The 11 checks. What Othor AI looks for, every night.

Every night, each check runs on every number and breakdown it applies to. A result becomes a finding only if it passes three gates: the evidence is strong once every test of the night is corrected together for false discoveries; the money at stake is above your floor (₹5 lakh unless you change it); and it isn’t explained by a known pattern in your Ontology.

The checks are statistics and signal analysis, written in code. No model decides what is a finding.

Breaks a total into its parts and tests whether offsetting moves are hiding a real change.

Why reports miss it: The total barely moved, so no one looked underneath: one region fell while another rose.

Finds the week a series settled at a new level, and confirms the shift is real and has held.

Why reports miss it: Week-on-week reports compare each week with the one before, and each looks like the last.

Holds each actual against a forecast band fixed before it arrived, and each goal against where the year is heading.

Why reports miss it: A target is usually checked at month end, after the gap has opened.

Watches the spread of a process, not just its average, and tests when it widens.

Why reports miss it: Reports show the average, and the average often doesn’t move until later.

Tracks how two related numbers move together, and tests when that link breaks.

Why reports miss it: Each number on its own looks normal; only the pair has changed.

Searches for numbers that move ahead of others, and tests that the lead isn’t chance.

Why reports miss it: Reports put numbers side by side for the same period, not one ahead of another. It needs to know which numbers should move together, so it may ask you first.

Measures how much rests on a few customers or suppliers, and whether that is growing.

Why reports miss it: No report ranks names by their share, so the dependence grows quietly.

Learns each customer’s or supplier’s own rhythm, and notices when one goes quiet.

Why reports miss it: A customer who stops ordering creates no row, so nothing in a report changes.

Looks for values bunching just under limits and at round numbers, a sign of how a process is rounded or worked around.

Why reports miss it: Every value is valid on its own; the pattern only shows in the distribution.

Reconciles the same number from two systems and flags the gaps that matter.

Why reports miss it: Each system is consistent on its own, and they are owned by different teams.

Tests whether an average is hiding two different groups, and describes them.

Why reports miss it: An average is one number; it can’t show that it describes no one.

Some checks need history, a second source, or an answer from you. Sweep’s Coverage says which checks ran fully, partly or not at all, and what would let each one run.