Specifications/Technical Information
CrowdANALYTIX
CrowdANALYTIX
CrowdANALYTIX DataX - Supporting product data onboarding process -
Product data onboarding process
The process of data onboarding * of product data is performed manually by the person in charge at many companies, and there are various issues.
*Data onboarding: Collecting, creating, and organizing data so that it can be used in business
Below is an example of a manual data onboarding process in a company.
- The person in charge visually checks the catalog data sent in PDF, transcribes the contents, and enters the product data.
- Since the product data sent from multiple suppliers/manufacturers has different formats, each file is checked and organized manually and manually entered into the system.
- In order to ensure the quality of the product data, multiple people in charge visually check the same product data many times.
- Because product data categories and classification systems are complex, the person in charge makes assumptions based on past experience and existing product category classifications, and performs product classification work.
Manual data onboarding process
Product data onboarding process challenges
- It is not possible to eliminate the dependence of the work on the individual, and the business continuity of the work is not guaranteed
- It is necessary to perform categorization and labeling work in consideration of multiple product characteristics and elements, and there are many things that cannot be handled based on rules, making judgments dependent on the person in charge.
- Depending on the label and category, it becomes a task that requires specialized knowledge to make decisions.
- As a result, only specific personnel accumulate business know-how and knowledge, making it difficult to continue operations when personnel are changed or new personnel are assigned.
- It is difficult to ensure uniform data quality
- During work, even skilled workers can make human errors in input, transcription, normalization, labeling, and category judgment, resulting in missing data and incorrect values.
- When the original data is changed, it must correspond to the change of the already registered data.
- In addition, due to the shortage of human resources, it is difficult to continue to assign sufficient human resources to maintain data quality for the ever-increasing number of products and product data items.
- As a result, it becomes difficult to maintain consistent data quality.
- 商品点数・データ量に対して作業工数が比例的に増え、コストが増大する
- Confirmation of original data, input and transcription, judgment of labels and categories, reconfirmation, approval, and work occur in each process
- Since it is necessary to create a database with a certain quality, as the number of products and product information columns increases, the work increases proportionally and it is not possible to reduce the work.
- Long wait times for data to be available as a result of data being passed back and forth between multiple people
- As a result, as the number of products and the amount of product data expand, the corresponding costs will increase.
CrowdANALYTIX DataX
CrowdANALYTIX DataX is an AI solution service that supports the creation of clean structured data using AI technology. While interlocking with existing systems such as MDM/PIM/ERP, it eliminates the dependence on individual skills in the product data onboarding process, streamlines it, and realizes the generation, maintenance, and provision of product data with consistent quality.
- Create clean product data faster, more accurately, and more completely.
- Make the data onboarding process less dependent, improve the quality of your product data, and streamline your work.
- It is possible to combine function modules according to requirements and expand functions.
This service provides a combination of different functional modules according to your requirements.
Some of the functions that can be provided are as follows. Please contact us for details.
| Extraction of product attribute items | Example: Extract product description from PDF catalog |
| Automatic setting of product classification | Example: Estimate the product category from the product name and assign a label |
| Product title generation | Example: Automatically generate product names for EC publication |
| digital asset collection | Example: Automatic cropping of product images from a PDF catalog |
| Commodity price monitoring | Example: Watching and Monitoring Competitor Prices |
| Automation of product information collection | Example: Supplier Web Monitoring and Information Retrieval |
Solution brief
- Upload raw data for product information to be processed to the DataX platform.
- AI (machine learning model) customized for each customer performs various processing on each product data to generate new product data.
1. Categorization
2. Addition of characteristic labels
3. Data normalization
4. Generating additional product information (eg. generating product titles from product numbers, functions, and manufacturer names)
*The above is just an example of the function. The functions can be expanded by adding function modules according to your requirements.
- Among the generated product data, only the product data judged to have a low accuracy rate (probability) will be sent to the review/correction process. Users can check, modify, and approve processing on the WEB GUI.
- The product data generated through the above process is integrated with the normalized original data and output/provided as structured product data.
- The structured product data can be downloaded via WEB GUI or used in conjunction with existing systems such as MDM/PIM/ERP using Web API.
upload data
Upload data to the cloud via WEB GUI or API.
- Product description/review
- Product image
- .pdf, .xls, .csv, .msg, .docx etc…
Data processing by AI
Various processing is performed by an optimized AI algorithm, and additional information is added to each SKU.
- 自動カテゴリ分類
- automatic labeling
- Automatic product title generation, etc.
review/correction
Check the newly given data by AI
Data with low confidence scores are selected and corrected by manual review
- Only check data with low confidence scores
- Add/edit information
- Final approval process
Data management system integration
Final approved data via system linkage API or exported data will be linked with PIM/MDM/ERP systems, etc.
- .json, .csv, .xml etc…
- Integrate with existing systems via API
- Backup data stored in the cloud
Benefits
AI transforms the product data onboarding process.
Case study
Consideration step
Considerations
- This AI solution service creates a learning dataset based on the raw data (images, text, etc.) owned by the customer and tunes the AI model. When classifying and labeling product categories, a certain amount of raw data with categories and labels is required.
- The service is provided using a cloud platform, and data is input and output via WebGUI or API.
- We do not provide trained AI models or training datasets created.
Document download
video on demand
Inquiry
Strengths of macnica.ai
- We provide one-stop resources for data analysis, AI model development, and implementation to support projects.
- Resolve issues such as lack of AI personnel, lack of knowledge, and inability to operate
Latest Information
Case Study
Blog
Document List