Your client's app just had its best month ever. Signups tripled, the workflows you built are humming, and the founder is thrilled. Then the platform invoice lands: two million workload units over the plan allowance, billed at thirty cents per thousand. That is a $600 overage on top of a $119 plan — and if your contract says "hosting included" inside a fixed monthly fee, that $600 comes out of your margin, not the client's budget. Their success just became your loss, and nothing in your scope of work anticipated it.
This is the defining pricing risk of running a no-code agency in 2026. Bubble bills on consumption, Webflow meters bandwidth and content volume, and both reprice periodically. A fixed fee pretends those costs are predictable. They are not. The agencies that survive this treat platform spend as what it actually is: a pass-through cost that flows from the vendor, through your books, to the client — with your real income being the markup and the management fee, never the reimbursement itself.
How Bubble actually bills you in 2026
Bubble's headline prices look tame. On annual billing, web plans run $29 a month for Starter, $119 for Growth, and $349 for Team, with a free tier at $0 and Enterprise priced on request. Monthly billing costs more: $32, $134, and $399 respectively. But the plan fee is the smallest part of most production bills. The real meter is the workload unit.
Every paid plan includes a monthly workload-unit allowance: 175,000 units on Starter, 250,000 on Growth, and 500,000 on Team. The free plan includes 50,000 but cannot deploy a live app. A workload unit measures server work — database searches, workflow runs, API calls, page loads, file operations — and consumption varies wildly by operation. A simple page load costs roughly half a unit to two units; a single complex search can cost 10 to 50. An app with chatty workflows or unoptimized searches can burn through a Starter allowance long before its traffic looks impressive.
When the allowance runs out, the overage rate without any add-on tier is $0.30 per 1,000 units. One million extra units in a month — easy for a growing app with heavy backend workflows — costs $300. Pre-purchased workload tiers lower the unit rate substantially, down to about $0.10 per 1,000 at the largest tier, but you have to buy the tier before the spike, and tiers are a monthly commitment of their own.
Three mechanics make this treacherous for agencies specifically. First, consumption is genuinely hard to predict in advance; two apps with identical user counts can differ by an order of magnitude in workload depending on how their workflows are built. Second, overages are switched on by default, and Bubble emails the account owner at 75 percent and 100 percent of the allowance — but if your agency holds the account, that email goes to you, not the client whose users caused the spike. Third, switching overages off does not cap the bill gracefully: the app goes offline when it hits the limit until overages are re-enabled, a tier is bought, or the next cycle starts. "The site went down because we hit a usage cap nobody was watching" is not a conversation you want to have with a client.
Webflow has the same shape with different meters
Webflow does not use workload units, but its pricing has the same volatility profile from an agency's perspective: a modest base fee plus meters that move with the client's success.
In May 2026 Webflow merged its old CMS and Business site plans into a single Premium plan: $25 a month per site on annual billing ($39 monthly), covering 300 static pages, 20,000 CMS items, and 50 GB of bandwidth. The Basic plan at $15 a month covers a custom domain and 10 GB of bandwidth but no CMS at all. Anything beyond the included bandwidth is sold as paid add-ons, up to 2.5 terabytes — so a client's viral launch or a heavy quarter of image and video traffic converts directly into a bigger Webflow invoice.
And the site plan is only one of three bills. Workspace plans for design collaboration, extra designer seats, additional locales, and e-commerce transaction capacity are all billed separately, on top of the per-site fee. A client who asks for "just one more language" or "access for two more editors" is asking you to increase a metered stack — and if your contract quotes a single flat monthly number for "website plus hosting," every one of those yeses shrinks your margin.
Both platforms also reprice. Bubble adjusted Starter pricing in October 2026; Webflow restructured its entire site-plan lineup mid-year. Any multi-month fixed quote that embeds today's platform prices is a bet that the vendor will sit still. That bet has been losing for years.
The contract mistake: bundling metered costs into a fixed fee
The failure mode is almost always the same sentence, buried in a proposal or a master services agreement: a fixed monthly fee that "includes hosting." It reads as simplicity for the client. In practice it makes your agency the underwriter of vendor volatility — you collect a fixed premium and absorb unlimited metered downside.
Industry reporting on agency builds puts typical Bubble project fees anywhere from $20,000 to $250,000, and notes the specific margin trap: when you deliver a client app and the client's business succeeds, consumption rises, and either the client pays more they never budgeted for or the agency absorbs the increase. Fixed-price-on-vague-scope is already the most common source of agency-client disputes in software services generally; adding a metered third-party bill inside the fixed price turns every usage spike into a renegotiation.
The same trap appears one level down in AI and automation work, where agencies that bundle API and token costs into flat retainers discover that a vendor price change becomes their loss overnight. Platform spend behaves exactly like token spend: third-party-metered, client-driven, and outside your control. Costs with those three properties do not belong inside a fixed fee. They belong in a pass-through clause.
There is a second, quieter cost to bundling: it hides the numbers from the client. A client who never sees the workload-unit trend has no reason to approve the optimization work that would cut it — restructuring a wasteful search, moving heavy logic to backend workflows, buying a workload tier before the spike instead of paying raw overage after it. Transparent pass-through billing turns usage into a shared dashboard you both watch, which is the only arrangement where optimization work sells itself.
The fix: make platform spend a pass-through cost
Pass-through costs are expenses your agency pays on a client's behalf — vendor subscriptions, usage bills, license fees — that get billed back to the client, either at cost or with a handling markup. The concept is standard in agency accounting: the client's own costs flow through your books without becoming your income. Applied to no-code platforms, there are three workable models. Pick one per client, in writing, before the app goes live.
Model 1: the client holds the vendor account and pays directly. The client creates their own Bubble or Webflow account, adds their own card, and owns the plan, the tiers, and the overages. Your agency gets invited as a collaborator or guest editor. This is the cleanest arrangement — the metered bill never touches your books — and it is the right default for larger clients with their own finance function. The known downside: clients sometimes under-invest in their plan tier to save money, which creates performance bottlenecks and limit-hit outages that you then have to diagnose and explain. If you choose this model, put a minimum-tier recommendation in the handover docs and a clause stating that running below it is at the client's risk.
Model 2: the agency pays and rebills at cost plus a markup. You hold the vendor account, pay the platform bill on your own card, and invoice the client each month for the exact platform spend plus a handling fee — typically 10 to 20 percent of the pass-through, the range most agencies use to cover vendor management, reconciliation, and fronting the cash. This model suits smaller clients who want one invoice and zero vendor admin. Your contract must state the markup percentage, the invoicing cadence, and — critically — that vendor price increases flow through automatically. Without that last sentence, every Bubble or Webflow repricing reopens your pricing.
Model 3: a bundled care plan with a usage cap and an overage clause. You quote a flat monthly retainer that includes platform spend up to a stated cap — for example, "includes the Growth plan plus up to $150 a month in workload overages" — with anything above the cap billed through at cost plus markup. This gives the client budget predictability for normal months while keeping spike risk where it belongs. The cap must name both the plan tier and a dollar overage allowance; "includes hosting" without numbers is Model 0, the trap you are trying to escape.
Whichever model you choose, five clauses belong in the statement of work: who holds the vendor account and whose card is on file; who pays overages and tier upgrades, and above what dollar threshold you need written approval before buying a tier; the markup percentage on rebilled spend; a monthly usage report you send with the invoice, showing allowance, consumption, and trend; and the client's acknowledgment that vendor list-price changes pass through without a contract amendment. That fifth clause is the one agencies forget and the one they later need most.
Book it right: reimbursements are not your revenue
Once platform spend flows through your agency, the bookkeeping has to keep it visibly separate from what you actually earn. Suppose Bubble charges your card $300 for a client's October overage — one million extra units at $0.30 per thousand — and you rebill the client $345: the $300 at cost plus a 15 percent handling fee. Your real income from that invoice is $45, not $345. Booking the full $345 as revenue overstates your top line, inflates any revenue-based tax or valuation math, and makes a subsequent refund or credit a revenue reversal instead of what it really is: client money passing through.
The accounting standard agrees. Under ASC 606, the principal-versus-agent framework asks whether you control the service before it transfers to the customer. You do not control Bubble's compute or Webflow's bandwidth — the vendor does — so on rebilled platform spend your agency is an agent, and the correct presentation is net: the markup as fee income, with the reimbursement washing through matched pass-through lines. Concretely, keep two pairs of accounts: pass-through revenue matched against pass-through expense, and fee income standing alone. The pass-through pair should net to roughly zero every month, give or take timing; the markup lives in fee income where your margins are measured.
Here is what that looks like as plain-text accounting. The agency pays Bubble's $300 overage on its own card, then invoices the client $345:
2026-10-01 open Assets:Checking:Operating USD
2026-10-01 open Assets:Receivable:Client-Acme USD
2026-10-01 open Expenses:Passthrough:Bubble-WU USD
2026-10-01 open Income:Passthrough-Reimbursement USD
2026-10-01 open Income:Fees:Hosting-Markup USD
2026-10-31 * "Bubble" "October WU overage, client Acme app, 1M units at 0.30 per 1K"
Expenses:Passthrough:Bubble-WU 300.00 USD
Assets:Checking:Operating -300.00 USD
2026-10-31 * "Acme Corp" "October platform reimbursement plus 15 percent hosting admin fee"
Assets:Receivable:Client-Acme 345.00 USD
Income:Passthrough-Reimbursement -300.00 USD
Income:Fees:Hosting-Markup -45.00 USDWalk through the logic. The first transaction records the vendor charge as a pass-through expense, not as your own software cost — parking it in ordinary operating expenses is the mistake that makes your cost base look bloated and your margins look thin. The second transaction splits the client invoice: $300 clears against the reimbursement line, which nets with the expense to zero, and only the $45 markup lands in fee income. If the client disputes the overage or Bubble issues a credit, the adjustment hits the pass-through pair and your reported revenue never moves. The docs cover account-structure patterns like this in more depth, and Fava renders the pass-through pair as its own slice so you can confirm it nets to zero at a glance each month.
Two bookkeeping habits make this stick. First, reconcile platform spend per client every month, against both the vendor invoice and your rebill — workload dashboards drift from invoices more often than you would expect, and the month you skip is the month a tier renewed unnoticed. Second, never let a client's platform balance ride on your card across quarters; invoice rebills monthly with the same payment terms as your fees, because an unrebilled $600 overage aging in your expenses is an interest-free loan you never agreed to make.
Your next step: audit the contracts you already signed
Pull your current master agreements and statements of work this week and search for the sentence that creates this risk: any variant of "hosting included" attached to a fixed number. For each one, check which of the three models it actually describes — client-pays, rebill-with-markup, or capped care plan — and whether the five clauses are present: account holder, overage responsibility, markup percentage, monthly usage reporting, and automatic pass-through of vendor price changes. Most agencies find at least one client on an uncapped bundle and at least one contract silent on repricing.
Then fix them in renewal order, starting with the client whose app is growing fastest — that is where the next spike lands. Move new builds to client-held accounts or explicit rebill terms from day one, put the usage dashboard on the monthly invoice, and let the metered bill do what it is supposed to do: rise and fall with the client's success, on the client's budget, while your margin stays yours.
