Playbooks · Everyday tasks
Screen Upwork postings before spending Connects
Written October 26, 2026 · Edited November 11, 2026
For a freelancer × choosing which Upwork jobs to bid on × an agent doing the first read, ending with proposals only on jobs that pass your criteria and are worth their Connects price — and a running tally of what each client you win actually cost to acquire.
Before you start
You need an Upwork freelancer account and a saved search that matches your skills, plus a written list of what a job must have before you bid. The agent's job here is the first read and the arithmetic, not the decision: it never submits a proposal, and you never let it invent experience, past clients, or results you do not have. The prices and rules below are Upwork's own published Connects rules, and the screening workflow is built on top of them.
The one discipline the whole workflow rests on: the Connects price is visible on the post before you apply, so read the price, do the arithmetic, then decide — and re-check the number immediately before you send, since it can change while the post is live.
What "done" looks like
At the end of one sitting you have a screened shortlist: each posting marked pass or fail against your criteria, each passing post priced in dollars, and a short list of proposals to write — plus a running tally of Connects spent, proposals sent, and clients won, so your cost per client won is a number you know rather than a feeling you have.
Steps
- Set a monthly Connects budget first. Upwork's own Connects page says additional Connects cost $0.15 each, sold in bundles, and that members on the Basic plan may receive 10 free Connects a month while Freelancer Plus members receive 100. Start from what you actually get: whatever free Connects you are eligible for — some Basic members may receive 10 a month, and Freelancer Plus members receive 100 a month once they have been on the plan for a month, 90 to 100 in the first month — plus whatever you are willing to buy. You need a verified billing method on your account before you can buy Connects. A budget written down before you browse is what stops the tenth marginal bid of the week; a budget decided mid-browse is not a budget.
- Write pass/fail criteria before opening any post. Five lines, each a yes or no: the skills match work you have actually done; the budget fits your rate for the scope described; the scope, timeline, and deliverables are stated clearly enough to price; the client history shown on the post gives you no reason to walk away; and the job is still open and recent enough to be worth a bid. A criterion that needs judgment ("interesting project") is not a criterion; rewrite it until it is a question with a yes or no answer.
- Have the agent do the first read, not the first bid. Feed the agent one posting at a time and ask for exactly three things: which criteria pass and which fail, with the quoted line from the post behind each verdict; what is missing from the post that you would need before bidding; and the Connects price shown on the post. Anything the agent cannot quote from the post is marked unknown, not filled in. The agent's output is a screening note, and a screening note is not a proposal.
- Price every passing post before writing anything. Upwork sets the Connects a proposal requires per job, based on the project's size, scope, and market demand, and shows the number on the post before you apply — some projects require none at all. Multiply the number shown by $0.15 to get the bid's price: 6 Connects is $0.90, 16 Connects is $2.40, and a post that needs no Connects costs nothing to bid on. Then compare that price against the job's value: a $2.40 bid on a small one-off job needs a much stronger match than the same bid on a large ongoing contract. If the price is higher than the job is worth to you, the post fails, however well it matches your skills.
- Write proposals only for passing posts, and check every claim. Draft with the agent if that helps, but read every sentence of the draft against your real history before sending. A proposal that invents a past client, a result, or a skill you do not have is not a stronger proposal; it is a lie a client can check in one message. Send the proposal yourself, from your account, after the check — the agent never submits anything.
- Record every bid the day you place it. For each proposal, write down the post, the Connects spent, the dollar price, and the date. This takes a minute per bid and it is the whole basis for step 7: without the record, your cost per client is a guess, and guesses do not change behavior.
- Track cost per client won, weekly. Divide Connects spent, net of returns, by clients won — that is your Connects figure. Keep a separate cash figure: multiply only the Connects you paid for by $0.15, and add any plan fee you pay for the month, since free Connects cost nothing and a plan fee is cash spent whether or not it bought Connects. Returned Connects go back into your balance for reuse, never as cash: they come back if a client cancels a project before a contract is made or if Upwork removes a post for violating its terms. You may also receive Connects when you land an interview with an established client — Upwork does not share the exact criteria, not every interview earns them, and the amount varies; when they are awarded, they arrive as a refund of the Connects you spent to apply, plus any extra earned. They do not come back if the client rejects your proposal, hires someone else, does not view or respond to it, or if the post expires — or if you withdraw your own proposal, or if you are outbid after the client already interacted with your boosted proposal. Every Connect expires one year after it was issued, and Upwork sends no reminder before they expire, so check your history. If your cost per client is climbing, the fix is stricter criteria, not more bids.
- Review the criteria against outcomes, monthly. Look at which passing posts turned into interviews and which did not, and which failing criteria, if any, predicted the failures. A criterion that never separates winners from losers gets rewritten or dropped; a pattern in the losers — a budget band, a post age, a missing scope detail — becomes a new criterion. The list gets sharper every month it is used, and a list that never changes is a list nobody is learning from.
Check the agent's work
- Check that every pass/fail verdict quotes a line from the post. A verdict with no quote behind it is the agent's guess, and guesses are what the criteria exist to replace.
- Check the Connects price against the post itself before sending. The number shown on the post is the price; a number the agent remembered from another post is not.
- Check the arithmetic: Connects shown multiplied by $0.15. A bid priced wrong is a budget broken quietly.
- Check every claim in a draft proposal against your real history. One invented client or result in a sent proposal is worse than a week of unanswered bids.
- Check that withdrawn or expired bids are recorded as spent, not as pending. Upwork does not return Connects for those, so your tally must not pretend otherwise.
What can go wrong
- Bidding before reading the post. The Connects price and the missing scope detail are both on the post. A bid placed without reading it spends money on a job you might have failed on sight.
- Treating a bid on a stale post as free. Connects are not returned when a post expires. A post that has been sitting for weeks is the most likely bid to evaporate with nothing to show for it.
- Withdrawing a proposal to "save" the Connects. Withdrawal does not return them. Once a bid is placed, the Connects are spent whether the proposal stays up or not — so the decision that matters is the one before sending, not after.
- Boosting a weak proposal. Boosting spends additional Connects to bid for one of the top four boosted slots on the client's list, in a first-price auction that closes after seven days or on first hire, whichever comes first. Whether you can boost at all depends on how strong a match you are for the job, and the bid table shows what other freelancers are bidding. You pay your full bid only if you are in the top four when the auction closes or the client interacts with your proposal while it is boosted — opening, shortlisting, messaging, archiving, declining, or sending an offer. If you are knocked out before the auction closes and the client never interacted with it, only the boost Connects come back; the Connects spent to submit the proposal do not. A boosted weak proposal is a more expensive rejection, not a better chance: boost only a post that already passed every criterion.
Limits
- This workflow screens postings; it does not predict which clients will reply, and no reply rate or win rate is promised here. The criteria improve your odds per bid; they do not guarantee any outcome.
- Connects rules change: the per-proposal number is set per job and can change with size, scope, and demand, and monthly allocations depend on your plan and eligibility. The numbers on this page are Upwork's published rules — check the post and your Connects page before spending, every time.
- The Connects a proposal needs can change while the post is live, and if it changes while you are drafting or submitting a boosted proposal, Upwork pulls the extra Connects from your total bid to meet the new minimum. Re-check the number shown immediately before you send, every time.
Related playbooks
- GeoVisA11y: maps a screen reader can explore
- Luma: an NVDA add-on that describes the screen
- Write Upwork proposals with an agent, without inventing anything
- Cheapest way into an always-on agent
Sources
Sources: Upwork on Connects, Upwork's Connects help page, Upwork on Connects charges, returns, refunds, and history, Upwork on boosting a proposal, Upwork on Freelancer Plus.
Back to: Everyday playbooks hub · Also: Chase an unpaid invoice with an agent