Samstag, 18. März 2017

Delve and the Office Graph Inside Out Part II

In addition to my first post about the insides of Office Graph and Delve (Delve and the Office Graph Inside Out) this article is focusing on which signals are used by the Graph to generate the individual Delve experience.

Signals used by the Graph

You can find all the information you need about signals used by the Graph in this msdn article: https://msdn.microsoft.com/office/office365/howto/query-Office-graph-using-gql-with-search-rest-api
Based on this article we have the following Action Types:
  • PersonalFeed
  • Modified
  • OrgColleague
  • OrgDirect
  • OrgManager
  • OrgSkipLevelManager
  • WorkingWith
  • TrendingAround
  • Viewed
  • WorkingWithPublic


As you can see in the msdn article the list of signals can be dived in private signals and public signals so that data privacy is respected all the time:

In addition Mark Kashman published an article on Microsoft techcommunity about Understanding security and privacy of Delve and intelligent experiences in Office 365. In this article, we can find the following diagram:
So we can extend the list taken from the msdn article to this aggregated version:
  • PersonalFeed
  • Modified
  • OrgColleague
  • OrgDirect
  • OrgManager
  • OrgSkipLevelManager
  • WorkingWith
  • TrendingAround
  • Viewed
  • WorkingWithPublic

  • Member of
  • Created by
  • Shared with me
  • Direct reports
  • Public

Some of this signals are clear like for example Modified, Viewed, Created by, etc. some others are a little bit mystic like TrendingAround. We can imagine what TrendingAround means, but we cannot get an information about how this signal is processed in all details.
Anyway, it is easy to understand how this signals are used to generate the individual Delve experience.

The myth about the People suggestion in Delve

It is easy to imagine how content suggestions are generated based on signals. But one of the most asked questions about the Delve experience is about the difference between People list on the left and Related People in my personal Delve feet. Based on the signals list we can definitely get a better understanding about this. People on the left are other users we visited in Office 365 respectively we have clicked on their Delve profile. Related People are based on signals like “Member of”.
So for example if you are Member of
  • Member of a Distribution List in Exchange Online
  • Member of a Office 365 Group
  • Member of the same Manger or “Direct Reports” entity

this is processed by the Graph to generate the Related People overview in your Delve feet.
As we can see also in this scenario data privacy and data security is respected by the Office Graph and Delve. If you are a “Member of” the same Distribution List or Office 365 Group, you can see all the other members anyway.
More details about this and also about compliance in Delve can be found in Mark Kashmans article “Understanding security and privacy of Delve and intelligent experiences in Office 365” I mentioned above.

Related articles: