Materiality & Anomaly
The two questions behind every close.
Is this normal, and does it matter enough for someone to look? Every experienced finance person asks both, dozens of times a week. The problem was never that the questions go unasked. It is where the answers live: in someone's head, or on the item itself.
Two questions, asked thousands of times a month
On day three of the month an invoice arrives. Coded to freight, $60K, sitting in a batch with four hundred others. Nobody looks at it twice.
On day six of the close, five weeks later, a reconciliation is off by $60K and three people spend most of a morning working out why. It turns out to be a freight rebate. It was fine all along, and the team had reached exactly that conclusion, on a similar item, the previous quarter.
Nothing failed. The invoice processed correctly, the reconciliation worked as designed, and everyone involved was good at their job. What went wrong is that two questions got asked five weeks late.
The two questions
Is this normal? That is anomaly. Does it matter enough for someone to look? That is materiality.
They are not a reconciliation feature
It is tempting to file both under reconciliation, because that is where they become visible. They actually run through every operation, all month.
The same two questions, at every step
| Operation | Anomaly asks | Materiality asks |
|---|---|---|
| Payables | Is this invoice unusual for this vendor: amount, timing, a possible duplicate? | Does it clear on its own, or route to a manager, a controller, the CFO? |
| Cash application | Is this short payment unusual for this customer, or simply how they always pay? | Is the unapplied residual worth a person chasing? |
| Collections | Is this customer paying unlike their own history? | Which overdue accounts genuinely move the reported position? |
| Reconciliation | Is this account behaving unlike its own history? | Which variances need certifying, and which can certify themselves? |
| Journals | Is an entry missing that appeared every prior period? | Is the adjustment large enough to need review? |
| Close | Which task is running long against its own normal? | What actually needs sign-off before the books can be called closed? |
How both get answered today
Anomaly is a person noticing
Someone experienced looks at a number and thinks *that is not right*. That judgement is real, and usually correct. It is also unwritten, applied unevenly depending on who is working the account, and it leaves when they do. No system knows what normal looks like for that account, so no system can raise it.
Materiality is a threshold in a policy document
It exists, it is agreed, and it is applied by hand at reconciliation time by whoever happens to be on that account. Because the number lives in a document rather than in the work, it gets re-interpreted, and occasionally re-argued, each period.
Both get asked during the close
The oddly coded invoice on day three is not questioned until the reconciliation on day six of close. By then the person who coded it has moved on to next month, and the reason has gone quiet.
And the answer is not kept
Last quarter's conclusion, freight rebate, immaterial, no action, sat in a message thread. This quarter the same pattern appears fresh, and the same morning gets spent again.
None of this is a competence problem. It is a placement problem. Both questions are being answered at the end, by hand, from memory.
What changes when they become parameters
The alternative is to treat materiality and anomaly the way you already treat an amount or a posting date: as properties measured and stored on the item itself, carried with it as it moves.
Materiality set once, routing continuously
Tiers are configured, and every item then routes itself on arrival rather than waiting for someone to apply the policy by hand.
Example materiality tiers and a period's volumes
| Tier | Above | Routed to | Items |
|---|---|---|---|
| Low | - | Clears automatically | ~1,180 |
| Medium | $10K | Accounting manager | ~46 |
| High | $50K | Controller | ~12 |
| Executive | $250K | CFO sign-off | ~2 |
Anomaly measured per account, not per rule
A blanket threshold flags the same things for a vendor who invoices monthly and one who invoices twice a year. Measured against each account's own history, *unusual* means unusual for this account, which is what the experienced controller was doing in their head all along.
Both applied on arrival, not at close
The odd invoice is flagged on day three. Nothing waits five weeks to become visible.
The conclusion stored
When a controller rules a variance immaterial and explains why, that decision stays attached to the item and its pattern. The next occurrence arrives already knowing how the last one was treated. Nobody researches it twice.
Why this decides whether the month-end close lands on time
A close runs late when it becomes an investigation. Investigations happen when items arrive carrying no history, and you have the number, but not whether it is unusual, not whether it matters, not what anyone decided last time.
Answer both questions during the month and the close inherits a queue that is already sorted. The exceptions are known, ranked and explained. Certification follows the materiality already recorded rather than a fresh judgement on every account.
The difference is not people working faster. It is the same two questions answered continuously, instead of all at once, at the end, under time pressure.
This is what connected finance looks like in practice, and it is the part that makes the close slow when it is missing.
A test worth running
Pick a variance your team resolved last quarter, and ask two things. How long after it happened did anyone notice? And if it recurred tomorrow, would the system know what you concluded?
Most finance functions answer *weeks* and *no*. Both answers are fixable, and neither requires anyone to work harder.