The Switch

Timing a switch around the renewal date

A switch has one external deadline and it is the renewal date. Everything else is your own pace, which is why the schedule tends to drift until a charge arrives and turns the whole thing into a rush.

Find out the actual date, work backwards from it, and keep the export and verification steps out of the last few days.

Find out what the date is, rather than estimating

Most people are wrong about their renewal date by a week or two, and the ones on an annual term are often wrong by months.

The account’s billing or subscription area states it. Look at three things: the next renewal date, the term you are on, and what the account reverts to when the paid term ends. All three vary by product and by how you signed up, and the last one matters most for planning.

Do this before scheduling anything else. The whole plan hangs off it, and it is a thirty-second lookup.

Work backwards, not forwards

The natural mistake is to plan forwards from today and hope the renewal is far enough away.

Plan backwards instead. Put the renewal on a calendar, then place the steps before it:

  • cancellation: a few days before renewal, not on the day
  • shared links replaced and integrations revoked: before that
  • final export, verified: before that
  • end of the parallel period: before that

The gap between cancellation and renewal exists because cancellation is not always instantaneous in effect, and because discovering a problem on renewal day means either a rushed decision or a charge you did not intend.

Do not put the export near the deadline

The export step is the one that must never be compressed, because it is the only step with a waiting period you do not control.

Processing time varies by product, by account size, and over time. It might be minutes; it might be considerably longer. You cannot plan around a number nobody can give you, so plan around the shape: request early in the final window, wait for it to arrive, verify it, and only then move on.

Two days of margin turns the unknown into a non-issue. Two hours does not.

Annual terms change the arithmetic

If you are on an annual term, the calculation is different in one specific way: the amount of money already spent is not recoverable by hurrying, and it should not shape the sequence.

Do not assume a refund is available, and do not assume one is not — that varies by product and by jurisdiction, and the account’s own billing terms are the only place to check. What is worth knowing is that a partially used annual term usually means there is no deadline pressure at all, because the renewal is months away.

That is the good case: run the parallel period properly, export properly, and cancel whenever you are done, with the paid term simply expiring. Nothing about a switch requires the account to end on any particular day.

Confirm what you are cancelling

A consumer subscription and API access are separate products with separate billing. Cancelling one does not stop the other, and if you set up API access at some point — for a script, a tool, a one-off experiment — that is a distinct thing with its own usage and its own charges.

Check both. The failure here is small in money and annoying in shape: a charge arriving on an account you believe you closed, two months after the switch was finished.

Team and shared plans add a third case, where cancelling a seat and cancelling the plan are different actions with different consequences for other people.

What moves

WHAT MOVES — billing and timing

  · The paid term you already bought
                    → runs to its end date. Cancelling
                      usually stops renewal, not access.

  · Your archive
                    → moves out before cancellation, or
                      not at all.

  · Access to generate a new export
                    → commonly STAYS BEHIND with the
                      account. This is why export
                      precedes cancellation.

  · API access under the same identity
                    → STAYS BEHIND, still billing.
                      Separate product, separate
                      cancellation.

  · An expired export request at the
    deadline
                    → STAYS BEHIND. No time left to
                      reissue.

  · Cancelling before verifying the export
                    → IRREVERSIBLE in practice. The
                      deadline is not worth this.

Let the date decide, not the feeling

The renewal date is genuinely useful as a forcing function, and that is the argument for setting the end of the parallel period against it rather than against a sense of readiness.

A period with no fixed end runs for months, costs a subscription each cycle, and decides nothing — because splitting work across two assistants means neither accumulates context and both keep feeling provisional. The date resolves that without requiring you to feel certain.

If the deadline arrives and you are not ready

Let it renew.

One extra cycle is cheap compared to any of the alternatives: a rushed export nobody verified, an unverified archive, links you did not replace, a cancellation ahead of the sequence that protects the data. The cost of one more month is a known, small, recoverable number.

The cost of doing the irreversible steps in the wrong order to hit a date is not recoverable at all, and the money saved is never the size of the thing lost.

billingcancelling