Anomalies

Revenue swung 62% at one customer.
Was it a deal, or a data entry?

Close blockers are binary. Either the data is broken or it isn't. Anomalies are different. A customer's revenue jumped 38% month over month. A plan duration is 11 months instead of 12. An element sold at 56% above fair value. None of these are errors. All of them need someone to look and decide.

Anomaly checks
About anomalies
Findings that need human judgment, not data fixes. Review each one and mark as expected if it's a known event.
47anomalies need review across 3 types
Boreal Systems LLCCustomer↓ 62%
Revenue dropped from $412K to $157K
$412K → $157K
−$255K
Mark expectedOpen plan →
Plan #1015611-month
Boreal Systems · Annual Support · expected 12 months
11 months
$94,200
Mark expectedOpen plan →
Element RE-4401Fair value
Sold at $12,800 vs fair value $8,200 (±56%)
±56%
$4,600 variance
Mark expectedOpen plan →
The problem

Close blockers have answers. Anomalies have judgment calls.

An arrangement total that doesn't match its element allocations is wrong. You fix it. But a customer whose recognized revenue dropped 62% from last month? That might be a lost renewal. It might be a billing timing issue. It might be a data entry that posted to the wrong entity. You can't fix it without looking, and you can't ignore it during close.

Sentry runs three anomaly types across every subsidiary, flags the ones that cross configurable thresholds, and gives you the same triage workflow (mark expected, structured reason, fingerprint, resurface) as the rest of the product.

Revenue change by customer

Period-over-period swings, ranked by dollar impact.

Sentry compares each customer's recognized revenue this period against last period. Anyone crossing the threshold (default: ±20% and ≥$5K) shows up as a card with a 6-month sparkline, the dollar delta, and a direction tag (↓ Down or ↑ Up).

Rollup stripNet change, revenue down, revenue up. Three numbers across the top that frame the period’s overall movement.
Direction toggleFilter to only decreases or only increases. The toggle shows live counts for each direction.
SparklinesSix-month trend per customer. The red dot marks the current period. You see whether this is a sudden drop or a gradual decline.
Net change
−$221K
12 customers
Revenue down
−$312K
8 customers
Revenue up
+$91K
4 customers
All 12↓ Down 8↑ Up 4
Boreal Systems↓ 62%
$412K → $157K
−$255K
Northvane Solutions↓ 41%
$130K → $77K
−$53K
Summit Digital↑ 38%
$89K → $123K
+$34K
Detecting plans that match: off-by-one from quarterly, half-year, annual, and every 12-month cadence beyond
Plan #1015611-month
Boreal Systems · Annual Support
Jan 15 – Dec 1411 months · $94,200 deferred
Plan #1084213-month
Pacific Wire & Cable · Platform License
Mar 1 – Mar 3113 months · $48,000 deferred
Plan #1120411-month
Summit Digital · SaaS Subscription
Feb 1 – Dec 3111 months · $33,200 deferred
Plan #1039823-month
Cascade Industries · Managed Services
Apr 1 – Feb 2823 months · $187,440 deferred
Suspicious plan dates

An 11-month plan on a 12-month contract is almost always a typo.

Sentry scans every active revenue plan and flags durations that are one month off a standard cadence: quarterly (3), half-year (6), annual (12), and every 12-month multiple beyond. An 11-month or 13-month plan passes NetSuite validation, recognizes revenue, and posts to the GL. But each period posts a different amount than the contracted cadence would.

Pattern tagsEach card shows its detected pattern: 11-month, 13-month, 23-month. You see the exact off-cadence duration, not a generic flag.
Duration + deferredThe plan’s start and end dates, how many months it spans, and the deferred balance sitting on the schedule.
Tunable patternsAdjust which cadences to detect. Disable 13-month if your contracts legitimately span 13 months.
Thresholds

Tighter thresholds, fewer cards. Looser, more. You pick.

Every anomaly type has configurable thresholds. Revenue change has a percentage and dollar floor. Suspicious dates has a pattern list. Fair value deviation has a variance band. Adjust them from the popover or the full threshold configuration page.

The popover shows a live preview: “Would show 28 anomalies (currently 28).” You see the effect before you apply. If every anomaly type returns zero, the summary page shows a green hero: “Nothing unusual this period.”

Live previewAdjust a pattern or threshold, see the count change before you commit. No guessing.
Per-user overridesYour thresholds don’t affect your colleague’s view. Each user can tune independently.
Clean stateWhen all anomaly types pass, the summary shows a green banner with the scan count. Not a blank page.
Suspicious Plan Dates
Currently flagging 28 plans
Patterns to detect
11-month plans
Annual contract missing one period
14
13-month plans
Annual contract with extra period
14
Would show 28 anomalies · currently 28
Reset to defaults
CancelApply
Nothing unusual this period
All 3 anomaly types ran across 33 subsidiaries. None crossed thresholds.

See anomalies on your sandbox.

30-minute walkthrough. Your customers, your plan dates, your thresholds.

Book a Demo

No commitment. One business day response.