How it works
Arithmetic on your own order history. Nothing more.
Six steps from an order export to a short list of accounts that went quiet — and a way to check the whole method against your own past before you trust it.
Load your order history
Connect Prophet 21, Eclipse, or QuickBooks, or drop a CSV export. All that is strictly required is a customer, a date, and an amount, on every order you have ever shipped.
Learn each account's rhythm
Every account gets its own baseline: how many days normally pass between its orders. Orders placed within three days of each other are treated as one purchase, so a split shipment does not read as two.
Measure the current gap against it
Each day, the gap since an account's last order is compared to its own baseline. A customer that orders every 45 days and a customer that orders every 6 days are judged on different clocks — nothing is compared to an average.
Get a short list, not a report
The accounts that broke their pattern land on one list, capped so it stays workable, one row per account. Each row says what changed and what to do about it.
Work it and log the outcome
A rep calls, logs what happened, and the account leaves today's list. When it orders again, the save is recorded against the call that preceded it.
Check it against your own past
Replay runs the same rules over your history and shows which accounts it would have flagged, on what date, and how many days of warning you would have had. You can do this before you talk to anyone.
Data Requirements
What data do I need?
No black box
Why you can argue with it
Nothing decides who to call except arithmetic
Detection, level, priority, the sentence explaining the flag, and the recommended action are all computed from your order dates and amounts. No model participates in any of it. Run the same data twice and you get the same list.
Every flag shows its work
Each account opens to the baseline it was measured against, the gap that broke it, what changed in its order mix, what it is worth to you, and the date this would have fired. If you disagree with a flag, you can see exactly which number produced it.
Where AI is actually used
The assistant that answers questions in the app, and the column matcher that reads an unfamiliar CSV header, both use OpenAI under its API terms. Neither one decides anything about an account. Your data is never used to train a model.
Get Started
See which of your accounts already drifted.
Upload an order export and get the answer in about a minute. It runs in your browser — the file never leaves your machine, and there is no account to create.