Social listening can be vital in understanding how your messaging, products, and brand are resonating with your customers so you can adjust them before you’ve spent your entire marketing budget on a campaign customers don’t care about.
Are you ready to hear what consumers are saying about your products and services? Then Microsoft Social Listening (MSL) is the service for you. The service will be part of the upcoming Spring Wave, and was showcased at the recently held Convergence 2014 conference in Atlanta, GA
In MSL you setup searches to listen to social media conversation and learn the sentiment (how people feel about your brand on social).
As a natural next step you might decide for one or more campaings to execute in Microsoft Dynamics Marketing to influence the sentiment and generate more leads, and subsequently work with those leads in Dynamics CRM to bring in more revenue. A brilliant closed loop.
In this blog post I'll show you how to set up search terms in MSL, and help you understand what MSL can do for you in terms of sentiment analysis.
Set up searches to listen to social media conversations
Using MSL you can listen to conversations on social media around your brand, business, products, and competitors. The first step in successful social listening is setting up your search topics to ensure you’re capturing the right conversations. Once you’ve captured these conversations, you can fine-tune your searches, analyze your data, and drill into the information that matters most.
Available sources for search topics
MSL covers a set of sources where public posts are published.
The following sources are available:
To create a search topic
It’s important to start with a clear idea of what you want to listen to. Then proceed to set up a new search topic and add at least one search query to your topic.
Here is an example of one of my running Search Topics ("CRM Online"), with three queries in it
And here is the result of the above Search Topic; Buzz (# posts), Trend, Share of voice, Sentiment, Languages (majority of posts in English), and Buzz history in my time frame.
Note that the Sentiment Indicator (grey/red/green donut chart) tells me that the majority of posts I've picked up about CRM Online are positive (more about Sentiment later in this blog post). I can click eg the red part of the Sentiment Indicator to drill into the negative post and perhaps learn whats behind the sentiments and improve.
Keywords define the words and phrases you want MSL to listen for. Keywords are exact but not case sensitive.
I suggest you include variations of the keywords (eg singular and plural form of the word) to ensure you get the desired result from your query. Separate each keyword with a comma to invoke an “OR” condition.
If you have added more than one search term, your query looks to find at least one of the listed terms. It’s important to review keywords regularly. If your keywords yield too many results, think about narrowing the query by adding inclusions and exclusions or reduce the number of keywords.
Sentiment analysis in Microsoft Social Listening is target-specific. A post’s sentiment value always refers to the keywords of a search query only. If you are looking for sentiment values about your brand, you must add your brand name as a keyword.
Inclusions narrow down your search so you get a much higher quality selection of posts and results. Think of inclusions as an “AND” conditiion. Your search will be filtered so that posts are selected only if they contain both the keyword AND the inclusion. Inclusions aren’t case-sensitive.
When you set up your search query, you can choose from the following options to decide how close a keyword and an inclusion must appear in a post:
I recommend starting with the default option (paragraph). If your search topic yields too many irrelevant results, try narrowing it to the proximity in a sentence. Note that this may also remove relevant posts because all combinations of inclusions and keywords outside of a sentence will no longer be picked up by the application.
Inclusions are an efficient way to reduce the number of posts resulting from your search, and are a great way to make sure you stay within your quota.
If you in the Keyword field put X1, X2 and in the Inclusiion field put Y1, Y2 then you have in fact built the query (X1 OR X2) AND (Y1 OR Y2)
Sometimes a specific word can flood your results with irrelevant posts. Exclusions enable you to narrow down your searches and improve your results. Enter the words to exclude separated by commas and your searches will ignore posts containing these terms.
Exclusions aren’t case-sensitive. For every term you add to the exclusions, your search will be filtered so that posts won’t be selected if they contain both the search term AND the exclusion in the same paragraph.
Exclusions are an efficient way to reduce the number of posts resulting from your search, and a great way to make sure you stay within your post quota. Choose your exclusions carefully to avoid missing relevant posts.
Understand sentiment analysis
Sentiment analysis determines the attitude of an author towards the context of a topic. It reflects the public perception of a post’s content in relation to the keywords that were used to find the post.
Sentiment calculations only work if you select a search topic because sentiment values are always related to a certain keyword. section.
Select a search topic from the drop-down menu in the top left corner of the filter
After you’ve selected a search topic, you'll get a "Sentiment" page with several Sentiment sections.
The "Sentiment summary" appears at the top of the page and displays the total number of posts and the sentiment index for your chosen search topic. You’ll also see the percentage of posts where you either confirmed or edited the sentiment.
The "Sentiment summary by sources" shows an overview of the sentiment index on all active sources.
It provides the total number of posts per source, the sentiment index for your time frame, and the change in sentiment index, compared to the past five time frames (the reverse colored numbers inside the bars). The bar below the sources indicates the relative distribution of sentiment values in your data set.
The "Sentiment history" diagram visualizes the sentiment index over time. The black line indicates the sentiment index while the green line shows the average sentiment index in your time frame.
In the "Sentiment volume" section you see the sentiment values of the posts in your time frame, separated by day, By default, neutral posts aren’t displayed. To see the neutral posts, click Neutral in the legend.
I can even drill down to post to see whats actually written (eg by clicking the bars in the above chart).
In the below screen shot I've drilled down into a (positive) post in which the author has tweeted about (my keyword) "Microsoft Social Listening". The green part of the sentence is what MSL has interpreted as a positive statement
I can filter posts in a variety of ways, eg to see only post where the author has a high "Reach" (similar to Klout score). Access the filters using the "+" icon in the upper left corner.
In the result set, I can even open the original post (in this case on Twitter)
About Sentiment value
Each post that results from your defined search queries is processed by the sentiment algorithm in the original language and annotated with a calculated sentiment value. Sentiment values are provided for the following languages:
The sentiment value results in a positive, negative, or neutral sentiment for a post.
Occasionally, the algorithm identifies positive and negative parts of a sentence and still rates the post as neutral. This happens because the amount of a post’s text identified as positive or negative cancel each other out. A post is also classified as neutral if there are no positive or negative statements detected in it. Note that the sentiment algorithm is not a self-learning system, even if you can edit any post’s sentiment value in the post list.
The sentiment values from posts with positive or negative sentiment that match your defined filters are normalized and result in the sentiment index for your search topic.
About Sentiment index
The sentiment index represents the general perception of the active search topic. It indicates the ratio of positive and negative posts over all posts that match your filters. Neutral posts aren’t taken into account. Sentiment values are calculated in relation to the keywords that define a search topic. That’s why you need to provide context for the calculations by defining the focus of your analysis as the target for sentiment. In analytics, you need to focus on a search topic to provide a target for the sentiment index.
For example, you maintain two search topics: Your Brand and Your Competitor. A blog post mentions your brand and your competitor. The author writes a negative statement about your competitor and a positive statement about your brand. If your target for sentiment calculation is your brand, the sentiment value of the post is positive. If your target for sentiment calculation is your competitor, the sentiment value of the post is negative.
The sentiment index is normalized to a value between -10 and 10. All your active filters and parameters are taken into account to define the data set for which the sentiment index is calculated.
A sentiment index of:
You can calculate the sentiment index with the following formula:
Sentiment index = (Positive posts – Negative posts)/(Positive posts + Negative posts) * 10
You find the change in trend next to the sentiment index. Microsoft Social Listening compares the sentiment index of the five previous time frames to the current value of the sentiment index in your time frame.
To see more about the analyzing features of MSL, please see "Are you learning? Analyzing using Microsoft Social Listening" - link
Great posting. I have a much better idea of how this works now
Very nice description of Social Listening! Thanks
I actually understand this now. Thanks for the walk through!