Sentiment Analysis

Understanding the data that you have is essential to building the right solution and utilising the correct classification methods.


What is Sentiment Analysis?


Sentiment analysis uses the wealth of data available from social media to provide vital information on brand perception over time or how a brand is viewed in relation to competitors. This text could be in the form of tweets, LinkedIn posts, Facebook comments or free text from a feedback form. By analysing this data with modern analytics techniques (such as Natural Language Processing) we can uncover much deeper insights than a simple positive or negative brand score.

  • Establish the link between brand sentiment and revenue
  • Grow brand value by up to 30%
  • Pre-empt and prevent negative customer experiences

Why use Sentiment Analysis?


Our clients typically adopt our sentiment services for one of the following reasons:

Inform brand personality and positioning


Understand how your brand sits in the customer’s mind. Using machine-learning techniques combined with sentiment score, we can understand which adjectives are most associated with positive or negative value. This enables us to identify blue ocean terminology to inform brand personality and strategy and ensure that marketing campaigns are based on distinctive positioning.

Monitor brand dynamics and respond to crises


Sentiment Analysis enables granular understanding of how brand perception is evolving at a daily, or even hourly level. This analysis could reveal which product launches generate positive or negative responses or understand the impact of a marketing campaign. Alternatively, a Brand Monitoring System could be setup to detect negative anomalies in brand perception and enable a rapid response to an emerging brand crisis on social media.

Improve the customer experience


By combining Sentiment Analysis with customer service data, we can uncover insights into how easy it is for customers to solve. These insights can then be embedded into the customer care process, where customer service response is prioritised based on a Sentiment Algorithm pre-emptively identifying particularly unhappy customers. By identifying and acting on these insights a business can ensure a smooth and hassle-free user experience for their customers.

How does Lynchpin do it?


Data review and preparation

Understanding the data that you have is essential to building the right solution and utilising the correct classification methods. Cleaning the data and transforming free form text into a more structured format can be time consuming however it builds the foundation for a successful outcome.

Classifier selection

There are a number of classifiers that have been used for sentiment analysis, and we can ensure the industry leading classification algorithms will be used in sentiment analysis. It may be that the best classifier is making use of more than one, an ensemble of classifiers.

Setup Reporting

Depending on reporting needs, we have experience and access to many different dashboarding techniques and tools to convey the information and results in a clear and reliable manner, allowing for real time actionable insights.

Feature Engineering

Once text has been normalised, feature engineering is performed to identify which features will prove most useful when deciding the sentiment of the text. There are several approaches this can take, for example n-gram models, bag of words approach or word vectors can all be used to represent the structure of the text. Drawing on our experience in this area we can select the correct method for your needs.

Train Classifier

Training the classifier requires a large dataset which has been converted into features and labelled with a polarity. The actual label can depend on the granularity. At this stage, the data will need to be labelled, for which we will make use of experienced services ensuring high levels of consistency. Once the classifier has been trained and tested to an optimum level of accuracy, it will be production ready for real-time reporting.

Analysis on live results

Working side by side with you we look to constantly monitor and improve the solution. With in-depth analysis of results and recommendations for refining the classifiers going forwards we ensure the solution is fit for purpose and scalable within your organisation.

Case studies


As an independent, full-service analytics consultancy, our team are flexible – equipped with a range of practical experience, skill, and commercial awareness.

Contact us


If you’d like to explore how we can help, please get in touch with our data and analytics consultants. We don’t have salespeople and your first point of contact will be a subject matter expert from our leadership team.