How It Works

From raw data to
specific action.

Six layers turn your commerce data into decisions. Most tools stop at layer one.

The Pipeline

Six layers. One decision.

Each layer adds intelligence. By the time data reaches your team, it's not a number — it's an instruction.

01

Data Collection

Shopify orders and customers. NetSuite inventory and B2B revenue. GA4 traffic. Google Ads and Meta spend. Klaviyo email performance. All synced every 15 minutes. Every data point carries a freshness timestamp.

Shopify
NetSuite
GA4
Google Ads
Meta
Klaviyo
Yotpo
Gorgias
02

Signal Detection

The engine scans three domains continuously. Revenue anomalies: day-over-day drops or spikes detected using 30-day rolling percentile bands. Customer churn: purchase intervals exceeding 1.5x their normal pattern. Inventory risk: critical stock levels with meaningful revenue at risk.

📈 Revenue drop detected: -42% vs yesterday
👥 182 customers overdue (avg 34 days)
📦 10 items critical ($4,376/day at risk)
03

Trust Scoring

Every signal gets a trust score from 0% to 100%. Factors: source reliability (metric engine 90%, customer data 85%, inventory 80%), data completeness, and time-of-day. At 6 AM, when only 20% of revenue data has arrived, signals are automatically suppressed — not hidden, but explicitly marked as unreliable.

Metric Engine 90% base trust
× Day progress 20%
= Effective trust 27%
→ Signal suppressed
04

Insight Grouping

Related signals merge into insights. A revenue drop + customer churn = one unified insight, not two separate alerts. Each insight is scored on three dimensions: revenue impact (0-40 points), urgency (0-35 points), and confidence (0-25 points). The highest-scoring insight surfaces first.

Revenue ↓ —┐
├→ Revenue Crisis (priority: 85)
Churn ↑    —┘
05

Story Generation

Insights connect into a narrative. “Revenue spiked 169% but 182 customers are churning. The spike may mask a retention problem.” Seven business patterns recognized: unstable growth, supply crisis, demand erosion, retention issue, inventory alert, revenue anomaly, and full crisis. Every number comes from real data.

Unstable Growth Revenue spiked but customers are leaving. The spike may mask an underlying problem. Focus: Customer Retention
06

Action Generation

Each insight produces concrete tasks assigned to the right team. “Call this customer — $3,866 LTV, 9 days inactive.” “Reorder this SKU — $1,298/day at risk.” Tasks have targets, instructions, deadlines, and suggested owners. Sales reps see tasks, not dashboards.

📞 Call jorgejr@jwsteelinc.com THIS WEEK
📞 Call beth.taylor@dawson3d.com THIS WEEK
📦 Reorder Fireplace Grate THIS WEEK
Why Trust Matters

What happens at 6 AM when data is incomplete?

This is where most analytics tools fail — and where Platox is fundamentally different.

Every other dashboard at 6 AM
Revenue: $9.59
Change: -100% vs yesterday

❌ Technically true
❌ Practically useless
❌ Creates false alarm
❌ No explanation provided
Platox at 6 AM
⚠ Data Status: Incomplete (21% of day)
🔇 3 revenue signals suppressed as unreliable

✓ 182 customers at churn risk ($53K LTV)
✓ 10 inventory items critically low

○ Revenue trend direction
Confidence: 40%  |  Risk Level: Early  |  Focus: Customer Retention
“Experienced operators don't ignore incomplete data. They don't overreact to it either. Platox behaves the same way.”
Market Intelligence

Is this your problem, or the market's?

Platox compares your performance against category context — broad for directional signals, deep for specific patterns.

Broad Category Context

When a signal is weak or directional, Platox looks at the broader category. Is the entire segment declining, or just your brand?

Deep Subcategory Context

When a signal is specific, Platox zooms into subcategory data. A problem in “outdoor lighting” is different from a general “home & garden” slowdown.

Market context is a decision layer, not a separate dashboard. It helps the engine explain why.

Live Example

What an operator sees.

This is a real output from the Decision Engine.

platox decision engine
HEADLINE Data incomplete (21% of day) — early signals show retention issue ANALYSIS Pattern: Unstable Growth Severity: High Confidence: 40% KEY FINDINGS182 customers overdue for reorder (avg 34 days inactive) • Top at-risk account: jorgejr@jwsteelinc.com ($3,866 LTV) • Revenue data covers 21% of today — signals suppressed SIGNAL FLOW Revenue 169% ↑ →masks→ Churn 182 ↓ →drives→ Risk $53K ↓ WHY THIS MATTERS Too early to confirm revenue trend. Customer churn without a revenue drop means there is still time. Act now to prevent future revenue loss. RECOMMENDED FOCUS: Customer Retention ACTIONS GENERATED: 5 tasks
FAQ

Common questions

Is Platox a dashboard?
Platox includes dashboards, but the core is a Decision Engine. It detects problems, explains what's happening, and generates actions. Dashboards are the view layer. The engine is the value.
How does it handle incomplete data?
Revenue signals are suppressed when less than 30% of the day's data has arrived. The system generates a “data incomplete” marker and tells operators what it can confirm and what it can't. No false alarms from partial comparisons.
What types of actions does it generate?
Specific tasks: “Call this customer — $3,866 LTV, 9 days inactive.” “Reorder this SKU — $1,298/day at risk.” Each with a target, instruction, deadline, and suggested owner.
How does trust scoring work?
Every signal gets a 0-100% trust score. Based on source reliability, data completeness, and signal diversity. The system tells you when confidence is low.
What integrations are supported?
Shopify, NetSuite, GA4, Google Ads, Meta Ads, Klaviyo, Yotpo, Gorgias, Google Search Console, MNTN, Google Trends.
How fresh is the data?
Synced every 15 minutes. Every data point carries a freshness timestamp. Trust scoring factors in data age automatically.

Ready to see decisions, not dashboards?

Connect your data. Get your first insight today.

Start free trial