For Ops Managers · Supply Chain · Logistics Teams

Catch inventory failures
and SLA breaches before they escalate.

Upload your operations report, inventory export, or KPI spreadsheet. ThresholdIQ's 9 ML methods automatically detect supply chain anomalies, delivery SLA breaches, production deviations and correlated KPI failures — no coding, no BI tools, no thresholds to configure.

7-day unlimited trial · No credit card · Data never leaves your browser

79%
of supply chain leaders report data quality as a top challenge
9
ML detection methods running simultaneously
<60s
from file upload to full anomaly report
0
threshold rules or config required

Operations teams drown in data, starve for signals

You have the KPI spreadsheet. You have the inventory export. But by the time you manually scan 500 rows, the stockout has already happened and the SLA has already been breached.

📦

Inventory count mismatches

Warehouse counts drift from system records over weeks. Small discrepancies compound into unforecast write-offs and stockouts that only surface at month-end cycle count.

🚚

Delivery SLA breaches

A single carrier's on-time rate drops from 94% to 78% — but it's buried in an average across 12 carriers. The blended number looks fine while customer escalations pile up.

📉

Production yield anomalies

A production line's reject rate creeps from 2.1% to 3.8% over four weeks. Within static tolerance? Maybe. But the trend is abnormal and the quality cost is accumulating.

🔗

Correlated supplier failures

Lead time, defect rate and unit cost all move together for one supplier — but no single metric crosses a threshold. The correlation pattern is the signal. Static rules never see it.

From export to anomaly report in 60 seconds

No dashboarding tool to learn. No formulas to write. Export from your WMS, ERP, or the spreadsheet your team already maintains.

1

Export your operations data

Inventory counts, delivery logs, production KPIs, supplier scorecards, SLA reports — any structured data from any system. ThresholdIQ supports .xlsx, .csv, .json and .xml. No column mapping required.

2

Upload — processing stays local

Drop the file into ThresholdIQ. All 9 ML detection methods run entirely in your browser using Web Workers. No server upload, no compliance paperwork, no IT involvement needed.

3

9 ML methods detect anomalies automatically

EWMA, SARIMA, Isolation Forest, DBSCAN, correlation deviation, seasonal baselines, trend detection, Z-score and stuck/zero detection all evaluate your data simultaneously. Every deviation is graded Warning, Critical, or Emergency.

4

Act before it escalates

Export the anomaly report as CSV for your WMS or PDF for your operations review. See exactly which metrics deviated, whether the trend is seasonal or genuinely abnormal, and which metrics moved together as a correlated failure.

Works across the entire operations function

If your data has a timestamp and numeric metrics, ThresholdIQ finds what's abnormal — whether it's a stock count, a delivery SLA, a supplier scorecard, or a service level KPI.

📦

Inventory & Stock Monitoring

Multi-SKU, multi-location inventory counts. Stuck-value detection catches when stock levels stop moving — a sign of counting system failures or data pipeline breaks. Trend detection surfaces gradual stock erosion across multiple periods before a stockout occurs.

🚚

Delivery & SLA Performance

On-time delivery rates, transit times, and carrier performance by lane. SARIMA seasonal baselines exclude peak-season slowdowns from triggering false alerts. Correlation deviation flags when on-time rate, damage rate and customer complaints all deteriorate together for a specific carrier.

🏭

Production & Yield Tracking

Units per shift, reject rates, cycle times, and throughput by line. Trend detection identifies gradual yield degradation before it breaches any tolerance. Isolation Forest catches multivariate anomalies — when output, waste, and energy consumption all shift simultaneously.

👥

Supplier Scorecard Analysis

Lead times, defect rates, fill rates and pricing across supplier base. DBSCAN cluster noise identifies suppliers that behave unlike any normal vendor pattern. Correlation deviation catches when a supplier's quality, delivery and cost metrics all deteriorate together — an early supplier failure signal.

📋

Customer Service & Ticket KPIs

Response times, resolution rates, CSAT scores, first-contact resolution. EWMA detects sudden spikes in ticket volume or handling time. Seasonal baselines account for predictable volume patterns — Monday morning surges won't trigger alerts that genuine Thursday evening spikes will.

Warehouse & Fulfilment Metrics

Pick rates, pack rates, dock-to-stock times, order accuracy by zone. Multi-window Z-score escalates picking rate anomalies from Warning to Emergency as the deviation persists. Stuck detection identifies zones where activity has frozen — catching conveyor outages and system disconnections before orders are missed.

What ThresholdIQ finds in a supplier scorecard

A multi-metric correlated failure across one supplier — invisible to individual column reviews, immediately visible to ThresholdIQ.

SupplierOn-Time %Defect Rate %Lead Time (days)Unit Cost ChangeThresholdIQ
Supplier A96.20.812+0.2%Normal
Supplier B88.12.416+4.1%⚠️ Warning — correlated 4-metric drift
Supplier C94.81.111-0.1%Normal
Supplier D79.35.722+9.8%🟠 Critical — multi-window escalation
Supplier E97.10.610+0.5%Normal
Supplier F61.012.30+22%🔴 Emergency — sensor + isolation forest

Supplier B flagged for correlated deterioration across all four metrics — no single metric crossed a threshold but the combined pattern indicates a supplier under stress. Supplier D escalated to Critical. Supplier F flagged Emergency with zero lead time (data feed failure) and extreme cost and defect anomalies.

How each ML method applies to operations data

All 9 methods run in parallel across every metric in your file. Here's what each one specifically catches in operations and supply chain data.

Multi-Window Z-Score

Evaluates each KPI against rolling baselines at 50, 100, 200 and 500 periods. An SLA rate that drops and stays down across multiple windows escalates from Warning to Emergency — separating a one-day blip from a structural supplier breakdown.

Primary severity driver
📈
EWMA Spike Detection

Catches sudden ticket volume spikes, unexpected inventory drops, and instantaneous delivery rate anomalies. Fast-reacting — detects the sudden event before it compounds across multiple reporting windows.

Sudden event detection
🌀
SARIMA Seasonal Residuals

Models peak-season demand patterns, month-end order surges, and weekly delivery cycles. A predictable peak-season SLA dip won't trigger alerts. A dip that exceeds the seasonal expectation will.

Peak-season aware
🌲
Isolation Forest

Detects globally unusual combinations across all metrics simultaneously — e.g. on-time rate fine but defect rate, lead time and cost all abnormal together. This multivariate pattern flags a supplier under stress before any individual metric breaches its limit.

Multivariate outlier
🔗
Correlation Deviation

Monitors the statistical relationships between KPIs. When normally-correlated metrics diverge — pick rate stays flat but order accuracy drops and handling time spikes — a process breakdown is flagged. The pattern is the signal.

Correlated failure
📍
DBSCAN Cluster Noise

Identifies suppliers, carriers, or SKUs that behave unlike any normal operational cluster. Catches systematic outliers — a warehouse zone that consistently underperforms, a carrier that behaves unlike all peers in the same lane.

Systematic outlier
🌙
Seasonal Baseline

Maintains separate normal ranges per day-of-week and time-of-day. Monday morning order surges and weekend warehouse slowdowns are learned — alerts only fire when a KPI deviates from its own expected level at that specific time.

Context-aware
📉
Trend Detection

Identifies monotonic drift in fill rates, gradual SLA erosion across periods, and slowly increasing lead times. Flags the trend at Warning level weeks before it becomes a customer escalation — giving procurement teams time to act.

Early erosion warning
🚫
Stuck & Zero Detection

Repeated identical stock counts or sudden zero inventory from an active SKU indicates a counting system failure or data pipeline break. Fires Emergency immediately — catching WMS connectivity issues before they become undetected stockouts.

System failure detection

Your ops data already contains
the warnings.

Upload your spreadsheet. Nine ML methods detect anomalies in seconds — graded Warning, Critical or Emergency — with zero configuration and zero data leaving your browser.

Just enter your email — no password required

No setup · Works with any WMS, ERP or TMS export