How to read an automation seller's portfolio before you order

The short answer

An automation portfolio answers one question well: has this seller built something with this shape before? It answers almost nothing about the questions that cost you money later. Did it keep running after the demo? What happened the first time a step failed? What did the client get to keep?

So read the portfolio for fit, not for proof. Look for samples that match your job’s trigger, size and tools, then ask the seller for the one sample that matters most: the closest thing they have built to your project, sent in reply to your first message. In my own orders, that reply was what I decided on, and as far as I remember, what the sellers attached were mockup designs rather than live links.

What a portfolio sample can and cannot show

Most automation portfolios are a few images or a short video: a canvas of connected boxes in n8n or Make, a before-and-after of a spreadsheet, a message that arrived in a chat app. That is useful, and it is limited.

What a sample can show:

• The seller has worked with the tools you use, or something close.
• They have built a workflow with roughly your number of steps and branches.
• They think about the result the client sees, not only the canvas.

What it cannot show:

1. Unattended behaviour. A screenshot is one moment. It cannot tell you whether the job still ran on night thirty, after a token expired or a machine slept. That is what the acceptance checklist for unattended automations is for.
2. Failure handling. A clean canvas often hides the fact that there is no error branch at all. The happy path photographs well.
3. What happened after delivery. Whether the client got the export, the credentials and a way to restart it is not visible in a picture.

My own content pipeline is a good example of why a single working run says little. Its scheduled publisher worked on the post format it was built for. When I changed the format, a dry run showed that it would have published every post with one part silently missing, because three places in the code still expected the old shape. Any screenshot of that pipeline, before the change, would have looked perfect.

Six things to read in each sample

Open each sample and go through these. You do not need to understand the tool to do it.

A sample that shows four of these is strong. A portfolio where every sample is only a canvas is not a bad sign on its own, but it means you will get the rest of the answers from the conversation, not the page.

Size the sample against your job

Portfolios mix small and large work, and the gig page does not separate them for you. The reviews help here, because each review shows the order’s price range and how long it took.

When I read 317 recent reviews of eleven automation sellers, the orders were spread widely: 91 were under $100, 124 were $100–400, 65 were $400–1,000 and 37 were $1,000 or more. Delivery ranged just as widely: 105 within three days, 74 in four to seven, 66 in eight to fourteen and 51 at fifteen days or more. The full breakdown is in what 317 Fiverr reviews say about hiring an automation freelancer.

That spread is the point. A seller whose portfolio and reviews are mostly small, quick orders may be excellent at exactly that, and new to a two-week multi-tool project. The opposite is also true: a seller used to large pipelines may not be the fastest choice for one clean connection. Match the samples and the review sizes to your job before you match the stars, which barely differ between sellers.

Your job	A matching sample shows	A weaker match
One trigger, two tools	A short workflow with a clear output	Only large dashboards or AI agents
Scheduled, unattended	A schedule trigger and some failure alert	Only form or button-triggered demos
Several tools, branching rules	Branches with labels, a run-book	One linear chain of steps
AI step in the middle	Example input and output of that step	A canvas with an AI box and no result

Ask for the sample you actually need

The best portfolio item is not on the page. It is the one the seller sends when you describe your project.

My own first messages to sellers have been short: a greeting, the project I wanted to start, and a question about whether they had done something like it before. What I decided on was the reply, specifically the similar past work the seller attached to it. A seller who has built your kind of workflow will usually reach for it straight away. One who has not will send the general portfolio again, which is also an answer.

Paste this into your first message, with your own details:

I want to automate [one sentence: the task and the result]. It should run [on a schedule / when X happens] and connect [tool A] to [tool B]. Have you built something similar? If so, could you share the closest example, with any client data removed, and tell me:

1. How long it has been running, and whether it runs unattended.
2. What happens when a step fails, and who is told.
3. What the client received at the end (export, run-book, walkthrough).

The first message guide covers the rest of that message, and the workflow automation brief gives you the fields to send once you have a shortlist.

Reading the answers

The three questions in the message are there to test how the seller thinks about the parts a portfolio cannot show.

• “How long has it been running?” A good answer is specific: weeks or months, on whose account, with or without the seller still involved. “It works great” is not an answer to the question.
• “What happens when a step fails?” Listen for retries, an alert to a named person, and a log. If the answer is “it won’t fail”, ask what happens when the login it uses expires. Every login eventually does.
• “What did the client receive?” The answer you want includes the workflow export or repository, credentials on the client’s own accounts, and some written way to restart it. These are also the items to check later in the custom offer.

A reply that asks you something back about your case is a stronger signal than any sample. It means the seller has started thinking about your workflow, not theirs.

Where it goes wrong

• A sample with someone else’s real data. If a portfolio image shows a past client’s real customer names or emails, that is how your data may be treated too. Ask how they handle test data.
• Judging the canvas by its looks. A neat, colourful workflow is not a more reliable one. The failure path and the hand-over matter more than the layout.
• Choosing on the portfolio alone. It narrows the list. The reply to your brief chooses between the names left on it.

Wrap-up

Read each sample for six things: trigger, size, tools at both ends, a failure path, the real output, and hand-over traces. Size it against your job using the price and duration on the reviews. Then ask for the one sample that matters, the seller’s closest past project to yours, and decide on how they answer the three questions about what the picture leaves out.