How Often Should Credits Reset, and Do They Roll Over?

Idea in Brief

The Problem. Every credit-based contract has to answer a baseline question regardless of whether usage is seasonal: does an unused credit balance expire at the end of the period, or carry into the next one?

The Instinct That's Wrong. Reaching for a burst allowance or a spike-pricing clause before the underlying reset-and-rollover default is even settled.

The Fix. Set the reset cadence and rollover default first — it's a different clause than burst pricing, which only matters for usage spikes that don't fit any reset cadence at all.

The Question Every Contract Has to Answer

Before a deal desk gets anywhere near seasonal usage or burst behavior, every credit-based contract has to settle one default: what happens to a credit balance the customer didn't use. Two regimes have hardened around this question, split cleanly by buyer type.

For self-serve and subscription-embedded AI credits, the standard is monthly reset with no rollover, and it's documented first-party across the market. Anthropic's Claude usage credits "refresh monthly with the billing cycle and do not roll over." GitHub Copilot's new AI-credit model states unused credits "do not roll over from month to month" and "reset at the start of each monthly billing cycle." HubSpot's credit system works the same way: "on your reset date, your account will automatically go back to your original monthly credit limit."

For enterprise capacity deals, the picture is different — rollover is negotiable and increasingly expected, though still not the contractual default. Snowflake explicitly discloses that customers "may have the ability to roll over unused capacity to future periods, generally upon the purchase of additional capacity at renewal," and frames this as a differentiator versus ratable subscription software that "may not permit rollover." Metronome's enterprise commit guide confirms the baseline underneath that flexibility: "most commit contracts are 'use it or lose it,'" with negotiated exceptions — "a small portion (e.g. 10–20%) can carry forward if the customer renews," or a monthly commit can allow a single period of rollover with a defined cap.

The Exhibit: Credit Reset & Rollover

Plain-English variants of the same order-form clause, sized for a roughly $100K Order Form. All three share one clause title — Credit Reset & Rollover — so switching tiers means swapping the body text only.

Preferred: Standard Monthly Reset, No Rollover

Credits allocated under the Total Commit Amount are available for use during the applicable Term and do not carry over to any subsequent Term. Unused credits at the end of the Term expire with no cash or credit value.

Use this when: this is the market default and should be the opening position on nearly every deal — it's the documented standard across self-serve credit products (Anthropic, GitHub Copilot, HubSpot) and Metronome's own "use it or lose it" baseline for enterprise commit contracts. Most customers won't push back on this if the initial commit is right-sized.

Fallback: Capped, Time-Bound Rolling Reserve

Unused credits from any given month within the Term may be applied to cover usage in the immediately following month only ("Rolling Reserve"), subject to a maximum carryover of one (1) month's allotment at any time. Credits unused for more than two (2) consecutive months expire.

Use this when: the account has a large, strategic enterprise footprint with genuinely uneven usage across the year — this is the Snowflake/Metronome enterprise pattern, real rollover but capped and time-bound rather than open-ended. It changes revenue-recognition timing enough that deal desk should see it before it's offered.

Approval-Required: Full, Indefinite Rollover

Unused credits from any period within the Term carry forward without limitation or expiration until fully consumed or the Term ends.

Use this when: essentially never without explicit finance sign-off — indefinite, uncapped rollover breaks the breakage-revenue assumptions built into the pricing model. This tier exists in the ladder specifically so a rep has a clearly labeled reason to escalate rather than freelance the concession.

Burst Pricing: A Different Clause for a Different Problem

Once the reset-and-rollover default is set, a separate question remains: what happens when usage doesn't fit any reset cadence at all — a short, sharp spike rather than steady seasonal drift. That's a burst allowance, not a rollover policy, and it's a distinct clause slot.

The Exhibit: Burst Allowance

All three share one clause title — Burst Allowance — so switching tiers means swapping the body text only.

Preferred: Modest Grace Band

Customer may exceed its Monthly Credit Allotment by up to ten percent (10%) in any given month at no additional charge ("Burst Allowance"), provided such excess usage does not recur for more than two (2) consecutive months. Usage beyond the Burst Allowance will be billed per the parties' True-Up Clause.

Use this when: usage is expected to be mostly steady with occasional one-off spikes — a self-funding grace band protects against those spikes without opening the door to sustained overage at no cost.

Fallback: Accrued Burst Reserve

Customer may accrue unused credits from months where actual usage falls below the Monthly Credit Allotment into a Burst Reserve, up to a maximum of one (1) month's allotment. Accrued Burst Reserve credits may be drawn down in any subsequent month to cover usage exceeding that month's allotment, at no additional charge.

Use this when: usage is genuinely uneven and the customer can point to real underuse in other months — this is L.E.K. Consulting's "burstable reserve" model, where the customer earns the buffer through underuse elsewhere rather than getting it for free.

Approval-Required: Free Band Movement Tied to Renewal

Customer may exceed its current Band without additional charge for the remainder of the current Term, provided that the higher usage Band observed will govern pricing and minimum commitment terms at renewal.

Use this when: the account's usage has genuinely stepped up and the business wants to avoid friction now — L.E.K. calls this "free band movement." The concession is real today, but it locks in a higher renewal anchor, so deal desk needs visibility into that renewal implication before it's offered.

The Vantage Point

Revolear sets up dozens of new Order Forms every quarter for usage-based businesses and assists our customers' sellers in structuring exactly these two clauses. The most common drafting mistake we see isn't picking the wrong tier — it's conflating rollover and burst into a single ask when a customer says "we don't want to lose unused credits," when in fact one customer means a seasonal usage pattern and another means a single unpredictable spike, and those need different clauses entirely.

The Takeaway

Settle the reset-and-rollover default before reaching for a burst allowance — they solve different problems, and a burst clause bolted onto an undefined rollover policy just recreates the "use it or lose it" ambiguity one level up.

Related in this series: this post is part of Revolear's Usage-Based Contracting series on the business clauses governing the primary subscription term. Read more from the series:

What Contract Terms Are Becoming Standard in AI/Credit-Based SaaS? (pillar post)

Making Usage Revenue Look More Like SaaS

Renewal Is Not a Blank Slate