Start with a question your numbers can answer
A Connects budget becomes useful when it tells you what an attempt to win work costs. Looking only at the number of proposals sent leaves out both the money and the hours behind them. Looking only at contracts won can hide several weeks of work that produced those contracts.
Choose a group of proposals to track, such as those sent during one calendar month. Record their outcomes as they develop. A client who signs next month still belongs to that original group. Keep new clients separate from repeat projects so one existing relationship does not make your acquisition cost look artificially low.
Use the current price and your actual history
Upwork lists purchased Connects at US$0.15 each. The number required for a proposal varies by job and can change while the job is open. Boosts use additional Connects. Check the displayed requirement before submitting and record the amount actually charged. Upwork: Understanding and using Connects
Keep two records: cash paid for Connects and Connects consumed by your proposals. A purchase increases your balance; it does not mean you used that whole balance that month. For comparisons, you can value consumed Connects at the current purchase price, but label that figure as a replacement-cost estimate. It is not necessarily your cash expenditure.
Reconcile the account balance without mixing it with acquisition cost
Keep a simple balance record alongside the proposal sheet. Start from the opening balance, add purchases and other credits, then subtract spending and expirations and add any returns. Use the actual account history to resolve a difference. A spreadsheet covering only this month cannot reconstruct an older balance unless it includes that starting figure.
Upwork distinguishes temporary boost adjustments from final charges in its history. Match related entries before treating a returned amount and a later charge as two separate application costs. Read the transaction-history guidance.
Give cash purchases their own line. Buying a bundle for later use affects cash now, while the proposals consuming it may belong to several periods. Label the measure you report: purchased this month, consumed by this proposal group, or replacement value. That small distinction makes your numbers easier to compare and explain.
Calculate proposal cost and client cost separately
Add the base Connects and the extra Connects actually charged for boosts. Subtract returns attributable to those proposals when they arrive. Keep unrelated spending, such as profile advertising, in a separate line unless you are deliberately calculating the cost of the whole channel.
Connects cost per proposal equals the value of Connects consumed divided by proposals sent. Connects cost per new client equals that same value divided by new clients won from those proposals. State the period and whether any proposals are still open alongside the result.
Suppose 40 proposals consume 480 base Connects and 120 extra Connects in total, with no returns. That is 600 Connects × US$0.15 = US$90. The average is 15 Connects, or US$2.25, per proposal. If this group eventually produces two new clients, its Connects cost per client is US$90 ÷ 2 = US$45. These are invented inputs for explaining the calculation, not expected results.
Model reply and hire rates without calling them results
The calculator now accepts an optional reply rate and a hire rate from replies. Both are your assumptions. Expected replies equal proposal count multiplied by the reply rate as a decimal. Expected hires then multiply those replies by the assumed hire rate. Divide Connects spend by the corresponding expected outcome to obtain the scenario’s cost per reply or hire.
For an illustrative scenario, 20 proposals consuming 12 Connects each cost US$36 at US$0.15 per Connect. A 20% reply rate models four replies; a 25% hire rate from those replies models one hire. The estimated costs are US$9 per reply and US$36 per hire. If the reply-rate assumption falls to 10%, the model produces two replies and half an expected hire, making the estimated cost per hire US$72.
Half an expected hire is a planning average, not half a contract you can deliver or a promise that another batch will complete it. With zero expected hires, cost per hire is undefined. Keep this scenario beside actual outcomes and label both clearly.
Include the hours that Connects do not cover
Track job research, proposal writing, replies, discovery calls, and estimates. Use an hourly value you choose for planning. This makes the time visible; it does not mean every sales hour would otherwise have become a paid delivery hour.
Continue the illustrative example: 12 sales hours valued at US$20 per hour add US$240. Add that to US$90 of Connects usage and the total becomes US$330. Across two new clients, the cost including time is US$165 per client. Any management fee, subscription, or other acquisition expense would need adding separately.
The free Connects calculator includes optional sales-hours and hourly-value inputs. Use them to compare the Connects-only figure with the figure including your time. A contract's total value is not profit: delivery costs, platform fees, and other expenses still matter when you judge whether the work is worthwhile.
Set a spending limit before the next proposal
Choose an amount you can afford to test and a review date. Use recent observed Connects consumption to estimate how many suitable proposals it could cover. Do not turn the estimate into a quota that forces you to apply when the work is a poor fit.
Using the illustrative average of 15 Connects per proposal, a US$60 usage budget represents 400 Connects and covers 26 complete proposals at that average, with 10 Connects remaining. The actual count changes when requirements or boosts change. Set a separate limit for optional boosting so it cannot quietly consume the base application budget.
Separate an affordable test from a forecast you need to be true
Start with a cash and time limit that you can tolerate even if the group produces no new client. A plan that requires a particular conversion rate just to remain affordable is fragile when you do not yet have comparable observations. You can reduce the scope of the test by narrowing the offer, limiting optional boosts or spending more preparation time on relevant proof.
Decide in advance what would justify another test. That could be enough suitable opportunities to assess, clearer evidence of fit in replies, or a specific problem you can address before spending again. It need not be a fixed number borrowed from another freelancer. If the only response to weak results is a larger purchase, pause and inspect the selection and conversations first. Keep enough time available to handle a useful reply; more applications are not valuable when you cannot follow through.
Know what a published result cannot tell you about cost
Our 2025 account snapshot records 244 proposals and 13 hires, along with views and interviews. It does not provide the Connects consumed, cash purchases, sales hours, fees or collected revenue for that group. You therefore cannot calculate that account’s acquisition cost from the screenshot alone.
When a provider shows you a similar result, ask for the missing denominator and the cost basis before comparing it with your own activity. If the records are unavailable, report the available counts and leave the cost unknown. Do not multiply a headline average by today’s Connects price and describe the result as historical spend. The calculation can be useful as an explicitly hypothetical scenario, but it is not evidence of what the account actually paid.
Review costs with the person who owns the next decision
At the review meeting, bring the original proposal group, its later outcomes, actual Connects history and time estimate. Pick one action: continue the same bounded test, revise a specific part of the process, or stop spending on that offer for now. Record who owns the change and when you will inspect it again. Keep proposals that are still unresolved visible rather than quietly treating them as losses or future wins.
If you buy support with included Connects, check the allowance, how costs are documented and which other expenses are outside the agreement. Do not count the same reimbursed spend twice in a total. UGD’s pricing and Connects terms explain our offer; your own comparison should use the actual agreed arrangement. Open the Connects calculator to record actual costs and keep optional forecasts separate.
Review the reason for the spend
At your review date, look at suitable jobs found, proposals sent, meaningful replies, calls, new clients, and time used. Read the conversations too. Several replies asking for a service you do not offer point to a different problem from good conversations that stop after an unclear estimate.
Do not assume every unsuccessful application will be refunded. Upwork lists returns for a client cancelling before a contract and for posts removed for terms violations. Rejection, another freelancer being hired, withdrawal, or expiry without a hire do not qualify. Returns are reusable Connects, not cash. Upwork’s return guidance
Change one clear part of your approach, record why, and review another comparable group. A low cost per proposal is only useful if the proposals reach work you can deliver profitably.
Clear answers before you act.
What if I have not won a client yet?+
Report the spend, time, and zero clients won. Cost per client cannot be calculated with a zero denominator. Keep pending opportunities visible and avoid replacing missing results with an assumed conversion rate.
Should I count free Connects?+
Count them in usage. For cash budgeting, record what you actually paid. For consistent comparisons, you can value all consumed Connects at the purchase price and clearly label that as replacement cost.
How much should I spend each month?+
There is no useful universal amount. Start with an affordable test limit, your observed application costs, and the time you can give qualified opportunities. Review those together before increasing the budget.
Continue with a focused guide
Check the source, not just our word.
Platform details can change. This guide was reviewed on November 18, 2026. If something looks wrong, email team@upgrowthdesk.com.