Upstox MTF Smart Nudge & Risk Engine Prototype

[PROPOSED] Deterministic Decision Engine

Upstox MTF Smart Nudge & Risk Mitigation Engine

A customer-facing intervention layer designed to help MTF traders address margin shortfalls efficiently before forced square-off, while deferring to authoritative RMS controls.

⚠️ THE PROBLEM

Off-app SMS alerts lack in-context guidance, causing multi-step navigation friction and avoidable forced square-offs.

💡 THE SOLUTION

In-app Smart Nudge surfacing explicit risk state, shortfall breakdown, and recommended next action (Add Cash | Pledge | Trim).

🎯 PRIMARY PRODUCT METRIC [PROPOSED]

Successful Risk Resolution Rate

"Share of eligible MTF risk events where authoritative shortfall is cleared within the resolution window without forced square-off."

❓ WHY SMART NUDGE? [DECISION HYPOTHESIS]

"If research confirms that customers recognize the margin problem but struggle to decide or execute the next action, a guided intervention layer is a stronger V1 bet than simply increasing notification frequency."

🔍 01. PROBLEM EVIDENCE & DISCOVERY LAYER [PROPOSED RESEARCH DESIGN]

📋 A. PUBLIC EVIDENCE [PUBLIC EVIDENCE TO VALIDATE]

Public Broker / MTF Docs: Short deadlines mean resolution friction increases forced square-off risk. [SOURCE TO BE VERIFIED]

Customer Discussion Patterns: Retail traders cite confusion regarding auto square-offs and pledge flows. [SOURCE TO BE VERIFIED]

Competitor Observable Patterns: Off-app SMS alerts require manual tab-switching, adding latency. [SOURCE TO BE VERIFIED]

💡 B. PRODUCT INSIGHTS [POTENTIAL INSIGHT]

1. Telling customers margin is required without a clear next action causes delay.

2. A single default action fails customers who hold stock collateral but lack cash.

🧪 C. HYPOTHESES TO VALIDATE [HYPOTHESIS]

H1: Shortfall customers struggle to identify appropriate action quickly.

H2: Contextual recommendations improve successful risk resolution.

H3: Guided flows reduce avoidable support dependency.

🔬 VALIDATION PLAN & RESEARCH DESIGN

1. Interviews: Measure action confidence

2. Support Tickets: Measure root-cause queries

3. Funnel Analytics: Measure step conversion

4. Usability Testing: Measure task execution time

⚖️ 02. REGULATORY & RISK CONTEXT

"Smart Nudge is an intervention layer around authoritative broker/RMS controls, not a replacement for them."

1. MTF Margin / Risk Rules: Consume authoritative margin/risk state from broker RMS. [TO BE VALIDATED WITH RISK/COMPLIANCE]

2. RMS Authority & Liquidation Boundary: Never promise guaranteed resolution windows. Always defer to authoritative RMS state.

3. Pledge Workflow: Surface pledge option only when securities are eligible and CDSL/NSDL T-PIN can be initiated. [TO BE VALIDATED]

4. Regulatory Change Watch: Keep rules configurable for SEBI framework reviews. [PROPOSED / CONSULTATION]

⛔ DOES NOT CONTROL: Exchange formulas, RMS liquidation, haircuts, pledge approvals.

✅ CONTROLS: Risk explanation, option prioritization, guided workflows, metric tracking.

🎯 03. PRODUCT STRATEGY & DECISION FRAMEWORK

Decision Action	Condition
SCALE	Primary metric improves AND risk/customer guardrails remain within agreed limits.
HOLD / INVESTIGATE	Primary metric improves BUT one or more guardrails deteriorate.
ITERATE	Primary metric is flat AND funnel drop-off identifies experience bottleneck.
STOP	Primary metric worsens OR material guardrail deterioration occurs.

[HYPOTHESIS] FRICTION ANALYSIS: THE STATUS QUO PROBLEM

[HYPOTHESIS] Customers may delay action when margin-risk notifications require multiple steps across separate menus or do not clearly explain recommended next steps. Navigating sub-menus for top-ups or stock pledges creates friction before a margin shortfall can be resolved.

V1 AI Product Principle & Future Roadmap

V1 uses deterministic rules because risk interventions should be explainable and auditable. Generative AI should not make regulatory margin decisions.

📐 04. MEASUREMENT ARCHITECTURE & EXPERIMENT DESIGN

CUSTOMER MEASUREMENT FUNNEL
1. ELIGIBLE MTF RISK EVENT → 2. NUDGE DELIVERED → 3. NUDGE VIEWED → 4. RECOMMENDATION OPENED → 5. INTERVENTION STARTED → 6. INTERVENTION COMPLETED → 7. RISK STATE REFRESHED → 8. RISK EVENT RESOLVED

⚡ LEADING METRICS

Nudge Delivery & View Rate
Recommendation Engagement Rate
Intervention Start & Completion Rate

🎯 OUTCOME METRICS

Successful Risk Resolution Rate
Time to Risk Resolution
Forced Square-off Rate

🛡️ RISK & CUSTOMER GUARDRAILS: Forced square-off rate must not increase; monitor complaint rate, failed intervention drop-offs, nudge frequency, and support contact volume.

🧪 FIRST A/B TEST DESIGN [PROPOSED]

Control: Existing MTF risk communication experience [TO BE VALIDATED]

Treatment: Smart Nudge + Guided Intervention

Hypothesis: Recommending suitable interventions at risk moment improves successful risk resolution without worsening guardrails.

🏗️ 05. DECISION ENGINE ARCHITECTURE & INPUTS

            Broker RMS Authoritative Risk State

              ↓ Risk Input Normalisation

            Proposed Margin Risk State Engine (Healthy / Watch / Action Required / Critical)

              ↓ Intervention Recommendation Engine (Deterministic Resource Matching)

            Customer Smart Nudge UI (Contextual Why & Portfolio Signals)

              ↓ Customer Action (Cash | Pledge | Trim)

            Broker RMS Recalculation (Authoritative Refresh)

[PROPOSED] DECISION ENGINE INPUTS & CONTEXTUAL SIGNALS

INPUTS: Margin Shortfall, Available Cash, Eligible Demat Collateral, Urgency Level, Position Quantity.

CONTEXTUAL SIGNALS: Unrealised MTM Loss, Position Concentration %, Security Volatility Index.

"V1 intentionally avoids a composite risk score. Intervention recommendations remain deterministic and explainable."

Upstox MTF Smart Nudge & Risk Engine PM Portfolio Artifact