bill you

The day something stops working, the question is not whether the freelancer knows how to fix it. They do: they built it. The question is what that hour competes against inside their week.

If they charged you by the project, the fix is work that does not bill. There is no bad faith anywhere in that sentence: there is a person with a project in progress that is billing and a client from eight months ago writing on a Wednesday at eleven. They will do it, and they will do it well, but they will do it when they can. That "when they can" is exactly the product you bought without knowing you were buying it.

If they charge you by the month, your urgency competes with other urgencies worth the same to them, which is a far more even contest. Something less obvious changes too: someone who bills every month has an economic reason for it not to break, and someone who charged once has the opposite reason.

The table does not say the first column is wrong. It says that if you are going to buy the first one, you had better know it the day you sign and not the day it breaks.

Question 1: what of everything they built is still alive

The three questions are about the past. Promises about the future are free and everyone answers them well; a track record cannot be improvised on a thirty-minute call.

Exact wording: "Of everything you built in the last two years, what are you still maintaining today and what got switched off?"

They can sustain it if they answer with quantities and with dates. Four clients with things in production, the oldest from March last year. And they know why the ones that got switched off got switched off: one closed the business, another moved it to an internal team. Whoever sustains has a boring, specific story.

They only know how to build if the CV is all deliveries. "I've done more than forty automations" without a single one still alive under their care is not a merit, it is a data point: forty deliveries and zero maintenance means they never lived with a system in production for more than a month.

The follow-up that cannot be dodged: "Which of those four gave you the most trouble?" Whoever maintains has the anecdote on the tip of their tongue, with the name of the system that failed and the stupid detail that caused it. Whoever only builds goes generic.

Question 2: who found out the last time it broke

Exact wording: "The last time something broke in production, did you find out yourself or did the client tell you?"

They can sustain it if they say an alert reached them. And there is an even better answer, the one worth the most in the whole call: "the client told me, and from then on I set up alerts". That one shows two things that cannot be acted. That the failure cost them something, and that they changed the way they work because of their own experience and not because of a best practice they read. It is the closest thing to a guarantee you are going to hear.

They only know how to build if they answer that nothing has ever broken on them. It is the only answer across the three questions that admits no useful follow-up, because there is nothing underneath: what they are telling you, without meaning to, is how long they have lived with something in production.

The follow-up that cannot be dodged: "How long between it breaking and someone noticing?" With monitoring, the answer is measured in hours. Without monitoring, in days or in weeks. Both are honest answers and both help you decide; the one that does not help is the one with no number in it.

Question 3: what happens on a Tuesday when they are not around

Exact wording: "If this goes down on a Tuesday at three in the afternoon and you are buried in another project, what exactly happens?"

They can sustain it if they have a concrete, unglamorous plan. A colleague with access who they hand cases to when they are away. Two weeks' notice before going away. Or straight out: "nine to six I answer the same day, outside that I don't; if you need weekend coverage, I'm not your person". That last one is an honest no and it is worth gold, because it lets you decide with the information on the table.

They only know how to build if the answer has neither a name nor an hour in it. "I'll take a look", "I'm connected anyway", "it has never been a problem". That is not bad faith either: it is that until now it has never happened to them while they were genuinely busy, and the first time it does, you are the experiment.

The follow-up that cannot be dodged, and the most important of the six: "How many clients with things in production do you have today, and how many of those pay you a monthly fee?" Both numbers matter, but the second one matters more, and the math for why is right below.

The math that makes the answer to the third one verifiable

A workflow that depends on several outside platforms gets cut off some two to four times a year; the concrete causes and the math of what fixing them costs are in one-time payment or monthly fee. Let us take three outages a year per client, and a freelancer with nine clients with something running.

That is 27 incidents a year. Counting the diagnosis, the fix, the messages back and forth and the time lost remembering how that workflow was built, each one takes about two hours: around 54 hours a year, four and a half a month. Against a calendar of some 150 hours a month, that is 3 percent. Capacity is never the problem.

The problem is the other three properties of those hours.

None of them can be scheduled. They land on top of a project that already has a delivery date committed to another client. That is why the delay you will suffer looks nothing like the availability you were promised: it looks like the delivery date of a project you will never hear about.

If they charged by the project, those 54 hours do not get billed. At a rate of USD 50 that is USD 2,700 a year of free work, held up by goodwill and by fear of looking bad. Nobody sustains that indefinitely, and almost nobody announces it: freelancers with a long client list and no monthly fee do not quit, they start answering later each time.

They do not arrive evenly spread. The 27 incidents do not fall two a month: they fall on the day a platform changes something, and that day it breaks for the five clients who were using that integration. There is no calendar there, there is a queue, and your place in that queue was decided by someone else.

That is where the reading of the two numbers you asked for in the follow-up comes from. Divide clients with a monthly fee by clients in production. Close to zero with a long client list means you are joining a waiting list that is already subsidized with free work. One or two clients in total means the opposite and it is just as uncomfortable: you will matter a lot to them, and there is a real chance that within a year they get a full-time job and shut down. The healthy zone is the boring middle: between five and eight clients, with most of them paying a monthly fee.

Four cases where the freelancer wins and the agency is overpaying

• A closed, finished project. Migrating data once, building a report, connecting two systems for a one-off campaign. There is nothing to sustain, so there is nothing to buy from anyone by the month.
• When you already have someone inside who can operate it. The test is not that a systems person exists: it is whether that person was in the meetings where it was decided how it gets built. Whoever inherits a workflow they never saw does not sustain it, they rewrite it.
• When you are still testing whether automation is useful to you. A freelancer building a small workflow gives you more information about your own process than three sales meetings.
• When the process does not hit the money or the calendar. A report that arrives late is annoying. A badly booked appointment or an out-of-date stock level costs.

If you go with a freelancer anyway, ask them for these three things

• A review date, already quoted. Half an hour in month 3 and another in month 9, paid at today's price and scheduled before the project ends. It is the freelance version of the monthly fee: you are not buying availability, you are buying two moments in which someone looks. They cost less than the first diagnosis hour on the day it breaks, because in month 3 they still remember how they built it.
• A second name with a phone number. Not "I have a colleague": the name, the number and some proof that this colleague has already been inside your workflows once. If the name does not exist, go ahead anyway if the rest adds up, but write down that the continuity of your operation now depends on the health and the mood of one single person.
• A signal you can watch yourself without being technical. If leads come into the CRM every day and one day none come in, that is your alert. It is worth more than any dashboard you are never going to open, and it is the only item on this list that does not depend on them.

What gets signed —account ownership, exports, exit plan— is worth the same with a freelancer as with an agency and does not change because it is one person alone: it is all in the 12 questions before signing and in who owns your automations.

These three questions tell you whether that person can sustain what you have, not whether what you have needs sustaining; that is decided by counting how many outside platforms the workflow depends on. And there is something no good answer moves: even with the three answered perfectly, there is still one single person. How that gets split between names with a backup is in who maintains n8n inside a company.

If you already had the call, send us the three verbatim answers they gave you at info@striqtech.com and we will tell you what we read there, even if the reading is that this freelancer suits you better than we do. The one that weighs the most of the three is the second: if they answered with the name of a monitoring tool and an inbox that receives the alert, the rest of the conversation has already changed tone. If they answered "the client tells me", it has too.