Q&A with Phill Agnew: Smart machine learning for modern market research

Phill Agnew (@p_agnew)
Director, Product Marketing, Brandwatch
Publication date: 09/28/20

What do researchers use X to learn about? How is research conducted?

Researchers lean on X to better understand the live opinions of consumers across the globe. Using Brandwatch, a X Official Partner, researchers gain access to all posts dating back to 2010. 

That means they can measure consumer trends over time, spot new topics, and understand current consumer opinion. This data informs decisions at the world’s largest organizations on everything from marketing strategy to product development.

From an AI standpoint, it’s sometimes difficult to tell exactly what a user means. If a post said, “I love apple and tesla.” It’s very difficult for software to read this post and tell if the user means Apple the company, or apple juice.

How have researchers tried to tackle this problem?

For researchers, creating a search that only collects the desired context of certain words can be difficult. Let’s continue with Apple as an example. Searching for the term “Apple” will bring back mentions of the company, but also mentions of apple pie, apples (the fruit), apple juice, even Apple Records (the record company founded by the Beatles). 

To remove all these unintended contexts of the term “Apple”, researchers have to spend hours writing a complex search that only looks for posts that reference Apple the company and not any of the other uses of that term.

This type of search is hard to write. Only expert researchers can create them and even experts have to spend hours pulling a search like this together. 

What’s the solution? 

posts are great because they often contain a lot of helpful content. When we collect a post from X, we’re not just collecting the keywords (like “apple”), but also snippets of text before and after the keywords. 

Take this hypothetical post copy: “So much faster than before! The latest Apple update is fantastic.”

Writing a regular search to collect posts like this without including any and all mentions of Apple would be hard. There are no unambiguous words like “iPhone” or “Mac” in the post which would determine that the mention is referencing Apple, the business. 

Luckily, Brandwatch now has a simple, intuitive solution to the problem. Using smart machine learning, we’ve been able to train an algorithm that analyzes the structure of a post before and after the keyword. By taking that information into account, the system is able to accurately predict that this post is for Apple the company, rather than for fruit or the record company.

How will this new technology help researchers? 

Today, when researchers need to search for an ambiguous term like Apple, they no longer need to spend hours writing a detailed search. Instead, they can use Brandwatch Search (which leverages this smart algorithm) to only find the mentions they need.

How will this help companies that leverage X data?

With this new feature, companies that use X data will benefit from three huge advantages:

  • Time saved: A single post can be highly influential, so fast analysis is vital. Search will help companies create powerful X searches in seconds keeping them ahead of every new opportunity and threat. 
  • Cleaner, more accurate data: Search gives you confidence that the results returned are about the subject of your Query, allowing better decision making.
  • Everyone is an analyst: Gone are the days where research was left to a single person in your department. With Search, everyone can start to get value out of X data. Users can search in seconds with no training required, helping everyone across an organization make smarter, data-driven decisions.

If someone wants to try this tech out, where should they go?

Ready to get started?