Subscription Billing: Proration, Dunning and Churn

Proration means charging a customer only for the part of a billing period they used each plan. Recurly gives the standard formula:

Prorated amount = (days on the plan / days in the billing period) x full period price

A mid-cycle change produces two lines: a credit for the unused time on the old plan and a charge for the remaining time on the new plan. Using made-up numbers on a 30-day period, a customer on a 90 per month plan who upgrades to a 150 per month plan on day 20 gets:

That pattern follows Recurly's worked example, where the credit for the old plan and the charge for the new plan are calculated separately and then combined.

Billing systems add details that matter in practice. Stripe's documentation notes that it calculates prorations down to the second by default, and that you can change the granularity to days, hours, weeks or months. It also warns that negative prorations are not automatically refunded and positive prorations are not immediately billed: by default, both roll into the next invoice. If your customer expects to pay the difference today, you must invoice it explicitly.

Seat changes are where spreadsheets break. Turnstile describes the common error of collapsing several seat changes into one calculation: a customer who adds five seats on day 10 and three more on day 20 needs two separate proration lines, not one. Let your billing system do this math. Hand calculations drift, and every drift becomes a support ticket.

The formula is fixed. The policy is yours. Four decisions shape how customers experience plan changes.

Do upgrades take effect immediately? Almost always yes. PayPro Global recommends prorating upgrades so customers get instant access without paying for two plans at once. Making a customer wait until renewal to get what they want to pay for is leaving money on the table.

Do downgrades take effect immediately? This is a real choice. PayPro Global notes that some businesses do not prorate downgrades and apply them at the end of the paid cycle. End-of-cycle downgrades are simpler and avoid issuing credits. Immediate downgrades feel fairer to the customer. Pick one, state it on the pricing page and apply it the same way every time.

Are credits refunded or carried forward? Turnstile observes that most B2B SaaS companies apply credits to future invoices rather than issuing refunds. Refunds cost you processing time and sometimes fees. Carried credits keep the customer and the cash.

Does the billing date move? Treat an upgrade as a change to the existing subscription, not a cancel-and-restart. Turnstile recommends this to preserve the original billing anniversary date. A moving anniversary confuses customers and complicates your revenue reports.

Write these four answers down in one paragraph of policy. Support, sales and engineering should all be able to point to the same text.

Every recurring charge is a new authorization request to the customer's bank, and some fail. The reason for the failure decides what you should do next. Chargebee splits failures into two groups:

Chargebee warns that retrying a hard decline wastes attempts and risks card-network penalties. Stripe's retry engine follows the same rule: it cannot automatically retry a payment after a hard decline code without a new payment method.

Paddle lists the common causes behind these declines: expired cards, hard declines after fraud, soft declines from maxed-out credit limits, out-of-date billing information and charges not flagged as recurring. The last one is a billing configuration mistake, not a customer problem. If your charges are not marked as recurring, the issuer may treat a monthly charge as unusual and decline it.

Dunning is the automated process that recovers a failed renewal. Solidgate describes it as a structured sequence of automated retries and customer communication that escalates in urgency over a defined recovery window. A dunning system has three parts.

1. Retries. The system re-attempts the charge on a schedule. Fixed schedules work, but smarter ones time each retry using signals about when that card is most likely to clear. Stripe's Smart Retries setting, as documented on 2026-10-17, lets you choose a policy window and recommends a default of 8 tries within 2 weeks. Whatever tool you use, match the window to your billing cycle: a two-week window fits monthly plans, while annual contracts justify a longer one.

2. Messages. Customers need to know when their action is required. Zuora's example sequence escalates from email to an in-app notification or SMS if the first email goes unopened. Solidgate recommends that every message carry a direct link to update payment details so the fix takes one step. Start polite and specific: "Your card ending 4242 was declined. Update it here." Become more direct as the window closes.

3. A grace period. Keep access live while recovery runs. Solidgate notes that a subscriber whose card clears on the third retry should never lose service. Paddle's example shows why: a marketing tool that locks out a customer the day a charge fails stops their campaigns, and the customer's irate call to support turns an accident into a reason to leave.

A simple dunning plan for a monthly product looks like this:

Day 0   Charge fails. Soft decline: schedule retries. Hard decline: skip retries.
Day 0   Email: "Payment didn't go through" + update link. In-app banner on login.
Day 3   Retry (soft only). Second email if card not updated.
Day 7   Retry. In-app message or SMS.
Day 10  Retry. Email: "Your account will pause on day 14."
Day 14  Final retry. If it fails: suspend, keep data, move to win-back.

When the window closes without payment, suspend rather than delete. A customer who comes back next month should find their work where they left it.

Zuora puts it plainly: the best dunning email is the one you never have to send. Three prevention tools stop failures before they happen.

• Account updater services. Card networks offer services that refresh card details when a bank issues a replacement card. Chargebee describes an account updater as something that refreshes expired or reissued card details before the next charge, so the payment clears with no customer action. Turn it on if your provider supports it.
• Pre-dunning messages. You know when each card on file expires. Butter Payments recommends automated reminders to update card information before it expires. Paddle suggests delivering them inside your app, while customers are getting value, because that is when they are most willing to act.
• Correct transaction flags and more payment methods. Mark recurring charges as recurring, and offer payment methods that do not expire, such as bank debits where your customers use them. Solidgate's guide lists network tokenization, account updater and broader payment method coverage as the prevention layer that comes before any retry logic.

Churn is the share of customers or revenue you lose in a period. A single churn number hides the two causes that need different owners.

Paddle defines them: voluntary churn is when a customer actively decides to end their subscription, and involuntary churn is when they churn because of payment failure, outdated payment information or insufficient funds. Zuora's glossary, read on 2026-10-17, puts involuntary churn at 20% to 40% of all churn in subscription businesses. Treat that range as a reason to measure your own split, not as a benchmark to compare against.

Track these numbers every month, split by cause:

Voluntary churn responds to product work: better onboarding, features customers asked for and pricing that matches value. Paddle recommends acting on customer feedback and segmenting customers to understand why each group leaves. Involuntary churn responds to billing work: account updaters, pre-dunning, smart retries and a grace period. Do not send an exit survey to someone whose card expired. They did not leave, and they deserve a link to fix their card, not a question about why they are unhappy.