Keeping stock in step
Yessort holds the authoritative stock figure and pushes it to the channel. Even so the two can drift apart — someone edits stock in the store's admin, a push fails, or a channel is connected with listings already in place. Two mechanisms find those differences, and they treat them very differently.
Both are part of the wider family of scheduled re-checks Yessort runs against a channel — see safety nets for why they exist and safety nets and schedules for the full list.
Drift checks: find and report
Drift checks compare what the store is showing against what Yessort holds, and report. They change nothing.
A sample check runs nightly over a bounded subset of listings. It does not need the store's bulk export, so it never contends with a catalogue sync.
A full check runs weekly over everything. It needs the store's bulk export, and a store will only run one of those at a time, so it can find the slot already taken by a catalogue sync.
A check that finds the slot taken is recorded as skipped, which is a normal outcome and not a failure. Every pass is recorded whether it found anything or not, so "this channel has not been checked since Tuesday" is an answerable question — a check that silently stopped running looks exactly like one finding nothing, unless the runs themselves are recorded.
Where a pass covered less than everything — a skipped slot, or a cap on how many variants it looked at — that limitation is recorded with it. A partial sweep reporting "clean" would otherwise read as coverage it did not have.
Reconciles: compare and decide
A reconcile runs at the moments when Yessort first gains a link to a listing and has to establish what is actually there. It compares the store's on-hand figure against Yessort's total for that site — the sum across every location at the site, not one bin.
What it does with a difference depends on which reconcile it is.
When a channel is activated with listings already linked, nothing is booked. Every difference, however small, is held for a person to decide on.
The reason is specific. A store's reported on-hand includes units already committed to open orders. Treating that as stock on the shelf and booking the difference once created over a thousand units that did not exist at a live cutover — and the open orders then fulfilled against them, counting the same units twice.
When the review queue applies a decision, small differences are booked. A difference within 20% of the store's figure, or five units, whichever is larger is written to stock automatically. Anything larger is held.
This one books because a variant Yessort has only just created sits at zero, and the next stock push would send that zero out and wipe the stock the store was correctly showing. The threshold keeps that narrow: a difference bigger than it is far more likely to be a mapping problem than genuine drift, and booking it could push a badly wrong figure to the store and cause an oversell.
Every outcome is recorded
Both reconciles write a row for every difference they meet, including the ones they booked. So a row in the review is not always a job for you — some are a record of what Yessort did, kept visible on purpose.
Repeating a reconcile refreshes the existing row rather than adding another, so a recurring difference shows once with its current numbers.
See resolve a stock reconcile review for working through them, and reconcile and drift statuses for the states.