Shopify changed something on 55% of last year's working days

Every merchant has had the conversation. Orders drop, nobody deployed anything, everyone says they changed nothing, and the week disappears into working out what happened. The unspoken assumption in that conversation is that if nobody on your team changed anything, nothing changed.

So we counted how often the platform underneath you changes. Not opinions, not a vendor claim: all 1,583 entries in Shopify’s public changelog, going back to its first post in August 2018.

The headline numbers

On more than half of last year’s working days, Shopify shipped a change to the platform your store runs on. The surface most likely to cost you money, checkout and payments, was touched every two days.

It is accelerating, and that is the part that matters

The rate has roughly doubled. Full calendar years, from the same feed:

A merchant who built a mental model of “the platform is basically stable” in 2020 was not wrong at the time. They are wrong now, and nothing told them so.

What this does not mean

The reason that matters is arithmetic, not blame. If the platform changes on 55% of working days, then on any given day you investigate, there is a coin-flip chance that something changed underneath you as well as on top of it. “Nobody changed anything” is a statement about your team. It was never a statement about your store.

Method, so you can check it

• Source: Shopify’s own public changelog RSS feed at changelog.shopify.com/feed.xml, parsed on 9 September 2026. It carried 1,583 entries, the earliest dated 3 August 2018.
• “Last 365 days” is a rolling window from the parse date. Yearly figures are full calendar years only, so no partial year is compared against a whole one.
• Working days are weekdays, not adjusted for public holidays in any country, so the 55% figure is a slight under-estimate of weekday density.
• “Touching checkout or payments” is keyword matching on the title and description (checkout, payment, shop pay). It is approximate in both directions: it will catch entries that only mention checkout in passing, and miss ones that affect it without naming it.
• What we deliberately did not publish: a count of breaking changes. We tried, using the obvious keywords, and the matches were not trustworthy enough to report. A number we cannot defend is worse than no number.

What to actually do about it

The practical takeaway is a change to the first question you ask, not a purchase. When a number moves and your team is certain nobody touched anything:

• Check the platform changelog for the two days before the drop, not just your own deploys. Shopify’s is public and it is the cheapest lead you will get.
• Widen “we” to include everything that can change without a deploy — apps that auto-update, a theme edit, a price, a shipping rule, an email flow someone paused.
• Write down changes as they happen, not when something breaks. A shared channel or document works and costs nothing. Be warned it usually dies within a month, because the discipline falls on the people with the least reason to keep it up.

Where Rearview fits

We built Rearview because that last bullet keeps failing. It puts every production-affecting change across your tools onto one timeline you can ask in plain English, and platform changes like the ones counted above land on it alongside your own, so the afternoon a number moved shows both halves.

Related reading: when conversion drops and nothing changed and tracking what changed in your Shopify store.