Keyword Analysis for Microsoft Teams
Certain subjects or keywords will gain popularity in your company, but tracking their proliferation across the organization can be challenging. At tyGraph our rich data model allows you to easily understand what is being discussed in Teams or Yammer.
This article will focus on drawing insights from Microsoft Teams. For information on Keyword analysis in our tyGraph for Yammer tool please see this article:
Also this article focuses on using the default visual layouts in our tyGraph for Teams tool. See this article if you are looking for advanced analytics examples by manipulating the report model:
Threads vs Responses vs Messages – 0:40
You can search by threads or messages in these examples. Though many people use those terms interchangeably, at tyGraph they mean specific things. It’s worth explaining that we categorize a Thread as the first block of text that initiates a conversation in a channel. A response is any text underneath the thread. A message is any body of text at all (thread + responses beneath it).
Filtering by Keyword for Teams – 1:52
The simplest form of keyword searches can be done with an attribute in the right filter pane. Using the report level filter you can search for all messages that contain a word.
If you cannot find this attribute and you’re in the Power BI service or desktop you can open the model and add it from the Messages table. If you are in tyGraph Online and you don’t see this filter it means that your organization has disabled it for your report. Please reach out to me or a colleague at your company to see if we can get you permissions for a report that has this attribute exposed.
- Open the filter pane and scroll down to an Attribute named “Message” or “Message Body”
- Select the attribute and switch to “advanced filtering”
- Select “contains”
- Type the word you would like to find.
- Hit apply filter. Done!
This will show you all conversations that contained that keyword. You can move between all pages with Message data to understand that keyword. Keep in mind other measures, like Teams, channels and SharePoint Activity will still show everything. But you can see the number of teams you reached which we will cover next.
You can draw a raft of insights from keyword analysis. Below are frequently asked questions around this topic.
- How is this keyword trending over time?
- Are more or less people engaging with the keyword?
- Who has used this keyword the most?
- Where are people using this keyword? (teams/channels)
- Is the conversation far reaching or concentrated within select individuals?
- Are comments around this keyword of a positive or negative sentiment?
How is this keyword trending over time? – 3:20
We can use the dynamic date axis to move up and down within time frames. So if you need to answer how many messages are being posted over a larger time frame, then change the report period, and bring the axis up to an aggregation that makes sense for your situation.
Finally if you want to understand that time users are posting messages with this keyword the most. Drill down to the Time axis by pressing the double down arrows. This plot will now use Hour UTC as the aggregation.
For a detailed explanation on how we use hierarchies in our date visuals see our article here: http://tygraph.azurewebsites.net/tygraphkb/plotting-activity-over-time-in-tygraph/
Are more or less people engaging with the keyword? – 5:56
The simple way to plot this the unique users who posted messages is with your keyword is by looking to the message contributors line. However, we know some users may participate in a subject by just reacting to those messages without replying. Also we may want to know the percentage change in users or include things like users who reacted to the conversation. For this we will start using the “Engaged Members” visuals as well.
The engaged members count is any user who took any one of the acts of engagement that we have selected in the MAE Score Metrics. The reason this is important is because we can finesse this filter to fit our scenario.
First, if you select just Messages in the MAE Score metrics box, and click a single day you will see the engaged members match. This is because only showing users who were engaged by posting a message for the report period.
With that understanding we can now expand the MAE Score Metrics to include Reactions. Now, this number tells us the number of users who reacted to conversations with this keyword. Not just those who posted it. Which is why the numbers are different.
Who has used this keyword the most? – 10:05
For a simple list of users who posted messages with your keyword:
- go to the “Teams and Channels” page
- See the Team Members chart, any users with messages are people who posted a message with that keyword.
- From here you can sort by clicking the header or ellipsis
- If you would like you can also select the chart and filter the visual to only show people with messages.
- Finally you can export the list of those users.
As shown in the same Team Members table. You can also easily spot users who have given or received likes.
Another item to analyze is users who had the most influential messages with this keyword listed in the top influencers visual. The influencer score is our own algorithm that rewards users for gaining reactions on their posts, being mentioned, or generating responses from a number of colleagues.
Where are people using this keyword? – 12:35
In the bottom right corner of the recent activity page we have Teams and Channels by MAE Score. MAE Score is a non weighted total of the items selected in the “MAE Score Metrics” box for the report period.
- After filtering by your keyword go to recent activity page
- The Team MAE Score will be listing the number of actions by each team.
- Use the hierarchy to navigate between teams and the channels beneath them.
Is the conversation far reaching or concentrated within select individuals? – 14:36
By filtering to your key word in the Messages attribute you can map users who participated in the same conversation or 2nd degree users from salient in conversations on the same keyword. The purpose of this is not to find the flow of information but rather:
- To see what users are at the center of a discussion involving multiple users.
- To see if what users connecting otherwise separate tribes of a discussion.
- Spot users on the fringes of a discussion.
Are comments around this keyword of a positive or negative sentiment? – 18:48
You can tell if conversations around your keyword are positive or negative from looking to the bottom of the Team Sentiment Chart. This will tell you the overall average sentiment with the average sentiment by team and channel for messages with your keyword.
Plotting sentiment over time is done below so you can see if your keyword is changing nature over time or if there are events that are drastically negative/positive.
Finally it all comes together with the Key Influencers visual. This visual tells you why sentiment is at its current level. It analyzes your data, ranks the factors that matter, and displays them as key influencers. For full details on the visual itself check out Microsoft’s fantastic article here: https://docs.microsoft.com/en-us/power-bi/visuals/power-bi-visualization-influencers
The core of the visual are two plots, Key Influencers and Top Segments. Key influencers, will show you top attributes that cause your sentiment to increase or decrease by a certain amount. You can hover over the influencer bubble to have the explanation spelled out for you.
Each bubble will be highlighted in the chart to the right. This bar chart tells you how much messages with the influencing attribute are above the average sentiment (dotted line) for your selected messages. You will notice that the average dotted line though the bar chart is equal to the average at the bottom of the Team Sentiment table.
The key influencers visual will identify a person, team, channel, or other attribute as influential if it has a sentiment significantly outside of the mean with a decent sample size. You can tell by hovering on the bar chart that will be highlighted. Sometimes another user, team, attribute, will have a higher average sentiment but fewer messages or a higher standard deviation so the influence is not considered to be influential.
You can switch between what influences sentiment to increase or decrease using the drop down at the top. This is very important depending on what type of conversations you are targeting.
Finally if you are looking to focus on culprits that are causing your conversations around a topic, you can also switch to top segments. This is very helpful for generating actions for you and your team. For example posting messages where this user replies are the key factor in causing the conversation to be positive in the Running Gulp Mix – General Channel
Other insights may be more intuitive, such as sentiment is likely to be lower on days other than Friday!
You can answer a wealth of questions about a keywords without ever leaving the safety of the default tyGraph for Teams reports. As you should guess, this is only the surface. The tyGraph for Teams model can handle a variety of even more advanced scenarios which we will explore in our next article that you can find here: coming soon!