SDL Trados Business Manager can recommend vendors by applying different recommendation mechanisms:

1. Vendor groups
2. Machine learning
3. Previous work completed by vendors for particular customer, project and specialization
4. Vendor's willingness to work.

How to get recommendations

You can get recommendations in the following three places from the translation assignment editing window:

• Open drop-down menu in the Vendor field;
• Click Search icon in the Vendor field to open vendor selection popup window;
• Switch to the Candidates tab and click Add candidate button.

Depending which recommendation features are enabled, you could get four types of recommendation, each of which has own color code:

Orange - recommendation from vendor group;

Violet - recommendation by machine learning system;

Yellow - recommendation based on previous work completed by vendor;

Green - recommendation based on vendor's willingness to work.

Vendor groups

You can read detailed information about vendor groups in a separate article: Vendor groups.

Machine learning

You can enable and disable machine learning functionality from the application Settings > Translation assignments tab > Recommendations group > Disable machine learning for recommendations.

When enabled, application will be tracking vendors you select in translation assignments and build history model which will be used to predict vendors in new assignments. As you train the system (more assignments you create), you will be getting more accurate predictions. For predictions, system takes into account customer name, translation project, specialization, language pair, responsible person (manager), etc.

When you're viewing machine learning recommendations (violet), take a note on the Recommendation reason column. In this column, for machine learning recommendations, a score value is displayed, like this: 72.21/72.21. First part of this value is vendor's score, and second part is maximum score among all vendors. If score value equals or close to maximum score, and value is below 80, then such recommendation could be considered as not reliable. Recommendation with high probability of accuracy should be higher than 90.

Recommendations based on previous work

You can enable and disable machine learning functionality from the application Settings > Translation assignments tab > Recommendations group > Disable recommendations based on previous work.

Idea is simple - if particular vendor was already ever used for a current translation project, customer or specialization, this vendor will be recommended.

Willingness to work

In the vendor editing form, on the Settings tab, you can find a field called Willingness to work. This field allows to enter a value from 0 to 10. Zero value means that willingness to work isn't rated. Manager, depending on conversations with vendors, could rate their willingness to work using 1-10 scale. When deciding how to rate this, you could take into account different aspects: expressed intention to work with you, overall culture of communication, speed of replies, etc.

Depending on translation assignment parameters and number of found vendors with specified willingness to work, TBM could include up to 3 vendors into this recommendation group.