A pile of proposals is not a problem of judgement, it is a problem of process. A two-pass filter that gets you from twenty proposals to four serious candidates in an afternoon, the four questions that separate people who do the work from people who describe it, and a 20-minute call agenda that actually decides something.
Reading twenty proposals properly is a day's work, and you do not have a day. So most people read four or five carefully, get tired, and hire whoever felt reassuring — which is a decision made by the order the proposals arrived in. The way out is not better judgement, it is a process that spends your attention where it changes the outcome. This guide is that process: a two-pass filter, the questions worth asking, how to read a portfolio and a rating without over-trusting either, and how to close out the people you do not pick.
A two-pass filter beats reading everything
Split the job in two. Pass one is fast and brutal: read only the first three lines of every proposal and sort into two piles. Anything that opens with a paragraph about the sender rather than about your project goes straight to the reject pile, as does anything that could have been sent to any posting in your category. You are not judging ability yet — you are checking whether this person read your brief. It takes about fifteen seconds each, so twenty proposals is five minutes.
Pass two is slow and generous. Take the survivors — usually four to seven — and read them properly: the plan, the price, the questions they asked, the one sample they chose to attach. This is where your judgement is worth something, and now you have the time to use it.
If pass one leaves you nothing, that is information about your brief rather than about the market. A posting that no bidder engaged with specifically was probably not specific itself; writing a brief that gets good bids is the upstream fix, and reposting a sharper version will outperform grinding through the pile you have.
What a portfolio proves, and what it never proves
A portfolio is evidence, not a guarantee, and it is worth being precise about which is which. It reliably proves taste and range: you can see whether this person's aesthetic sense matches yours and whether they have worked at the scale you need. That is genuinely useful, and it is most of what people use a portfolio for. It also helps to know what a well-built one is trying to do to you, which is laid out from the other side in building a freelance portfolio that lands clients — a strong portfolio is a sales tool, and reading it as one makes you a better assessor of it.
- It does not prove authorship. On team projects, the person you are talking to may have done one part. Ask which part, on which item. A straight answer is common; a vague one is a signal.
- It does not prove they can work with you. The polished case study says nothing about whether they answer messages, miss deadlines or argue about scope. That is what the call and the references are for.
- It does not prove currency. Three beautiful pieces from four years ago and nothing since is a different proposition from three from last quarter. Check the dates.
- It does not prove they will repeat it. A portfolio shows the best outcome of a good engagement. Your engagement is a new one, with your brief, your feedback speed and your dependencies in it.
Four questions that separate doers from describers
Most interview questions invite a rehearsed answer. These four are hard to fake because they ask about specifics the person either has or has not lived through. Ask them in writing before the call if you like — a thoughtful written reply is a good sign in itself.
- “Walk me through how you would start this project in week one.” People who do the work describe a sequence with dependencies and unknowns in it. People who describe the work talk about qualities: “I will make it modern and user-friendly.”
- “What is the part of this job most likely to go wrong, and what would you do about it?” Anyone who has finished five projects like yours has a ready answer. “Nothing should go wrong” means either inexperience or a sales reflex, and both cost you later.
- “What do you need from me, and by when?” This is the question that predicts delivery. The honest answer usually includes something you had not planned to provide, which is exactly why it is worth hearing before you start rather than in week three.
- “Tell me about a project that went badly and what you changed afterwards.” You are not looking for drama; you are looking for a person who noticed a pattern and adjusted. Someone with no such story has either been lucky or is not telling you things.
Paid trial tasks, not free samples
You may reasonably want proof before committing to a full project. The right way to get it is to pay for a small, self-contained piece of the real work: one page of the design, one module of the build, one article of the series, the first data cleanup. Set it up as a small first milestone, so the amount is committed against that one piece of work and released only when you approve the output — what the funding step does and does not do today is spelled out in milestone payments and escrow for clients.
Asking for the same thing unpaid is a mistake in both directions. It filters out the strongest candidates, because people with a full pipeline decline unpaid spec work as a matter of policy, so you end up choosing from whoever had a free afternoon. And it tells the freelancers who do accept that your project is one where their time is assumed to be free, which is not the relationship you want to open with.
Reading reviews and skill badges without over-trusting them
A star rating is a summary, and a summary is all a profile here gives you: an average, and the number of reviews behind it. The review text itself is not on the profile, so do not plan to read between the lines of it. Read the pair for what it can honestly tell you — whether there is any history at all, and how much — alongside the count of completed projects next to it, which is the more informative half. What an average cannot distinguish is project shape: five five-star reviews on tiny tasks and one three-month build that went well produce the same number.
So get the detail where it is actually available: on the call. Ask which past project was closest to yours, what went wrong on it and what they changed afterwards, and — for anything substantial — whether there is a past client who would take a five-minute reference call. A freelancer who has done the work will offer one without flinching.
A skill test badge on a profile means the freelancer passed a platform assessment in that skill. Treat it as what it is: a floor, not a ceiling. It is genuine evidence that someone is not bluffing about a fundamental, and it is useful precisely when you cannot assess the craft yourself — but it says nothing about judgement, reliability or how they handle an ambiguous brief.
New profiles with no history are a real trade-off rather than an automatic no. A capable person starting out will often do excellent work to earn a first review. Manage the risk with structure rather than by refusing: a small paid first milestone, a short first scope, and a clear acceptance test. You can browse profiles directly in the freelancer directory rather than waiting for people to come to you.
A 20-minute call that actually decides something
Shortlist two or three people for a call, not six. The call is not for assessing skill — the proposal and the written answers did that. It is for assessing whether you can work together, and twenty minutes is enough if you run it to an agenda.
| Minutes | What you do | What you are listening for |
|---|---|---|
| 0–3 | You restate the outcome and the deadline in your own words. | Do they correct you or add something you missed? Good sign. |
| 3–10 | They walk you through week one and the riskiest part. | A sequence with dependencies, not a list of adjectives. |
| 10–15 | They tell you what they need from you and when. | Specific asks. Vagueness here becomes your delay later. |
| 15–18 | You cover working hours, response time and the check-in rhythm. | Whether their normal matches yours. Mismatched expectations, not skill, sink most projects. |
| 18–20 | You state the next step and the date you will decide. | That they are comfortable being told a decision comes later. |
Making the call, and closing the loop
Decide on fit against the brief, not on who impressed you most. The strongest interviewer is not always the strongest fit: for a well-defined build, reliability and clear communication usually beat flair, and for an ambiguous problem the reverse is often true. If two candidates are genuinely close, pick the one who asked the better questions — over a long project that habit compounds in your favour.
Then close out everyone you did not choose, in one line each: “Thanks for the detailed proposal — we have gone with someone else for this one, but I will keep your profile for the next project.” It costs you two minutes for the whole shortlist. It also protects something worth more than that, because the freelancers you rejected politely are the ones who will bid on your next posting, and the ones you left on read are the ones who will not.
The takeaway: run two passes, not one: fifteen seconds each to drop anything not written for your project, then real attention on the four to seven that survive. Treat a portfolio as evidence of taste and a rating as a thin summary — an average and a count, with no review text behind it — so get the detail on the call and ask which part of a team project was theirs. Pay for a small trial milestone instead of asking for free samples, interview two or three people to a fixed agenda, and decide against the brief rather than on charisma.