How do you find margin leakage hiding inside ERP data?
By separating measurement from interpretation. MISH ingests ERP and 3PL data, normalizes it, computes sales, cost, margin and baselines deterministically, flags exceptions, and only then lets AI explain the cause and recommend an action. Every recommendation is captured, and the outcome after the action is measured, so the system learns whether the advice worked.
Alex Scharifker · Product architect and systems lead
Designed and led by Alex Scharifker for Shippers Solutions as an applied decision-intelligence system.
MISH Sales & Margin Intelligence
Shippers Solutions needed clearer visibility into margin leakage, pricing drift, customer behavior, order cadence, and inventory return. The business had data, but the real problem was converting that data into trusted decisions sales and operations teams could act on.
How the system is put together
Sources in, deterministic logic in the middle, AI restricted to interpretation.
Data sources
- Acumatica ERP
- Elucidate 3PL
- Customer history
- Item / order data
- Cost & margin data
- Inventory movement
Layers
- Ingestion
- Normalization
- Deterministic analytics
- Exception detection
- AI interpretation
- Human feedback
- Decision ROI measurement
Deterministic logic
- Sales
- Cost
- Margin
- Weighted 3-month / 12-month baselines
- Negative margin flags
- Low margin flags
- Below-baseline flags
- Cadence shifts
- Volume anomalies
- Turns and Earns
Where AI is used
- Root cause analysis
- Recommended action
- Business context explanation
- User-facing summary
- Workflow guidance
Decision loop
- Detect exception
- Explain cause
- Recommend action
- Capture user response
- Measure post-action outcome
My role
- Product architecture
- Data model design
- Business-rule design
- AI use-case definition
- Stakeholder alignment
- Workflow adoption
- Training / SOP creation
- Roadmap toward Turns and Earns module
The architecture case file
Layer-by-layer architecture, deterministic model, AI guardrails, and the decision loop.
A dashboard tells you margin fell. MISH tells you which line broke, why, who owns it, and what to do.
Key takeaways
- —Client-specific implementation for Shippers Solutions across ERP + 3PL data, 19 users.
- —Deterministic SQL owns every number; AI only interprets it.
- —Exceptions routed to the account owner with a recommended next step.
- —Every decision captured and replayed against a locked baseline to measure ROI in dollars.
Architecture Principle
Calculated truth stays deterministic. AI only interprets.
- SQL owns sales, cost, margin, baselines, cadence, turns, and ROI.
- AI receives precomputed payloads and explains what the numbers mean.
- Human decisions and feedback feed the Decision ROI loop.
Ingest
ERP + 3PL data from Acumatica and Elucidate.
Normalize
Resolve customers, items, vendors, dedupe rows, preserve exclusions.
Detect
Run deterministic rules for Margin, Behavior, and Capital exceptions.
Interpret
AIAI explains root cause and recommends action using strict guardrails.
Measure
PayoffDecision ROI compares post-action outcomes against locked baselines.
What changed
- Three decision lenses in production: Margin, Behavior, and Capital.
- ERP + 3PL integration into a single normalized decision layer.
- Exceptions surfaced: negative margin, low margin, below-baseline pricing, cadence shifts, volume anomalies, and Turns and Earns opportunities.
- Deterministic math kept separate from AI interpretation so margins, baselines, turns, and ROI stay auditable.
- Decision ROI loop measures whether human action improved margin, revenue, volume, throughput, or inventory productivity.
Follow Me
LinkedInResume
Download PDF
From Chaos to Clarity
Get insights on product management and operations delivered to your inbox.
Subscribe on LinkedIn