SENTIMENT & TOPIC ANALYSIS

About Sampler.io

Sampler.io is a digital sampling platform that helps CPG brands created targeted sampling campaigns allowing consumers to try samples from the comfort of their homes.

DESIGN PROCESS

EMPATHISE

Studying the current state - Insights Report

At Sampler, at the end of every sampling campaign, our in-house client success team works on creating an insights report. Unlike traditional street corner or store handout sampling, targeted digital sampling can help brands know & reach their audience better. For this reason, these insights are useful for brands to help them understand how their product is being perceived when it comes to adoption, advocacy, differentiation and awareness.

The report looks like this:

To be able to reach this level of depth to provide brands helpful takeaways and actionable insights, the team has to go through the reviews manually to identify the sentiments, common themes and topics that have emerged in the ratings & reviews left by consumers who tried the product sample.

When the quantity of reviews is in thousands, which more often than not is the case, the team has to rely on external tools to help them with review analysis. This leads to added cost for business.

Why do brands need these insights?

In addition to reducing the manual effort that the team has to put in to create an insight report, interviews with client success team helped uncover how a Sentiment & Topic analysis dashboard Sentiment & Topic analysis dashboard can help CPG brands understand their target consumers better:

• Monitor their influence and reputation

• Identify meaningful trends at the product level

• Get qualitative product-specific insights that provide the why behind the what to drive ACTION, that otherwise would have gotten lost in a sea of reviews

• Respond effectively to their customers

• Significantly cheaper way to get an early understanding of how their products are being received by their target consumers in the market, as compared to traditional product feedback mechanisms

DEFINE

Sentiment Analysis

THE WHAT

Sentiment analysis is the automated process of understanding underlying sentiment or opinion of a given text. Implementation of this tool helps product reviews and sort them by positive, neutral, or negative sentiment, which is then presented to the brands.

In the Sampler universe, consumers leave ratings and reviews for the products samples they tried. These ratings & reviews come in huge volumes and can be extremely useful for brands to learn how consumers feel about their products. But reaching that stage of understanding can require a lot of time and manual processing of reviews to truly understand the sentiment behind every individual user review.

Topic Analysis

THE WHAT

Topic modelling or topic analysis is an unsupervised machine learning technique that is implemented to scan the sea of reviews received on samples by consumers to identify the common topics & top themes of conversation & present the analysis to brands.

In the Sampler universe, where we have a large text, like in case of reviews, what brands need is a way to see what are the common topics that are being discussed by consumers. For example, for a skincare brand sampling a moisturizer, it will be useful to learn if the consumers are having conversations on topics around the product’s gentleness, comedogenicity, hydration etc and what are the associated sentiments around those topics. Topic analysis dashboard can provide those insights within a few clicks.

The Challenge

Replace a currently manual & time taking end of the campaign ritual with in-house insightful dashboard to help CPG brands understand the perception, adoption, and advocacy of their products by their target audience.

IDEATE

Collaboration

The first step to creating a Sentiment & Topic analysis dashboard that presents insights to the CPG brands about their sampled products included understanding how the ratings & reviews left by the consumers will be analyzed. At this step I collaborated with our in-house data science team to understand the machine learning mechanism that would be employed for data processing of reviews before they could be presented on the user end.

TESTING

Usability Testing

Remote moderated usability testing sessions were conducted with 10 business users to validate and test the designs.

• Users found the new dashboard to offer insights quickly, and available to everyone in the team without any dependency.

• Users were excited to find insights around how consumers were feeling (sentiments) and what words were being used to describe their products in positive, neutral and negative sentiments

• Users liked the flexibility to play around in the dashboard and get insights at all levels - overview to in-depth.

Impact

Launching Sentiment & Topic analysis dashboard had the following impacts:

• Manual effort of processing reviews reduced by 99%. The dashboard enabled a self-serve approach where brands can come and access the insights and analysis readily available on their Sampler accounts.

• Brand engagement with Sampler analytics increased by 37% with brands having the ability to login & check their product performance & perception through ratings & review analysis. (Source: Hotjar)

• With Sentiment & Topic analysis becoming a new Sampler feature, business retention grew 25% (Source: SiSense)