disney.com

Deep dive on e-mail network-based Recommendations

Article Series on Organizational Network Analysis and Communication Content Analysis

Following from Part 1: We have combined various measures to infer further insights in the form of recommendations to employees on how to improve their communication, and to whom to talk for advice and support. In addition, through these recommendations employees are enabled to reach out to influencers and advisors to make an organizational impact.

These recommendations are modular and can be integrated in any system due its API service implementation. In this article, we focus only on two examples:

1. To increase your impact in organization reach out to <user>.

2. Improve communication and collaboration with <user>.

It is assumed that the <user> has given consent to be recommended.

Below you see how these recommendations look like:

1. Increase your impact in organization

Everyone wants to have more impact in own organization, but sometimes it is difficult to know how. It is hard to navigate in large complex organization, and even to have an idea what to do next to increase own impact, and to have your voice be heard. Employees need to be enabled, connected, supported, and trained for all the good reasons. We focus on social aspects and communication flow optimization.

How our impact recommender works?

Imagine a situation where an employee, Maria, is accountable, transparent, punctual, and keeps her manager, Laura, always informed. Maria is consistent and applies the same principles in communication with other relevant peers. Employees like Maria have the first ingredient to make an impact in organization. Imagine also that Maria is still not highly connected, or connected enough, to the right people in the organization. Next step is to confirm that that Maria is not already highly connected (otherwise less point in recommending, she knows the right people). Once we have filtered out the key influencers and we know Maria is not one of them, we continue to find out if Maria is already connected with any of the top influencers in the company — Maria should be loosely connected with at least a few influencers.

Next step, we make sure that these influencers (potentially line manager) are not within actual close communication range with Maria (if they were, it would either be her line manager or particular organizational communication setting that would have already enabled her to reach out). We identify the communication clusters of the influencers with whom Maria has reasonable, but yet not strong communication. From these clusters, we identify other employees John and Rita (not the identified influencers) that belong to the same cluster and have high edge betweenness centrality with the actual influencers (this means that influencers are in good communication with John and Rita, hence can provide a reference of support about Maria).

Once we have identified John and Rita, we go on and we recommend them to Maria — usually we recommend three people. Now, it is up to Maria and John or Rita to engage in communication and arrange their collaboration. Below, you see a simple form of a diagram for this use-case:

2. Improve your collaboration with colleagues

It is important for employees who rely on communication to be able to perform, to actually communicate effectively. This is true for most of office employees, whereas less true for blue-collar, field workers, etc. We make use of communication as a tool to identify counter-parts with whom your communication and collaboration should be improved.

How our collaboration recommender works?

Imagine a situation where Michael is highly involved employee and communication is one of his main tools. We focus on emails but this could be any written communication with source and destination, be it instant chat, comments, or feedback. As Michael communicates or exchanges emails, there will be communication patterns, decay, and high peaks — this depends on how work evolves, how relationships change, and how company structure changes in relation to overall communication. We analyze Michael’s communication clusters (people with whom he communicates frequently) while including temporal aspects, too. We conclude that Michael communication rate is relatively high — this means that he is a key contact point for a lot of topics.

Next, after we have identified people that are close to Micheal, we apply time-series analysis to understand how communication patterns evolved. We focus on communication patterns that tend to decay — for any reason whatsoever. Once we have identified few employees who communicated most with Michael, we apply Edge Betweenness to understand how it changed throughout time. We finally identify Tom and Jason, with whom Michael’s communication has changed. This change in communication could be due to anyone’s role, location, or organizational change, however none of these changed — and yet, communication patterns did change.

Finally, through our recommender implementation we suggest to Michael to meet Tom and Jason and discuss possible improvement options. It is obvious that there are other reasons outside of reach of our concepts but it is left to aforementioned employees to judge the situation and act upon recommendations. The main value from this concept is the opportunity that our recommender introduces to improve important lack of communication between employees.

The diagram below, shows the concept in a simplified form:

The concept of product definitions is outside of scope of this article. As a hint, Product Managers would define these products to be user-friendly and within privacy boundaries. Although, some measures are already in place — we don’t reveal the names of influencers that were utilized to identify the recommended colleagues. In addition, we assume that the recommended employees or coaches, have already agreed to be recommended and are willing to support other employees.

Our recommender can be used as an ‘excuse’ to initiate talks and improve collaboration. It is very often the case where employees communicate and work together only when there are no other options — this tends to be simply due to historical, contextual and perhaps even due to personal incompatibilities. Our recommender supports organizations to break the ice and bring people together, ultimately increasing communication effectiveness and productivity.

We combine and weight various algorithms, akin to ensemble methods. Few of the metrics that we utilize are: Hubs, Eigenvector Centrality, Louvian Modularity, Edge Betweenness Centrality. We also make use of contextual data such as locations and roles. In future, we plan to add business logic and contributions or goals that have been agreed with line managers.

This article is part of the series on ONA@Haufe

________________________________________________________________

  1. How to use corporate e-mail analysis to reveal hidden stars and ensure equal opportunities (Part 1)
  2. Technical overview of e-mail network-based Insights (Part 2)
  3. Deep dive on e-mail network-based Recommendations (Part 3)
  4. How to use trends to find hidden stars and work on a perfect project? People analytics will make you a star (Part 4)
  5. How to implement e-mail content-based analysis (Part 5)

________________________________________________________________

    Agron Fazliu

    Written by

    Head of Data Science at Haufe-umantis AG

    Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
    Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
    Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade