Here you will learn how to uncover what consumers think and feel about your brand, and why.

  • Understand consumer perception of your brand by monitoring large-scale, online conversations.
  • Identify the drivers of sentiment and use these insights to establish a scalable process for attracting new customers.

Brand Analysis Guide


Brand Analysis Case Study

Problem: Online conversation about Uber had skewed overwhelmingly negative for years. At the beginning of 2015, 44% of all tweets about Uber were negative. Strategists needed to quantify drivers of consumer opinion about the brand to identify actionable insights. Segmenting the conversation into custom categories using an Opinion Monitor provided a solution that scaled (45MM tweets).   

Insight: The research found only a small percentage of complaints about bad drivers, ride costs, or poor experiences. The data showed a working product that people liked to use. Instead, overwhelmingly, Uber’s negative sentiment was driven by press surrounding then CEO Travis Kalanick and several high-profile legal issues (1).

Action: This analysis proved that online conversation about negative press overshadowed positive consumer experiences. Uber teamed up with Cheezburger and the ASPCA to bring people kittens on demand. The #uberkittens campaign provided a positive subject of conversation and decreased negative sentiment around Uber during the campaign and after.


What is Brand Analysis?

A Brand Analysis evaluates brand performance using data derived from social media conversation of consumers. Performance indicators include the volume of brand conversation, tone of the conversation, topics discussed, and more.

Each brand (and industry) is unique, and therefore we must consider how to define what data is relevant and how to segment that data in order to extract meaning from it. The purpose of this guide is to provide researchers with a clear, straightforward roadmap to execute a custom brand analysis.

The process to undertake any brand analysis breaks down into 4 parts:

(1) Defining the data

(2) Segmenting results

(3) Synthesizing meaning

(4) Creating a report


1. Defining the data

Construct a sample frame to identify relevant data using keywords

With trillions of data points available, the biggest challenge for researchers using social data can be isolating the relevant data. Online conversation about a brand or organization is composed primarily of two streams.

First, “owned conversation” centers on social media posts directed at a brand–or more specifically, a social account or property owned by the brand. This compares loosely with traditional media analytics in terms of paid advertising. This can provide insight into a brand’s “fitness” by measuring its efforts to gain traction through the lens of consumer engagement.

Second, organic or “earned conversation” aligns with the concept of earned media, traditionally thought of as brand mentions in newspapers and magazine articles. Most simply, this measures what people are saying among themselves. This data has much to offer, however, it can be time-consuming to build a search query that excludes irrelevant data because the way people converse online is dynamic.

Isolate relevant vs. irrelevant data

The visualization below outlines the central challenge with defining relevant vs. irrelevant data for social listening when it comes to earned media. The ideal universe of data is represented by the red circle. Because of the nature of unstructured data we are sometimes forced to make compromises between precision and recall when establishing the research sample frame, represented in the grey boxes.

The following tools are available to help define your sample frame and determine the precision and recall of any given research question. More detail on the specific execution of each option is provided in later sections of the guide:

A) Keywords: Should be balanced in consideration of ambiguity. Even the perfect keyword set is subject to the unpredictability of unstructured text. For example, a person may accidentally type “ge” instead of “get” and their content would be included in a monitor for General Electric that includes even the strictest exclusions.

B) Machine Learning: Crimson Hexagon’s Brightview algorithm may be trained to recognize text patterns common among irrelevant content for your research. This is ideally leveraged in more advanced stages of brand analysis.

C) Image Analytics: Our image analytics technology provides over 5,000 classifiers for objects, scenes, and actions, including, for example, a classifier for NSFW content that may be applied to Twitter and Instagram (and potentially Imgur looking ahead to 2019).


2. Segmenting results

Create a customized data-set by using automated and custom categories

Not done. 


For an automated analysis (aka basic sentiment and emotion analysis) create a Buzz Monitor using keywords that include variations of your brand name and associated products, accounting for misspellings (see below for specific instructions).

Auto Sentiment



You can also create an Opinion Monitor for tracking overall brand sentiment and breaking it down into categories such as specific brand offerings, brand qualities (e.g. reliability or taste.), or drivers of positive and negative conversation.

Keyword Categories

Machine Learning


3. Synthesizing meaning

Surface actionable insights by leveraging business questions

Not done. 

If you are just getting started with brand analysis then reporting on topline results of your research may be enough. Once you begin regularly reporting on these metrics, you can then begin to expand the scope of your analysis.

Beginning with a clear business question can make a big impact on your ability to synthesize meaning from the data. Identify a clear, targeted question about what you wish to learn:

What do people like and dislike about my brand’s customer experience?

Are topics of conversation aligned with my brand’s positioning?

What services or product lines are customers speaking about most? 

Is the proportion of positive conversation outperforming my industry?

Is my brand reaching the appropriate audience?

Using these questions you can determine where you should direct your efforts in terms of building additional monitors for competitive brands, creating advanced filtering, or leveraging the capabilities of an opinion monitor to create custom categories. As you evolve your approach brand analysis to become more sophisticated keep these business questions front and center.


4. Creating the report

Not done. 

Employing a framework for the analysis helps keep communicate your findings in way that is focused and deliberate. Basic  frameworks have the following basic structure:

1. Volume

2. Sentiment

2. Topics




Creating your monitors

1. Keywords


An exclusionary approach is useful when a product or brand has an unambiguous name that is rarely used to refer to anything other than the brand. The boolean operator: AND   is the primary operator for an exclusionary keyword design.


Searching for the brand ‘Clean and Clear’ you may wish to add exclusions for other common uses for the words “clean” and “clear” to proactively eliminate irrelevant mentions.

“Clean and Clear” AND –(“so clean” OR lake OR sea OR ocean OR river OR sky OR skies)


Some words and phrases are commonly associated with content that most would consider “spam” or irrelevant. But what may count as spam for one user may be essential data for another. One man’s trash is another man’s treasure when it comes to online content, and so we recommend developing a set of exclusions that fit your definition. It doesn’t need to be perfect, start small and grow your list of exclusions over time.

AND –(“chance to win” OR contest OR giveaway OR contests OR giveaways OR #win OR eBay OR Amazon OR #deal OR dailydeal OR dailydeals OR #discount OR #discounts OR #follow OR #ff OR followfridays OR “follow Fridays” OR nsfw OR #free OR #freebie OR “click here” OR #job OR #jobs OR #hire OR #hiring OR #save OR #savings OR “promo code” OR “promo codes”)



When it’s difficult to identify word use because there are too many variations, we may choose to use an inclusionary approach to exert more control over the sample frame. We may also favor an inclusionary approach at the industry and category level. The boolean operator OR  is the primary operator for inclusionary design


Searching for the brand Express we might wish to include only posts which also have words directly related to the apparel category.

(iPhone OR iPhoneX) AND (“i want” OR “really want” OR “i need” OR “really need” OR “just want” OR “here i come” OR “can’t wait” OR “cant wait” OR “cannot wait” OR “can not wait” OR “gonna get” OR “gunna get” OR “going to get” OR “wanna get” OR “wanna buy” OR “gonna buy” OR “gunna buy” OR “going to buy” OR “should i buy” OR “should i get” OR “should get” OR “must buy” OR “waiting for” OR “waiting until” OR “wait for” OR “wait until” OR “waited for” OR “waited until”)

2. Data Sources


Track customer engagement with each of your brand’s social media accounts through Social Account Monitors.

Through HelioSight, you can quickly and efficiently search for a top-level overview of your audience and conversations surrounding your brand.

Request a Demo

Ready to transform your business?

Get a walkthrough of Crimson Hexagon and learn how consumer insights can help you make better business decisions.