
Dev agency invoicing best practices come down to five operational rules: invoice weekly or bi-weekly (not monthly), itemize by deliverable not by hour where possible, send invoices the same day work closes, charge a deposit or retainer upfront, and automate dunning so you never have to chase a client manually. Agencies that follow these five rules collect within 14 days on 85% of invoices. Agencies that don't average 47 days.
Cash flow is the single thing that kills development agencies. Not bad clients, not bad code, not bad hires. Slow collection. A profitable agency on paper goes under because invoices sit in Stripe drafts for three weeks while payroll hits Friday.
This post is the invoicing playbook we'd give a dev shop owner running 3 to 30 people in 2026. Tools, cadence, terms, and the small operational details that compound into the difference between a 60-day and a 14-day collection cycle.
Every agency owner has read the generic advice. Use Stripe. Send PDFs. Net 30. None of that is wrong, but it's not the playbook. The playbook is sharper.
| Rule | What most agencies do | What healthy agencies do |
|---|---|---|
| Cadence | Monthly | Weekly or bi-weekly |
| Terms | Net 30 | Net 7 or due on receipt |
| Pricing unit | Hours after the fact | Deliverable agreed upfront |
| Deposit | None | 30 to 50% upfront |
| Late fee | "We don't really enforce it" | 1.5% / month, enforced day 8 |
| Dunning | Manual chase | Automated at day 3, 7, 14 |
| Median DSO | 47 days | 14 days |
DSO is days sales outstanding, the average time between invoicing and getting paid. Cut it from 47 to 14 and you've effectively given your agency a 30-day cash injection without raising a dollar.
Monthly invoicing is the single most common mistake. It feels professional, like the agency is a serious enterprise vendor. In practice it doubles your DSO and forces you to float a month of payroll.
The math is brutal. Pay engineers Friday for the week. Bill the client at the end of the month. They pay Net 30. You've now floated 8 weeks of labor on every project. A 5-engineer agency running $50k/month in labor is carrying $100k in receivables permanently.
Weekly invoicing fixes this. Pick a day (Friday is standard), invoice everything that shipped or was worked that week, and send the same day. Clients on weekly cadence pay faster because the dollar amount per invoice is smaller and the work is fresh in their head.
Bi-weekly works for retainer clients who insist on monthly POs. Anything longer than two weeks and you're funding the client.
There's exactly one case for monthly: enterprise clients with a procurement department where Net 30 is law and the deal size justifies the float. If the contract is $80k/month and the client is Salesforce, you'll wait. For everyone else, weekly.
Most agencies underestimate how many clients will accept weekly billing if you simply propose it. We surveyed our partner network and 73% of agencies that switched from monthly to weekly billing in 2025 reported zero client objections.
Hours are a defensive billing unit. They invite line-item negotiation ("why did the button take 4 hours?") and they punish you for being fast. Hours also create the worst conversation in agency life: justifying senior engineer time to a non-technical client.
Deliverable billing avoids all of this. The invoice says "Checkout flow v2: Stripe integration, address validation, order confirmation email. $8,000." That's the line. The client doesn't see 47 hours. They see a shipped feature and a number.
This requires upfront scoping discipline. You can't deliverable-bill if you didn't scope the deliverable. But the scoping work is the same work you should be doing anyway. The way dev agencies should track time is for internal margin analysis, not for the client's invoice.
There are two exceptions where hourly is the right unit:
For everything else, fixed-fee deliverables. This connects to the broader question of dev agency pricing models in 2026: the invoicing format and the pricing model have to match.
Every project should start with money in the bank. The standard is a 30 to 50% deposit on signing, balance billed against milestones or weekly progress.
Founders flinch at asking for deposits. They feel like it signals distrust. It doesn't. It signals you've been burned before and built a real business. Mature vendors ask for deposits. Skittish freelancers don't.
A 30% deposit on a $40k project is $12,000 in the bank before the first commit. It covers your first month of engineer cost and removes the catastrophic risk (client ghosts after week 3 with $30k of work delivered and no payment).
For retainer clients, charge the retainer on the first of the month for that month, not in arrears. "January retainer billed January 1, due on receipt" is the structure. Anything else and you're funding the client's cash flow.
The right tool depends on agency size and how much accounting integration you need. The honest comparison:
| Tool | Best for | Monthly cost | Weak spot |
|---|---|---|---|
| Stripe Invoicing | 1 to 5 person agencies, card-paying clients | $0 + 2.9% + 30¢ | Limited project tracking, weak ACH handling |
| Wave | Solo freelancers, simple invoices | $0 | No real automation, limited integrations |
| Harvest | 5 to 20 person agencies, time-to-invoice flow | $13.75/user/month | Aging design, payment is bolted on |
| Bonsai | Freelancers and small studios | $25/month | Less robust for multi-currency |
| FreshBooks | Service businesses, accountant-friendly | $19 to $60/month | Expensive at scale |
| QuickBooks Online | 10+ person agencies, accounting-integrated | $30 to $90/month | Steeper learning curve |
| Xero | International agencies, multi-currency | $15 to $78/month | Reporting is shallow without add-ons |
The stack we recommend for most dev shops in the 5 to 20 person band: Harvest for time and invoicing, Stripe ACH for payment processing, Xero or QuickBooks for accounting. Total cost roughly $200/month for a 10-person shop.
For agencies with international clients, Wise (formerly TransferWise) handles multi-currency receiving accounts cheaper than your bank. You'll get paid in the client's currency and convert at near-spot rates instead of bleeding 2 to 3% to your bank's FX spread.
Net 30 is a relic. It made sense when invoices were paper and checks were mailed. In 2026, with ACH and card payments, Net 7 is reasonable and "due on receipt" is enforceable for established clients.
Default to Net 7. Negotiate to Net 14 for enterprise. Never accept Net 60 or Net 90 without pricing the float into the project (add 2% per 30 extra days).
Late fees are legal in every US state and most jurisdictions, but most agencies don't enforce them because they're afraid of damaging the client relationship. This is backwards. Clients who pay on time don't care about late fees. Clients who pay late will continue paying late until there's a cost.
Set 1.5% per month (18% annualized) and enforce it on day 8 for Net 7 invoices. Automated late-fee assessment is built into Harvest, FreshBooks, and QuickBooks. Turn it on and let the software handle it.
Most agencies "chase" invoices manually, which means they don't chase them. Automate it:
This sequence does two things. It collects faster. And it filters out clients who'll cost more in collection effort than they ever paid in invoices. The agencies that scale cleanly are ruthless about this filter.
If you bill international clients, three things matter that domestic-only agencies ignore.
Currency. Bill in the client's currency if you can afford the FX exposure (Wise multi-currency account handles this). Bill in USD if you can't, but understand you're shifting FX risk to the client, which can sour deals when their currency weakens.
VAT and reverse charge. EU B2B clients use reverse-charge VAT, which means you don't add VAT to the invoice if you have their VAT ID. Get the VAT ID, validate it on VIES, and note "Reverse charge applies, Article 196 of Council Directive 2006/112/EC" on the invoice. Skip this and you'll either eat 20% in VAT or lose the deal.
Sales tax US. If you're a US agency, software services are taxable in some states (Texas, Washington, parts of New York) and not others. If you have nexus in a taxable state and bill clients there, you owe sales tax. Use Avalara or TaxJar to track it. This is one place where the dev agency hourly rate benchmarks 2026 conversation matters less than the sales-tax conversation.
If you run a dev agency and book Cadence engineers to deliver client work, the invoicing structure shifts. You bill the client at your agency rate (say $200/hr or $15k/week deliverable pricing). You pay Cadence the weekly engineer rate ($500 to $2,000 depending on tier). The delta is your margin.
This is the white-label pattern, and it solves the spiky-demand problem agencies hit when a project lands faster than the hiring cycle. Every engineer on Cadence is AI-native by default, vetted on Cursor, Claude Code, and Copilot fluency before they unlock bookings, so the ramp-up time is short enough to slot into a client project without disrupting your delivery cadence.
The invoicing mechanics:
The margin sits in the gap. A senior engineer booked at $1,500/week works out to $37.50/hr if you bill them at 40 hours. You bill the client at $200/hr deliverable-equivalent. That's a 5x markup, in line with traditional agency margins, but without the hiring overhead or the bench cost.
The 10% partner referral on top is the cleaner revenue path for agencies whose clients want to manage engineers directly. Refer the client to Cadence, they sign up, you earn 10% recurring on every week they're active. No delivery management, no bench risk.
Pick the one operational change with the biggest cash impact:
If you run an agency and want a structural way to fold flexible engineer capacity into your delivery model, the Cadence partner program pays 10% recurring on every founder you refer, and lets you white-label booked engineers under your agency brand. It's the cleanest way we've found to scale agency capacity without taking on hiring risk.
Weekly is the default for project work, bi-weekly for retainers, monthly only when a client's procurement department requires it for contracts above roughly $80k/month. Weekly cadence cuts days sales outstanding from a typical 47 days to roughly 14 days, which is the difference between a healthy cash position and floating two months of payroll.
Net 7 is the modern default. Net 14 is acceptable for enterprise. Anything longer than Net 30 should carry a financing premium (add 2% per extra 30 days). Always pair payment terms with a 30 to 50% deposit on signing.
Yes, in every US state and most jurisdictions globally, provided the late-fee terms are in the signed contract or the invoice itself before the work begins. The standard is 1.5% per month (18% annualized). Modern invoicing tools (Harvest, FreshBooks, QuickBooks) assess and bill late fees automatically.
Stripe Invoicing for 1 to 5 person shops with mostly card-paying clients, Harvest for 5 to 20 person agencies that need time-to-invoice flow, and QuickBooks Online or Xero for 10+ person shops that need real accounting integration. The combined stack (Harvest plus Stripe plus QuickBooks) runs roughly $200/month for a 10-person agency.
Use a multi-currency receiving account (Wise is the standard) to avoid 2 to 3% FX spreads on bank conversions. For EU B2B clients, collect and validate the VAT ID via VIES, then note "Reverse charge applies, Article 196 of Council Directive 2006/112/EC" on the invoice. For US state sales tax on software services, use Avalara or TaxJar to track nexus and remit.
Deliverable wherever possible. Hourly billing invites line-item negotiation, punishes you for being fast, and forces uncomfortable conversations about senior engineer time. Reserve hourly billing for discovery phases where scope is genuinely undefined and for reactive maintenance work billed against prepaid hour blocks.