Inventory health findings

When you finish an inventory count, Rinvy runs a health check on the results. It compares what you counted against what it expected to see, based on your previous count, the receipts and waste logged in between, and each product's usage rate. When something doesn't line up, it flags the product, describes the likely cause in plain language, and shows the estimated dollars at stake.

What it looks for

Each finding covers one product and one likely explanation:

Missing receipt? — there's meaningfully more on the shelf than the numbers can explain, and no delivery was logged in the window. The most common cause: a delivery that arrived but never made it into Rinvy — logging it brings the books back in line with the shelf.
Unlogged waste? — stock dropped further than receipts and normal usage explain, with nothing in the waste log. Was there spoilage or a dropped case nobody recorded?
Count error? — stock came in higher than expected even though deliveries were logged. The previous count may have been low, or a delivery was logged short.
Usage shift — consumption ran ahead of its usual pace. Not an error, but worth knowing about.

Every finding shows the estimated dollar impact and the numbers behind it (expected vs. counted, received, waste logged).

Checked twice before you see it

Alerting on every deviation trains everyone to ignore the alerts, so Rinvy is deliberately conservative:

Rule-based checks only flag deviations that clear meaningful dollar and percentage bars, and stand down entirely when the data is too thin to judge (no usage history, too few counts, too long between counts).
Every candidate finding is then reviewed by AI before it surfaces. If a mundane explanation fits the numbers as well as the flagged cause — normal week-to-week swing, count timing, a skewed average — the finding is demoted instead of shown.
At most the top 5 findings per count surface, ranked by dollar impact.

Where findings show up

Right after the count — the count-complete screen shows what the review found (or "No anomalies found"), usually within a few seconds of finishing.
Dashboard — under Needs attention, alongside low stock and price alerts. Tap a row for the details and the recommended action.
Push notification — Owners and Managers get a push when a count surfaces a critical finding (the person who did the count doesn't — they just saw it in the app).
Reports → Health — the full cross-session history, filterable by severity and status.
Count snapshot — each past count's page lists everything recorded for that count.

Acting on a finding

Each finding carries a suggested action that deep-links to the fix — log the missing receipt, record the waste, or open the product's report. The receipt and waste forms open in record-only mode: your count already measured what's on the shelf, so the corrective entry updates the books without changing stock levels again. Acting on it resolves the finding automatically: logging a receipt clears a "Missing receipt?" finding for that product, and logging waste clears an "Unlogged waste?" one.

Dismissing

If a finding is wrong or not worth chasing, dismiss it from the detail view. Dismissals are respected: the same finding type for the same product won't resurface for a week, and later reviews take past dismissals for that product into account.

What it needs to work

Two completed counts. The first count of a product is a baseline — analysis starts from the second.
A few counts of history. Products with fewer than 3 counts in the last 30 days are observed but not alerted on; the usage expectation isn't stable enough yet.
Costs on products. Dollar impact comes from each product's case cost; products without cost data can't be ranked.
Counts no more than ~2 weeks apart. The longer the gap, the fuzzier the expectation — beyond 14 days, findings are recorded but not surfaced.