Towards the end of last year, I met up with employee advocacy and human network analysis guru Marie Wallace at IBM to discuss driving engagement and business results using scores.

 

Here’s what she had to say.

 

TB: Hi Marie, can you let us know a bit about your role at IBM and what you’re up to now?

MW:

I work in IBM’s Analytics Group on emerging technology. Previously I worked on the natural language processing technology which is now part of IBM Watson, our cognitive computing platform, although these days I’m all about analysis of human networks. Over the last couple of years I’ve been focused on an effort called Project Breadcrumb, which is an engagement analytics system that ingests collaboration and social data, from systems like IBM Connections, to measure employee engagement. The algorithms incorporate more than 15 years of IBM research into social networks and reputation. Our system measures engagement activity, reaction, eminence, and network to provide IBM Connections users with detailed scores on their internal social effectiveness. Now we’re expanding our vision to create algorithms that go beyond simply the social network but to other systems as well.

 

TB: What is an algorithm in your context?

MW:

Our algorithms take into consideration a wide variety of interactions, that are represented in the graph as nodes, edges, and their respective properties, in order  to generate a set of scores for each individual. The algorithm represents a significant competitive advantage for us and our clients as it’s based not only on good data science but also good social science, and has been proven through a number of business value experiments to accurately characterize individual engagement. This is absolutely critical because as they say “be careful what you measure as you might just get it” and if the algorithm is measuring and rewarding the wrong behaviors than it is detrimental to your organization.

 

TB: So do your clients ask you to create the algorithms for them?

MW:

Yes! We are seeing an increasing number of internal and external clients interested in using these types of network analysis techniques to meet a number of business goals. It might be an organization looking to implement organizational change programs where they could benefit in understanding who are the influencers or information brokers across their organizations in order to leverage them to maximize outcome. Or perhaps knowledge redistribution is a key objective where experts and expertise distribution, such as social sharing behaviors, are critical.

 

One question I always ask clients before they create any metric is to think very clearly about the outcome they are trying to steer. If you measure people to become more engaged then they engage more. If you measure number of deals closed then they will close more deals.

 

However, you have to be careful – sometimes you can optimise for the individual and not for the organisation.  Taking our personal social dashboard as an example; if you just chose to measure and reward individual “activity” then you’d generate a company of spammers, whereas if you include “reaction” and “eminence” then you drive more thoughtful sharing; which is critical to redistribution of knowledge throughout a company.

 

A great algorithm will in fact take “gaming of the system” into account and still ensure the right business outcome.

 

TB: So can an algorithm change behaviour?

MW: Absolutely, with social science we know that whatever culture you want in your organisation then the measurement and reward system you put in place can help drive that. By making this analysis available to every individual in the company, and simplifying the presentation so that everyone can understand what is driving the scores, much the same as you guys at Rise are doing, then we can use the scores from an algorithm to drive and reward the desired behaviour.

 

TB: So what advice do you have for business leaders wanting to create their own algorithms to drive behaviour?

MW:

Well firstly I would encourage you to bring all stakeholders, analyzers and analyzees, into a room to agree first what outcomes you want and then decide what the metrics should be. You are likely going to be constrained by the data you can get hold of, so secondly I would recommend that you start to identify the data you have or that you could start to collect; it’s always amazing to me how much data companies either have, but are not using for analysis, or that they could have but aren’t collecting. And finally consider potential privacy and security concerns early into the project and ensure that these are taken into account.

 

TB: Yes, we’ve also seen that the best scoring programs are those where the players have a real voice in how the algorithm creates the score.

MW: Yes, it’s really important that the analyzees – employees in our case – are part of the discussion and that they are comfortable and happy with the analysis being undertaken, and the privacy controls. Finally it’s important not to go too fast; getting the algorithm right and ensuring employees are happy with it, is more important than making it available for use by other parts of the business. For example HR usage in performance reviews. This will come over time, but we must walk before we run and bring all stakeholders with us on the journey.

 

TB: Great insights, thanks Marie. Do you have any other materials you’d like to share for people setting out on their own people analytics journey?

MW:

I’ve got heaps more content on this subject on my own blog and for those wanting to deep dive then there’s our Engagement Analytics team at IBM.