Upwork proposal benchmarks: what a good conversion rate actually looks like

A healthy Upwork account gets its proposals opened by the client between 30% and 50% of the time, and turns 10% to 25% of them into a chat. Those ranges come from 19,590 proposals we sent on behalf of 15 agencies over three years. They are the first thing we look at when an account is underperforming, because the two numbers point at completely different problems.

Upwork does not publish conversion benchmarks. Forums trade anecdotes. So the question every agency owner asks — *is this normal?* — is also the one they cannot answer alone. Here is our attempt, with the caveats that matter.

The three numbers worth tracking

The order matters. Viewed rate is mostly about whether you were reachable and worth clicking. Chat rate is about what you wrote once they clicked. Treating both as one blended 'response rate' averages two unrelated problems into a number you cannot act on.

What the client actually sees before deciding to open you

This is the part most advice gets wrong, and we got wrong in the first version of this article. The client is not looking at a list of names. Each proposal in their list already shows: your photo, your country, your profile title, job success score, completed jobs, total hours, total earned, average response time, your skills, your hourly rate, a Boosted badge if you bid for placement — and the first two lines of your cover letter, with a *View more* link.

So the opening of your letter is not hidden behind the click. It is part of the decision to click. A low viewed rate can absolutely be a writing problem, as long as you understand that only the first two lines are doing that work.

Why proposals go unopened, in order of how often we see it

1. 01You were never reached. On an active posting the queue fills fast, and clients stop scrolling long before the end. Placement decides whether you exist.
2. 02You were outbid for placement. Boosted proposals sit at the top, and the boost is an auction: a higher bid pushes yours down the ranking.
3. 03You landed in the bucket Upwork considers a poor match. Proposals the platform judges misaligned with the posting get grouped away from the main list. In practice they are read by nobody.
4. 04You were reached, and the card did not convince. Country, rate, badge, stats, photo, profile title — all of it is visible before the letter.
5. 05The first two lines said nothing. Opening with who you are rather than what you understood about their problem is the most common version of this.

Real accounts, real spreads

Averages hide the useful part. These are three accounts we ran, anonymised to the type of agency, so you can see how wide the spread gets between a mature account and a cold start.

Cost per chat means nothing on its own

Connects cost money, boosted placement costs more, and someone's hours go into every submission. Divide that by conversations started and you get cost per chat — a number that is only meaningful next to your average contract value.

Two accounts of ours, both healthy, illustrate the range:

On a new AI chatbot account starting from zero history, the first six weeks came out at $14.67 per view and $42.54 per chat. Whether that is expensive depends entirely on what a contract in that niche is worth. As a reference point, plenty of agencies pay a lead generation partner 5% to 10% of contract value for a client — against a $5,000 project, $42 a conversation is cheap; against a $300 project it is ruinous.

The rule we use: cost per chat, multiplied by the number of chats it takes you to close one client, has to stay comfortably below what that client pays you. If it does not, no amount of extra volume fixes it. We have turned down work on exactly this arithmetic, and it is a better conversation to have in week one than in month six.

About applying fast

Applying early helps, and it helps for a narrower reason than people think: it buys visibility before the queue fills. On an attractive posting we routinely see 30 to 50 proposals arrive within the first 15 to 20 minutes, and the overwhelming majority are automated or AI-written. Getting in before that wave is most of the benefit.

What early submission does not do is decide who gets hired, and that depends heavily on the size of the job:

• Small, well-defined work — a logo, a deck, a few hundred dollars — often gets hired quickly, sometimes from the first few credible replies. Speed genuinely matters here.
• Large projects — a $10,000 to $20,000 build — are not decided in an afternoon. The client is comparing, scheduling calls and reading portfolios. Being first buys you a look, nothing more.

What decides the large ones is fit. We once lost a contract after a call that went well, and asked the client why. The answer: *we liked how you think, but the freelancer we picked had delivered an almost identical project a month ago — same databases, same tooling.* No speed advantage closes that gap. A relevant, recent case does.

How to read your own numbers this week

1. 01Take the last 100 proposals, not the last 10. Below 50 submissions the numbers swing too hard to mean anything.
2. 02Split viewed and chat. They are different problems with different fixes.
3. 03Record the competing proposal count and whether you boosted, at the time of applying. Without those two fields your viewed rate is uninterpretable.
4. 04Compare by job post age, and keep small jobs separate from large ones. Mixing them hides both effects.
5. 05Only then rewrite the letter — and rewrite the first two lines first, because that is the part the client sees before deciding.

What the numbers cannot tell you

Conversion says nothing about whether the contracts are worth having. We have seen accounts with a 25% chat rate lose money, because every conversation was a small project with a client who wanted four calls first. Track contract value and hours alongside conversion, or you will optimise your way into a busy, unprofitable month.

And conversion is not visibility. If your profile never appears in client searches, every lead has to be bought through proposals and no invitations arrive the other way. That is a separate problem with separate numbers — we measured what actually moves a profile in Upwork search across 37 queries.