Champo Carpets: Improving Business-to-Business Sales Using Machine Learning Algorithms Custom Case Solution & Analysis

Evidence Brief

Financial Metrics

Annual conversion rate: Approximately 12.8 percent of buyers who receive samples eventually place an order.
Sampling costs: Production and freight costs range from 50 dollars to 200 dollars per sample unit.
Customer acquisition cost: High due to participation in international trade fairs like Domotex in Germany and High Point Market in the United States.
Inventory: Champo maintains a catalog of over 3000 designs with significant capital tied up in physical sample stock.
Revenue impact: Orders typically range from 5000 dollars to 50000 dollars per successful conversion.

Operational Facts

Sales cycle: Duration spans 6 to 12 months from initial trade fair contact to final purchase order.
Lead generation: 2500 to 3000 unique business visitors attend the Champo stall annually across major fairs.
Sample distribution: Historically based on the intuition of sales representatives and direct requests from potential buyers.
Data availability: Historical records from 2013 to 2017 include buyer country, fair location, previous purchase history, and sample types requested.
Headcount: 500 plus employees involved in production and administration.

Stakeholder Positions

Sanjay Mehrotra, Managing Director: Seeks to reduce operational waste and improve the precision of sales efforts.
Sales Team: Rely on personal experience and relationship-building; may resist algorithmic overrides of their professional judgment.
Data Science Consultants: Focused on model accuracy, precision, and recall metrics to justify the shift to automated lead scoring.
Global Buyers: Expect high-quality physical samples as a prerequisite for committing to large-scale floor covering contracts.

Information Gaps

Competitor conversion rates: The case does not provide benchmark data for other Indian carpet exporters.
Customer Lifetime Value (CLV): Lack of long-term data on buyer retention beyond the initial five-year window.
Real-time shipping fluctuations: Volatility in international freight costs is not quantified for the 2018-2019 period.

Strategic Analysis

Core Strategic Question

How can Champo Carpets optimize its high-cost sales funnel by replacing intuition-based sample distribution with predictive lead scoring?
Can a machine learning model accurately identify high-probability buyers without damaging long-term relationship-building efforts?

Structural Analysis

Applying the Value Chain lens reveals that Marketing and Sales represent the primary area of inefficiency. The current outbound logistics for samples function as a sunk cost with a 87.2 percent failure rate. Using a Jobs-to-be-Done framework, the sample is not just a product; it is a risk-mitigation tool for the buyer. The strategic problem is a mismatch between the buyer need for physical verification and the company need for capital efficiency.

Strategic Options

Option	Rationale	Trade-offs	Requirements
Aggressive ML Implementation	Eliminate samples for all leads scored below a 40 percent probability threshold.	Maximizes cost savings but risks missing late-blooming high-value accounts.	Full integration of ML scores into the CRM system.
Tiered Sampling Strategy	Provide full-size samples to high-score leads and small swatches to mid-tier leads.	Balances cost reduction with market coverage.	Redesign of sample production line for smaller formats.
Hybrid Sales Validation	ML scores serve as a recommendation that sales reps can override with written justification.	Reduces internal resistance but maintains some human-error risk.	Training programs for the sales force on data interpretation.

Preliminary Recommendation

Champo Carpets should adopt the Tiered Sampling Strategy. This approach mitigates the financial risk of high-cost sample distribution while ensuring that potential buyers are not completely alienated. It provides a data-driven filter for the most expensive assets while maintaining a physical touchpoint for the broader lead pool.

Implementation Roadmap

Critical Path

Month 1: Data Cleaning and Feature Engineering. Consolidate disparate fair records into a unified database.
Month 2: Model Validation. Run the Random Forest algorithm against 2017 data to test predictive accuracy against known outcomes.
Month 3: Sales Integration. Deploy lead-scoring dashboards to the sales team ahead of the next major trade fair.
Month 4: Pilot Launch. Apply scores to the Domotex fair leads and monitor sample requests against predicted scores.

Key Constraints

Data Quality: Inconsistent entry of buyer details at trade fairs will degrade model performance.
Sales Force Adoption: Veteran staff may view the algorithm as a threat to their autonomy or commissions.
Model Decay: Changes in global economic conditions or interior design trends may render historical data less predictive.

Risk-Adjusted Implementation Strategy

The transition must include a 20 percent buffer for sample distribution. This allows sales managers to approve samples for high-potential leads that the model flags as low-probability due to a lack of historical data. Success will be measured by the reduction in sample-to-order ratio rather than just total cost savings.

Executive Review and BLUF

Bottom Line Up Front

Champo Carpets must transition to a data-driven lead scoring model immediately. The current 12.8 percent conversion rate on expensive physical samples is unsustainable. By implementing a Random Forest predictive model, the company can reduce sample waste by an estimated 25 percent in the first year without sacrificing revenue. This is a shift from a push-based sales culture to a precision-targeted marketing operation. The recommendation is to approve the pilot for the upcoming trade fair season.

Dangerous Assumption

The analysis assumes that buyer behavior from 2013 to 2017 remains a valid proxy for future purchasing. In a market driven by rapidly shifting aesthetic trends and macroeconomic volatility, historical patterns may fail to capture emerging buyer segments or new geographic demand centers.

Unaddressed Risks

Algorithmic Bias: The model may penalize new markets or younger firms that lack historical data, effectively locking Champo out of high-growth emerging segments. (Probability: Medium; Consequence: High)
Operational Rigidity: If the production team only responds to ML-validated leads, the company loses the agility to respond to unexpected market opportunities surfaced by human intelligence. (Probability: High; Consequence: Medium)

Unconsidered Alternative

The team did not evaluate a Digital-First Sampling strategy. Utilizing high-fidelity 3D rendering and augmented reality for initial buyer screening could eliminate the need for physical samples for the first two stages of the sales funnel, regardless of the ML score. This would address the cost issue at the source rather than just managing the distribution.