DX Innovation

AI Buyer Potential Evaluation Model: Data-Driven Buyer Scoring

What Is the AI Buyer Potential Evaluation Model?

The AI Buyer Potential Evaluation Model is a data-driven system that identifies the most promising overseas buyers for Korean exporters by comprehensively analyzing purchase likelihood, financial stability, transaction history, and industry fit. Developed by the KOTRA DX Innovation Lab, the model addresses the subjectivity and inefficiency of traditional staff-dependent buyer assessment and enables scientific buyer management based on objective indicators.

When AI scoring is applied to the hundreds of thousands of buyer records registered in the KOTRA buyer database, tasks previously evaluated manually by trade office staff can be automated, while hidden high-potential buyers can be uncovered. The value of AI-based assessment is especially high in emerging markets such as Bangladesh, where buyer information is often not systematically organized.

35
Evaluation Factors
Multi-dimensional scoring
800K+
Buyer DB
KOTRA global DB
86%
Prediction Accuracy
Deal-close prediction
5 levels
Score Grades
Grades S-D
Real time
Refresh Cycle
On data change
86
Countries Covered
Linked to all trade offices

Five Core Scoring Dimensions

The AI buyer evaluation model analyzes 35 detailed factors across five dimensions to generate a composite potential score ranging from 0 to 100. The weight assigned to each dimension is adjusted dynamically according to industry and market characteristics.

Five Core Dimensions of Buyer Potential Assessment
DimensionWeightNo. of FactorsKey Indicators
Financial Stability25%8Revenue, credit rating, debt ratio, years in business
Purchase History25%7Import volume, transaction frequency, product fit
Commercial Fit20%6Product match, order size, payment terms
Growth Potential15%8Sales growth, new business, market expansion
Communication15%6Response speed, consultation history, interest level

AI Algorithm Architecture

The buyer scoring model uses an ensemble approach, combining the predictions of multiple AI models to produce a final score. This reduces the bias of any single model and enables more effective processing of diverse data types, including both structured and unstructured information.

Data Collection
Integrate the KOTRA buyer DB, external credit databases, trade statistics, and web-crawled data
Feature Engineering
Extract 35 evaluation factors, handle missing values, normalize variables, and create derived features
Model Ensemble
Combine three models: XGBoost for structured data, BERT for text, and LSTM for time-series signals
Score Output
Generate a 0-100 composite score, an S-D five-level grade, and dimension-level subscores
Dashboard Delivery
Provide buyer scorecards, grade-change alerts, and recommended follow-up actions
Grade S Buyer Profile
Overall Score85-100
Deal Conversion Rate65%+
Average Transaction Value$100K+
Recommended ActionPrioritize immediately
Grade D Buyer Profile
Overall Score0-30
Deal Conversion Rate< 5%
Average Transaction ValueUnverified
Recommended ActionFurther validation needed

Using Buyer Scoring in Bangladesh

Buyer evaluation in the Bangladesh market comes with distinct challenges. Formal financial information on local companies is limited, the informal economy occupies a large share, and corporate structures are highly diverse. The AI model applies Bangladesh-specific evaluation criteria designed to reflect these characteristics.

01
RMG Buyer-Specific Assessment
For buyers in apparel and textiles, the largest industrial segment in Bangladesh, the model emphasizes export performance, factory certifications such as LEED, WRAP, and BSCI, and supply records with global brands. Buyers holding factory certifications receive credibility bonus points automatically.
02
Use of Unstructured Data
When formal financial data are limited for Bangladeshi buyers, the model uses NLP to analyze unstructured information such as websites, social media, and news coverage in order to assess business activity and reputation.
03
Integration of Local Trade Office Intelligence
The model incorporates feedback from on-site visits and consultation records maintained by KOTRA Dhaka staff, supplementing indicators of actual transaction intent and reliability that data alone often miss.
04
Monitoring Grade Changes
Automatic alerts are triggered whenever the score grade of a buyer changes, whether upward or downward. If an existing Grade A buyer falls to Grade C, the system flags it as a risk signal and enables a proactive response.
AI Scoring for Export Consultation Dialogue AnalysisTechnology for assessing buyer seriousness in real time during consultations
KOTRA Navi Genuine Buyer Discovery SystemKOTRA official AI platform for discovering genuine buyers
Buyer ScoringAI Evaluation ModelGenuine BuyersData AnalyticsBuyer DatabaseDX Innovation
AI Buyer Potential Evaluation Model: Data-Driven Buyer Scoring | Dhaka Trade Portal