Automatic Schema Detection
Time columns, numeric metrics and dimension groups are inferred the moment you upload. No column mapping UI. No "Apply" button. Just results.
🤖 ML-powered · No rules to configure · Zero setup
Your data already knows
something is wrong.
ThresholdIQ finds it.
Upload any Excel, CSV or JSON file. ThresholdIQ's multi-method ML engine automatically detects anomalies, seasonal patterns, trends and correlated metric deviations — no thresholds to set, no formulas to write.
7-day unlimited trial · No credit card · Your data never leaves your browser
Drop in any structured file
.xlsx
.xls
.csv
.json
.xml
Google Sheets
Detection engine
9 methods. One upload.
Automatically graded.
Every metric in your file is passed through a full ML pipeline simultaneously. Results are fused into a composite anomaly score that drives three-tier severity classification.
Multi-Window Z-score
Rolling windows at 50, 100, 200 and 500 points compute local mean & std-dev. Breaches across multiple windows escalate severity automatically.
Warning → EmergencyEWMA Spike Detection
Exponentially weighted moving average filters the trend. Residuals that spike beyond 3σ of the smoothed series are flagged as sudden anomalies.
Sensitivity boosterSARIMA Seasonal Residuals
A seasonal ARIMA model accounts for daily and weekly cycles. Points that deviate from seasonal expectation — not just raw magnitude — are surfaced.
Pattern boosterIsolation Forest
Unsupervised multivariate outlier detection across all metrics simultaneously. Identifies globally unusual point-in-time combinations — e.g. meter reading zero while current is normal.
Global outlierCorrelation Deviation
Monitors whether correlated metrics deviate in the same direction simultaneously. Voltage + Current + Power Factor all abnormal together → emergency escalation.
Multi-metric alertDBSCAN Cluster Noise
Density-based clustering identifies points that don't belong to any normal behavioural cluster. Catches reverse-wiring, tampering, and sensor drift patterns.
Behavioural outlierSeasonal Baseline
Builds separate mean & std-dev for each hour-of-day and day-of-week bucket. Sunday-low consumption or overnight lulls won't fire false positives.
Context-awareTrend Detection
Compares consecutive 50-point window means. Monotonic drift across three windows — a gradual rise or fall — is flagged before it becomes a critical event.
Early warningStuck & Zero Detection
Repeated identical values across a rolling window, or a sudden drop to zero from a non-zero series, indicate sensor failure or data pipeline issues instantly.
Hardware failureScore fusion formula
final_score = multiWindow_score + min(0.25, ml_composite × 0.25)
ml_composite = EWMA(0.12) + SARIMA(0.22) + IForest(0.20) + Corr(0.12) + DBSCAN(0.06) + Seasonal(0.12) + Trend(0.10) + Stuck(0.06)
Multi-window is the primary severity driver. ML methods can only boost — never de-escalate — the severity decision. This prevents single-method false positives while ensuring no coordinated anomaly is missed.
Severity system
Three tiers. Automatic grading.
No manual threshold values required. Severity is derived from how many rolling windows simultaneously detect a deviation.
Warning
W50 window breach only
A value deviates beyond 2σ in the 50-point short-term window. Fast-reacting — catches sudden spikes before they escalate. May self-resolve.
Critical
W50 + W100 both breached
The deviation persists into the 100-point mid-term window. This confirms the anomaly is not noise. Immediate investigation recommended.
Emergency
W50 + W100 + W200 all breached
The anomaly is confirmed across short, mid and long-term baselines. A structural shift has occurred. Immediate escalation required — email report fires automatically.
How it works
Two steps. That's it.
No mapping. No rule configuration. No BI consultant. The ML engine handles everything automatically.
Upload your file
Drag in Excel, CSV, JSON, XML or connect Google Sheets. ThresholdIQ auto-detects the time column, numeric metrics and dimension groups — no mapping required.
Click “Detect Anomalies”
Nine ML methods run in parallel across every metric. Results appear on the timeline with severity colour coding, signal breakdown, and exportable alert logs — in seconds.
Features
Intelligence you'd expect
from a data scientist.
Built for Finance and Operations teams who need answers fast — not a six-week BI implementation project.
Detection Signal Breakdown
The "Signals" tab shows exactly which detection methods fired on each anomaly — EWMA, SARIMA, IForest, DBSCAN, Correlation, Seasonal — with counts and severity.
Timeline & Severity Charts
Anomalies are plotted on an interactive timeline with Warning / Critical / Emergency colour bands. Pan, zoom and filter by dimension group without rerunning detection.
CSV, PDF & Email Reports
Export the full anomaly log as CSV. Generate branded PDF reports with KPI summary cards. Email alert reports directly to stakeholders with one click.
100% Browser-Side. Private.
All parsing and ML detection runs locally in your browser using Web Workers. Your spreadsheet data never leaves your machine — not even a single byte.
Scales to Millions of Rows
Off-thread Web Worker processing keeps the UI fully responsive even with 1M+ row datasets. Detection runs on a background thread — the app never freezes.
See it in action
Watch ThresholdIQ detect anomalies automatically
From raw file upload to full ML anomaly report — under 60 seconds, zero configuration.
Who uses it
Works wherever your data lives in a spreadsheet
If your data is structured and time-stamped, ThresholdIQ's ML engine will find what's unusual — without you having to guess thresholds upfront.
Financial Monitoring
Daily cash flow, AR aging, budget variance. The seasonal baseline prevents month-end spikes triggering false emergency alerts. Trend detection surfaces budget drift weeks before quarter-end.
Inventory & Supply Chain
Multi-SKU stock monitoring. Stuck-value detection catches zero-quantity entries that signal data pipeline failures before they cascade. Correlation flags when demand and supply diverge simultaneously.
Sales Performance
Rep performance metrics exported from CRM. EWMA detects sudden drops mid-quarter. Trend detection flags gradual declines before they appear in end-of-quarter roll-ups.
Operations & SLA Tracking
Customer response times, delivery delays, quality metrics. Isolation Forest catches multivariate anomalies — when latency, error rate and queue depth all deviate simultaneously.
IoT & Sensor Data
Voltage, current, temperature, pressure — all analysed simultaneously. DBSCAN identifies tampering or reverse-wiring patterns. SARIMA prevents night-shift readings triggering day-shift alarms.
HR & Workforce Analytics
Overtime hours, absence rates, productivity metrics. Correlation detection flags when multiple workforce metrics degrade together — a leading signal of retention risk or burnout.
FAQ
Common questions
Do I need to set thresholds manually?
No — that's the whole point. ThresholdIQ uses rolling window statistics, seasonal baselines and ML models to learn what "normal" looks like for your specific data. It detects deviations automatically without any threshold values from you.
How is this different from a simple standard-deviation alert?
A single Z-score check misses seasonal patterns (Sunday lows, night-shift readings) and can't detect multi-metric correlated failures. ThresholdIQ runs 9 methods simultaneously: SARIMA handles seasonality, Isolation Forest catches global multivariate outliers, DBSCAN identifies behavioural cluster deviations, and the results are fused into a single scored alert.
Do I need to install anything?
No. ThresholdIQ runs entirely in your web browser. Upload your file and anomalies are detected immediately — no downloads, no plugins, no IT setup required.
What file formats does it support?
Excel (.xlsx, .xls), CSV, JSON, and XML. If your data has a time column and numeric metrics, ThresholdIQ auto-detects the schema and starts detection immediately. Google Sheets is also supported on paid plans.
Is my data safe? Does it leave my browser?
All file parsing and ML detection runs locally in your browser using Web Workers. Your spreadsheet data never leaves your machine and is never uploaded to any server. Not even a single row.
What's included in the free trial?
A full 7-day unlimited trial with all Pro features: all 9 detection methods, unlimited file sizes, multi-metric detection, anomaly log, timeline charts, CSV & PDF export, and email reports. No credit card required.
How is this different from Power BI?
Power BI is a general-purpose visualisation tool. It has no built-in anomaly detection engine — you'd need to build custom DAX measures, configure AI visuals, and set up data refresh pipelines. ThresholdIQ is purpose-built: upload a file and get ML anomaly detection in under 60 seconds, no BI skills needed.
Stop guessing where anomalies hide.
Upload your file. Nine ML methods run automatically. Anomalies surface in seconds — graded Warning, Critical or Emergency — with no thresholds to configure and no data leaving your browser.
Just enter your email — no password required
No setup · Your data never leaves your browser