Check what the sort actually did

Handing your collection order to an app is a real act of trust, and the honest way to earn it is to let you check the result rather than ask you to assume it. Everything below is a place in QSortby where you can see what happened and why — four of the five are on the Free plan.

This page is a map of those places. Each one is covered in depth elsewhere; the point here is knowing they exist before you need them.

Before anything publishes

On a collection’s Setup tab, turn Push to Shopify automatically off. A newly computed order is then staged instead of going live, and you approve it from the Preview tab — a side-by-side of what is live now against what the next run would produce, with added, removed and reordered counted for you.

Leave it on once you trust the rule. The point is that the first few runs do not have to be taken on faith. See Set the time window & keep it fresh.

The rule itself is yours to read

There is no hidden model deciding your order. On the Setup tab you can see the criteria the collection ranks by and the window it measures, in the same words you picked them in, and change either at any time.

That matters more than it sounds. If a result surprises you, the rule is right there — you can disagree with it and fix it, rather than concluding the feature is broken. See How ranking works.

Why a product was matched to another

Open any product and look at the Similar tab. Every row names the signal that produced it — same product type, shared tags, shared collections, similar price, sells well, same vendor — so a match you disagree with points you at the data behind it.

Only signals that could actually be compared are listed. If neither product has tags, the tag row is left out rather than shown as zero, because zero would read as “no overlap” when it means “nothing to compare”. See One product: what it sells with, and why.

Which rule was live, and when

Analytics → Sort projects keeps the ranking configuration on record with its dates. When a week goes unusually well or badly, you can tell which rule was actually running at the time instead of reconstructing it from memory.

What it earned — and what that number is not

Analytics → Impact reports influenced revenue: orders where a shopper interacted with a QSortby feed before buying. The word is deliberate. Attribution here is last-touch, which means the figure tells you the sort was involved in those orders — not that it caused them.

The same care applies to experiments. A result is reported as directional, never as proof, and the minimum sample per arm stays on screen whether or not you have reached it, so you know when a difference is still too early to act on.

See Experiments & attribution.

What is not here yet

Two things merchants reasonably expect that QSortby does not do today, said plainly so you do not go looking:

• Restoring a previous product order. Collections you archive can be restored, and configuration changes are on record — but there is no one-click rewind to the exact order a collection had on a given day.
• A per-product score breakdown for collection ranking. The reasoning behind a similar-product match is shown; the contribution of each criterion to a single product’s rank is not surfaced in the dashboard.