Annual vs monthly commitment: the 20% you pay for flexibility

Someone picked the term on the day the subscription was bought, and nobody has looked at it since. That is the normal state of a Microsoft 365 tenant. The choice quietly sets what a seat costs, how long you are stuck with it, and whether a leaver in March saves you anything or nothing at all.

Under the new commerce experience there are two independent settings that people talk about as if they were one: term duration (how long you are committed) and billing frequency (how often you are charged). Confusing them is expensive in both directions — paying the monthly premium for flexibility you already gave away, or committing annually on a team whose headcount moves every quarter.

Two knobs, and only one of them is the commitment

A new commerce subscription carries a single product, a single term and a single billing frequency — you can buy for 1, 12 or 36 months, and be billed upfront, monthly or annually. Those are separate fields, and only the first one is a commitment.

Row two is where the money leaks in the other direction: it looks monthly on the invoice, and behaves annually on the day you want to cut seats. Microsoft is explicit that after the seven-day window "the partner is billed for the full term even if the customer stops paying or using the subscription", and adds in brackets: applicable to any billing plan. Spreading the payment changes nothing about the commitment.

Where the flexibility actually lives: seven days

Both terms have the same escape hatch, and it is small. For new commerce offers a subscription can be cancelled with a prorated refund within the first seven days (168 hours) of any term — cancel inside 24 hours and the refund is full. Seat counts follow the same clock: licences can be increased at any time, but decreased only within seven days of when those licences were added, whether they were added at purchase, at renewal or mid-term.

So the real difference between the two terms is not price, it is how often that door opens.

Mixing terms, on purpose

The useful move is not picking one term for the tenant. It is splitting the headcount into the part that is stable and the part that is not, then buying each on the term that matches.

• Baseline on annual. The seats you would still need after a bad quarter — finance, ops, the people who have been there for years. This is usually 70–90% of the tenant and it earns the annual price.
• Volatile headcount on monthly. Contractors, seasonal staff, a team mid-hire, anything attached to a project with an end date. Here the premium is cheap: one E5 dropped after two months rather than carried for ten pays for a lot of monthly uplift on its neighbours.
• Never the leavers you already know about. A seat that has a departure date inside the next twelve months should not be sitting on an annual term, and neither should the ones nobody signs into — those are a different problem with the same invoice.

One structural catch when you mix: end dates. Coterminosity lets you align a new subscription to an existing one's end date, but annual and triennial term subscriptions can't be coterminated with a monthly term subscription — a monthly one can align to anything, not the other way round. Plan the annual group's renewal date first and let the monthly seats orbit it.

Changing your mind: the renewal is the only door

Term duration is not editable mid-term. What you can do is queue the change for the boundary: a scheduled change at renewal can modify the billing term, the billing frequency, the licence count, the SKU or the coterminous date. Two conditions and one trap:

• The subscription must be active and autorenew must be on — turning autorenew off to "keep options open" is what removes the option.
• Saved scheduled changes are deleted by mid-term changes: cancelling, turning off autorenew, changing a quantity, upgrading a SKU or converting a trial. Adding five seats in month nine silently erases the term change you queued in month two.

And the renewal itself takes up to 24 hours, after which the seven-day window reopens — timed from renewal completion, not from the calendar date. If your plan is "we'll cut the unused seats at renewal", that is a diarised week, not a vague intention.

What the tenant will and won't tell you

Here is the honest limit. The commercial terms live in Partner Center, not in the tenant. The companySubscription resource gives you skuPartNumber, totalLicenses, status, isTrial, ownerTenantId (which partner sold it) and the one date that matters operationally — nextLifecycleDateTime, the moment the subscription moves to its next state if it isn't renewed. There is no term-duration property and no price. You can see when the door opens; the term itself you confirm against the price list.

That date is still the whole game, because the decision is only actionable in the week around it. GraphPaaS reads those subscriptions per tenant on a schedule and keeps them next to the seat data — so "which subscriptions renew in the next 60 days, and how many of their licences are doing nothing" is a page rather than a fire drill in the last three days of the term.