Join our mailing list for intranet and digital workplace links from around the web.

We’re careful with your personal information. Read our privacy statement for more about how we manage your details, and your rights.

Get in touch

Make your intranet work harder for you. Contact us to see how we can help.
+44 (0)1244 458746

Microsoft Graph – Opportunity or echo chamber?


Microsoft Graph – Opportunity or echo chamber?

Microsoft Graph aims to make search smarter by prioritising content created by people close to you in your network. But is this a good idea? There is a trade-off between search efficiency and the reinforcement of silos. Organisations should think about the transparency of content promotion algorithms, particularly in light of the Facebook / Cambridge Analytica revelations.

This collaborative post by Matt Varney (KCTCS), Agnes Molnar (Search Explained) and Sam Marshall (ClearBox) came about from a Twitter exchange at a recent conference. We outline the risk, consider mitigations, and explain how Microsoft Graph works.

Modern search and the Graph

At Microsoft Ignite, September 2017, the big news was Satya Nadella’s demonstration of new search capabilities in Office 365. Modern, ’personalised’ search became a popular topic very quickly.

Why? – Because it was, and still is, very promising. Instead of using the classic search index, it relies on the Microsoft Graph, which provides not only search results, but recommendations based on your previous activities and relationships. The experience is always personalised. What I can see is targeted for me. Anyone and everyone else in the organisation will see a different set of content in a different order.

Modern search ranks results by what is closest to me in the Microsoft Graph. This is calculated based on my relationship to the documents, their authors, and also the timeliness of the content. For example, a document that has been modified yesterday by someone who I work with regularly will be ‘closer‘ to me than something that has not been opened for months, and belongs to someone who I rarely communicate with.

This way the suggested documents are always ‘relevant‘ to me. But this kind of relevance should be rather called ‘timely‘, I feel. Another challenge is that we don’t have real control over what is in the Graph, or what is displayed to our users.

Modern search does a good job when we want to get back to the recent work of ours or colleagues, or to discover content even if we don’t know if exists. But with the limitations above, we might have the impression that something is missing…

It’s important to mention that we still have the classic search center in Office 365, too, with its well known, customisable capabilities. However, classic search cannot get the benefits of Microsoft Graph today. See a detailed comparison by Search Explained.

Social distance and silos – what’s the risk?

When Britain voted to leave the EU, just about everyone was surprised at how narrow the vote was. A reason for this is that social media has created an echo-chamber effect. Everyone was hearing that their opinion was the dominant one because nearly everyone else in their network felt the same.

Companies can be similarly surprised when their internal networks become too siloed. We know that uniformity of opinion is bad for innovation. We also know that the echo-chamber is poor for risk-assessment, because everyone is blind to the same things.

The more we work with people, the more we think like them. The more our search results amplify the views of people close to us, the faster that will happen. That’s what Modern search is doing. Social capital isn’t neutral and can be damaging to companies if strong allegiances are not aligned with organisatonal goals.

However, silos are useful sometimes. For example, they can play a strong role in the diffusion of innovations once part of a group has chosen to adopt a new idea or approach.  Silos can be a great help in getting day-to-day work done because they build trust, strong tacit knowledge, and a specialist vocabulary (‘jargon’ to outsiders).

What should organisations do?

People naturally want to be around others with whom they agree, and simply want comfortable and familiar conditions and scenarios. This is nothing new and it means that silos are almost inevitable. The trick is to spot them, and then actively work to dilute them to break down the walls and separations where appropriate.

So we’re not talking about deliberate manipulation Cambridge Analytica style, but a natural tendency in organisations. Technologies like the Microsoft Graph do not necessarily create silos, but they can amplify them and allow them to grow larger and at a faster rate. Because their scale and scope is different, the normal analog tools and methods have had to adjust, and may not have kept pace.

Some traditional communications methods used to break down silos in the analog world are evolving and extending with ‘Work Out Loud‘ approaches that use the power and speed of technology to deliver communications and ideas to newer recipients. The walls can still come down.

However, using traditional ways to identify the silos in the first place can remain difficult.  The search indexes of the Microsoft Graph could provide the network mapping, for example, and help us understand where silos are forming.

The technology is able to not only index the content and information within an enterprise, but it can also track all the actions of the people creating and consuming that information.  It can provide the first truly omniscient view of what an organization does, and how it is doing it, and with whom the information is shared. Data driven decisions – in real time and at organisation scale – could potentially occur with regularity.

The organisation could use these data to identify the numerous internal silos from the view from the top, and then work to tear them down as needed. Further, the organisation could find and celebrate successful efforts to work out loud, which helps build momentum for working more openly.

What should individuals do?

Perhaps the most intriguing aspect of working with this new and powerful perspective is at the personal level. Individuals can see the organisation in new ways from their point of view and through the lens of the now exposed (to them) personal trends. They can see more clearly and visually that they work with certain people or groups more or less often than they may have thought.

Modern search still works to return results from right across Office 365 if you know how to use it.  Martin White makes many valid points about how transparent we should be with personalised (graph-based) search. But once you have this understanding it provides an increased chance to discover a new colleague or collaboration partner within your organisation, someone with similar interests and experiences.  Those serendipitous moments around connecting to new people and new content can add joy and satisfaction to everyday work.

Silos happen. Humans and technology working together can help diminish their negative influences such as isolation, groupthink, and blind spots of the bigger picture.  At the same time, those forces can accentuate positive actions, such as connecting with previously unknown content and collaboration partners.

Sam Marshall

I'm the director of ClearBox Consulting, advising on intranet and digital workplace strategy, SharePoint and online collaboration. I've specialised in intranets and knowledge Management for over 19 years, working with organisations such as Unilever, Astra Zeneca, Akzo Nobel, Sony, Rio Tinto and Diageo. I was responsible for Unilever’s Global Portal Implementation, overseeing the roll-out of over 700 online communities to 90,000 people and consolidating several thousand intranets into a single system.

  • Posted at 12:15 pm, 29 March, 2018

    I can certainly see, in principle. the benefits of personalised search even if I have some concerns about it. However the load on the search support team will be very significant, assuming they exist, There will be a constant requirement to explain and refine the filters. Moreover relationships are not static, as this excellent video from Autodesk illustrates https://www.autodeskresearch.com/projects/orgorgchart. Typically around 10% or perhaps more of employees leave or join each year, and perhaps another 10% (say) change roles. Employees may find that colleagues that they thought they were associated with seem to have vanished? But will their work prior to the organisational change or resignation still be tracked and visible. So that means that in a company with 10000 employees perhaps 2000 profiles will need to be re-curated each year. Let’s say there are 200 working days a year, so on average there will be 10 profile changes a day that might need care and attention from the search team. These are just some of the search management issues that I have come across with highly personalised search, and they cannot be solved with an IT Desk ticket.

      • Posted at 12:43 pm, 29 March, 2018

        If the employee is new to the organisation they will need to know who has been deemed to be their colleagues/nodes. It may be visible on the Graph but in increasingly virtual organisations the nearest person in terms of network links could be thousands of miles away. Someone is going to have to provide a guide as to what they will find when they search, and probably only the search team will have the admin privileges to know the connections, and also be able to tweak them if the nodes are not ideal. But that may still require a joint sign-off with the manager concerned as there could be security implications, and that could take time.

        Then of course there is the language issue – some of the members of the network may be working in local, rather than corporate language. How then will relevant information be found? I’m sorry about the long ramble but the overall concept of personalisation assumes a relatively steady-state/single role/mono-lingual business environment, and I think that is more unrealistic than search vendors appreciate. And of course all the above applies to “we can help you find all the experts in your organisation’ promises.

    • Posted at 11:18 am, 4 April, 2018

      Hi Martin,
      You have a good point. Curating people profiles has to be done in a similar but very different way than content curation. For example, if an employee leaves the organization, his/her profile should be removed from search. – But still, he/she is the author or contributor of many documents, therefore we cannot “remove” him/her from the index entirely.
      Also, how to help the users to find his/her successor?
      What if the successor, in the very same role, has a different job title? How will the system know it?
      There are many-many questions, and if we take your 20% rate above, it’s a tremendous job. Some might be automated, but in other cases it might be totally unclear how to deal with these challenges.
      I think the biggest challenge is that most organizations don’t even realize these challenges 😉

      The challenge with O365 / Microsoft Graph is that “removing” people from it cannot be done manually, we have to wait until they will be vanished. The other option is to remove them from the search index but it raises many follow-up questions, see above.

      This whole question might worth another blog post 😉

Post a Comment


This site uses Akismet to reduce spam. Learn how your comment data is processed.