Behaving as Expected
98.13%
325,892 / 332,106 records
Records classified as Normal or Slightly Unusual are considered to be behaving as expected.
Neural Engine for Anomaly Detection & Learning
Let the data speak for itself.
Neadl learns expected behaviour without human labelling, allowing anomalies to surface on their own
Test 1
Within Expectation Groups
Contextual anomalies - Neadl learns typical data patterns within each Expectation Group and flags records that behave unusually.
Test Results
332,106 recordsHighly Anomalous
110records
0.03%
These records deviate significantly from expected behaviour.
Anomalous
6,075records
1.83%
These records have notable, but not significant, deviation from expected behaviour.
Neadl in the haystack
Find the record that deserves attention.
Neadl reviews millions of rows in minutes, ranks the most unusual records, and explains why each one stands out. Internal risk and control teams can move from broad sampling to individual instances of error, fraud, or control failure.
Coverage
Every row considered
No sample design, no narrow rule set, and no need to know the risk pattern before the review starts.
Priority
One ranked queue
Records are ordered by unusualness so reviewers can start with the exceptions most worth investigating.
Self-serve review
Analytics the business can run without a data science project.
Neadl is built for teams that need answers without waiting for specialist modelling, manual data preparation, or long project cycles.
Self-serve 01
No specialist support
Business users can upload a dataset and begin reviewing without relying on data scientists or external analytics teams.
Self-serve 02
No manual preparation
Classification, encoding, and readiness decisions are handled automatically before scoring begins.
Self-serve 03
No technical translation
Plain-language outputs show what is unusual, which fields contributed, and why the record is worth a closer look.
Any dataset
If you have the data, Neadl can find the anomalies.
Neadl is not limited to finance data. It can surface unusual records across operational, people, environmental, cyber, procurement, claims, payroll, and transaction datasets.
GHG emissions reportsEmployee wellbeing surveysOffice attendance logsTill receiptsPhishing training resultsVendor master filesExpense claimsProcurement records
Dataset intake
Profiling complete01
Upload CSV
02
Profile fields
03
Start ranking
ExpensesPayrollProcurementJournal entriesClaimsTransactionsVendor paymentsCredit cardsTravel bookingsPurchase ordersInvoicesTimesheetsInventoryRefundsGrantsDonationsGHG emissions reportsEmployee wellbeing surveysOffice attendance logsTill receiptsPhishing training resultsVendor master filesExpense claimsProcurement records
Dataset preview
Explore with Explanation.
No 'Black Box' Analytics
Inspect every anomaly in context with distribution overlays, bucket-level evidence, and plain-language reasoning tied back to the selected record.
Distribution PM2.5
Population New York (City)
No. of Global Records 52,704
Highlighted value: 69.200
---- Global reference
Highlighted value69.200
PM2.5 range67.5 - 70.0
No. records4
non-blank New York records<0.1%
No. global records142
non-blank global records0.3%
Neadl Insights
The selected anomaly is 1 of 4 records for City "New York", where the PM2.5 value falls between 67.5 and 70.0.
This equates to 0.05% of the Expectation Group and 0.3% of global records.
Globally, there are 142 records, where PM2.5 is "67.5 - 70.0", which equates to 0.3%.
Bucket Detail
67.5 - 70.0Exact values67.8
67.5
68.2
67.6
69.2
1 record
12 records
67.7
69.4
67.9
68.0
68.4
68.5
68.6
68.7
68.9
69.3
69.6
69.7
69.8
Workflow
A review rhythm analysts already understand.
Behind the scenes
6 automated steps transform raw data into an actionable anomaly report.
LOAD
Data Ingestion
A focused signal layer designed to make row-level review easier to trust.
CLASSIFY
Data Profiling
A focused signal layer designed to make row-level review easier to trust.
ENCODE
Model Readiness
A focused signal layer designed to make row-level review easier to trust.
SELECTION
Expectation Setting
A focused signal layer designed to make row-level review easier to trust.
MODEL
Anomaly Scoring
A focused signal layer designed to make row-level review easier to trust.
EXPLORE
Exploration
A focused signal layer designed to make row-level review easier to trust.
Explainable intelligence
Academic methods, built for business review.
Neadl applies advanced anomaly detection methods that were previously practical only through bespoke data science work, then adds proprietary influence calculations to make the results understandable.
Proven modelling foundations
Use advanced techniques without commissioning a custom model for every new dataset.
Automated data readiness
Neadl prepares fields for modelling so reviewers do not need to make technical classification and encoding decisions.
Influence, not just scores
Reviewer-friendly explanations identify the parts of a record that pushed it away from expected behaviour.
Isolation Forest
Short paths expose unusual records.
Model family
Isolation Forest
Ensemble
4 sample trees
Signal
short path depth
Tree 01
2
path depth
Tree 02
7
path depth
Tree 03
3
path depth
Tree 04
9
path depth
Path depth profileshorter = more unusual
Records isolated in fewer splits receive stronger anomaly signals across the ensemble.
End-to-end workflow
From upload to ranked anomalies before the coffee gets cold.
Neadl brings ingestion, profiling, readiness, expectation setting, anomaly scoring, and exploration into one workflow. The result is a ranked list of unusual records without labelled examples, hand-built rules, or bespoke modelling for every dataset.
Evidence export
Manager packRanked anomaly listIncluded
Explanation summaryIncluded
Reviewer notesIncluded
CSV and PDF exportIncluded
Step 01
Upload any structured dataset
Step 02
Automate readiness decisions
Step 03
Rank unusual records
Step 04
Explain why each record matters
Comparison
Built for prioritization, not just detection.
DimensionManual samplingRules-based checksNeadl
CoverageSmall samplesKnown scenariosEvery row ranked
SetupManual judgmentRules to maintainUpload and profile
SignalEasy to missHigh false positivesPrioritized by unusualness
EvidenceScattered notesRule outputsExplanation and export trail
Access plans
Start with the dataset that matters now.
FAQ
Early access
Start with one dataset.
Bring the review file your team already has. Neadl will help rank, explain, and package the rows worth investigating.
Join waitlist