Onedot Platform FAQ

The supplier catalog is processed directly (straight through processing) and an ERP/PIM import file is generated. The intermediate results of the categorization, attribute mapping, normalization, etc. are displayed to the user via so-called feedback tasks, in which validation and overriding of the automated suggestions is possible.

The feedback is then implemented in a new processing and an updated ERP/PIM import file is created. In the case of a catalog update, the feedback effort is minimal, since the mapping profile is applied from the initial connection.

At the beginning of the collaboration, the product data model (PDM) is mapped by the dealer in the Onedot platform and continuously kept up-to-date via a synchronization process with the ERP/PIM.

This means that Onedot knows the categories, attributes, value lists, etc. for creating the automated mapping proposals. The product data export (PDE), i.e. the extract from the merchant ERP/PIM, is used for the initial training of the Onedot platform. This provides the merchant with a merchant-specific machine learning model for automated onboarding.

Onedot supports the automated import of common formats such as CSV, Excel, XML, TXT. For specific formats, the import may need to be configured once by an Onedot data specialist. There is no catalog structure requirement and the structure is allowed to change. Onedot automatically maps new categories, attributes, values to the merchant target model. When mapping for the first time, it is recommended that the user validates the automated suggestions. The more the supplier changes on his side, the more validation is recommended by Onedot.

The time required depends on the size of the catalog, i.e. number of SKUs, attributes, categories, and can take from a few minutes to several hours. To keep the feedback effort for the user as low as possible, intermediate results are displayed aggregated. This means that when categorizing, for example, the user may not check each product, but at the product segment level. The attribute mapping is structured in such a way that attributes can be checked as "globally" mapped, but the software maps this on a category-specific basis.

The customer can define the relevant onboarding modules at the beginning of the collaboration. The onboarding process usually includes the following modules:

1. import product data
2. Product data synchronization with ERP PIM
3. Extraction (attributes are extracted from continuous text)
4. Categorization (products are assigned to dealer categories)
5. Attribute mapping (supplier attributes are assigned to dealer attributes)
6. Normalization (of numbers, units, value lists)
7. Creation of ERP/PIM import file

Onedot has a consumption-based pricing model, which means that you only have to pay for what you actually use. The price is made up of 2 components.

a) basic fee of the platform, divided into so-called tiers

b) Single price per supplier, the more suppliers the cheaper the single price

The basic fee is usually paid by the merchant, the unit price is usually paid by the supplier itself. The number of SKUs and the number of users are not a price driver and are unlimited.

This is highly dependent on the data quality from the supplier catalog as well as the data quality from their current PIM/ERP, which is used to train the Onedot Machine Learning algorithms.

During initial onboarding, the accuracy of the results is usually between 60% and 70%. The accuracy of the results increases through the iterations, whereby values of over 80%, and in some cases even up to 96%, are achieved in a timely manner. For comparison, humans have an accuracy of about 80%.

There are four ways in which product data can be brought onto the platform:

1. onedot application: manual upload/download (drag and drop) via app.onedot.com

2. onedot SFTP: automated data exchange via sftp.onedot.com

3. onedot API: automated data exchange(www.onedot.com/api)

4. email: send the product data to a specific Onedot email address, which will deposit the catalog in the Onedot app.

A distinction is made between PDFs for product data onboarding and PDFs for asset data onboarding. The former are read semi-manually/automatically and converted into a machine-readable format which then goes through the normal onboarding process. Which data is extracted from the PDF is defined in advance with the customer. Asset PDFs go through normal steps such as file renaming or provision as a public link.

Onedot runs on Amazon Web Services (AWS) and the data is stored in the EU jurisdiction. SLAs define how long the data is stored. The rights of the processing results are fully owned by the customer. For more information, please see the Onedot Terms of Use: www.onedot.com/terms-of-service

Machine learning is used in Onedot in such a way that the mappings for the individual onboarding steps do not have to be created by the user himself, but that the software performs these mappings automatically. Various syntactic and semantic text analyses are used, including word embedding techniques as well as deep learning / transformer-based language models. The Onedot Machine Learning Research Team ensures that the Onedot Platform is based on the latest machine learning techniques and that proprietary algorithms are developed that differentiate themselves in the market.