For freelance platforms
Freelance platforms: milestones, cancellations, and the transfer that never reversed
Freelance platforms hold money deliberately: funded milestones sit on the platform until work approves. That escrow-shaped flow is client-friendly and leak-prone — because cancelled projects, partial releases and client-side chargebacks interact with Connect defaults in combinations neither the billing docs nor your intuition cover. Here is the profile, mapped vector by vector, with the arithmetic that shows where margin actually exits.
Escrow behaviour, Connect reality
True escrow means separate charges & transfers: platform charges the client, holds funds, releases transfers on approval. Liability is unambiguous (platform), and every release is a hand-built transfer — meaning every cancellation-before-release, partial-release and post-release refund is manual-path code where flags get forgotten.
Modelled against a mid-size platform
500 projects monthly, average milestone $450 across 2.4 milestones, platform fee 12%, early-cancel incidence 8%, post-release dispute requests maybe 1%. Roughly $540k monthly transacted; refund-adjacent flows touch tens of thousands; typical leakage strands four figures monthly in freelancer accounts — concentrated in the cancelled-mid-project cohort nobody enjoys chasing.
- Projects × avg milestone value
- 500 × $1,080
- Early-cancel share
- ~8% → 40 projects
- Funded-but-unreleased at cancel
- $17k+ held
- Leakage if flags misfire on release/refund mix
- $1.4–3.5k/mo
Client-initiated chargebacks land on you
Clients dispute via card networks, not your resolution centre. Destination-style liability means the platform absorbs the debit while the freelancer keeps released transfers — clawback territory, with the added wrinkle that freelancers cash out fast. Acting on dispute.created rather than waiting for closure is worth real money here.
Repeat-offender patterns concentrate losses
A small cohort of clients generates serial disputes; a small cohort of freelancers attracts them. Neither cohort is visible without event-level attribution — aggregates hide them inside acceptable averages. Findings-per-account ranking surfaces both within weeks of monitoring.
What FeeGuard does about it
Milestone flows map cleanly onto detectors: release-time checks catch the refund-before-release race, dispute findings arrive with recoverable figures pre-computed, and risk scores rank the repeat offenders automatically. Weekly payouts common in this vertical actually help recovery odds — funds stay reachable longer than gig-marketplace norms.
Common questions
Do milestone releases change reversal math?
No — a released transfer reverses like any other, proportionally against whatever refunds apply to its source charge.
We pay freelancers weekly. Better or worse?
Better for recovery odds: shorter payout cycles mean balances are more often still reachable when findings arrive.