Reducing Time-to-Hire and Finding Niche Candidates

Summary: Maison Battat, a global toy company, recently hired a uniquely qualified data science candidate in only 26 days. Phase describes how we worked with Maison Battat to build a text mining (i.e., “Natural Language Processing” or “NLP”) approach that could scan 1,000s of candidates in minutes and look for unique attributes and experiences shared by applicants and broader candidate pools alike.

Jump to: Problem | Solution | Impact | Testimonial | Request a Demo

Problem: too many candidates and poor screening options

Most applicant screening systems are limited in how they enable searching and filtering of candidates. This means that they rarely allow you to include unique combinations of skills or score candidate profiles in a more flexible manner.

Worse still, many job applications have hundreds or thousands of applicants. If you expand this to a passive pool, you can be looking at 10,000+ candidates for a single role. Hiring managers and recruiters struggle with reviewing so many profiles, and resort to keyword searches.

Solution: use recent advances in text mining to make candidate searches descriptive and holistic

To address the challenge above, Phase uses two strategies to solve this problem.

1. Make the search a conversational process

Rather than simply providing keywords, we ask recruiters to use a freeform description of the candidate they are looking for. To use the Battat example above...

A conventional exact-match keyword search might look like:
Python”, “SQL”, “analyst”, “engineer” or “toys
But through our search, the candidate is described as:
“A data scientist who has with experience with toys, education, or children’s products.”

We use descriptions because our search understands concepts and ideas. The algorithm knows that a “data scientist” is a person that is likely to know languages like “Python”, “SQL”, or others and takes on roles such as an “analyst” or “engineer”. By teaching the algorithm to seek out candidates who have experience with “toys”, “education” and “children’s products”, we can find people with relevant experience in related areas like “gaming” or “youth development”.

Another way to think about this interface is that the recruiter simply has to write an answer to the question, “What sort of candidate are you looking for?” You need not worry about the specific structure of your response, or fitting keywords into specific parts of a search form. The above example could easily be “A former toy designer interested in analytics” or “A multilingual French-speaker who can design products and analyze data.”

Our goal is to give the algorithm an idea of what type of candidate to look for. It will identify and make connections between concepts to find the strongest candidates. It is not limited by specific keywords. This means that a recruiter or hiring manager saves time by automating searches, while generating a broader diversity of qualified candidates.

2. Search the whole resume, not just skills lists or keywords

Our semantic approach makes it easier for us to scan an entire resume to understand the person as a whole. For instance, a candidate might outline an interest in “children’s products” in one part of their resume, but not include this in their core skillsets elsewhere. This semantic approach tracks the entire resume and scores the themes that come up rather than just individual skills or keyword flags.

Impact: 26 days for time-to-hire and a great candidate experience

Maison Battat is a family-owned toy company that encourages kids to be bold, curious, and playful. For over 45 years, they have offered a range of engaging toys for babies, toddlers, and young children including Driven™ tough trucks to dolls of Our Generation™.

Battat wanted to hire a unique candidate with experience in marketing, e-commerce, data science and analytics. They sought out someone who was a self-starter, fast learner, proficient in another language, has lived abroad, and shares their passion for improving the lives of children through play and education.

Our text mining approach above was used to analyze over 1,000 data science candidate profiles. The top result was Sogra, a bilingual data analyst with international experience. Importantly, she was the ultimate self-starter having created an award-winning smart toy while she was working at a toy startup.

Sogra was the first and only candidate interviewed – she was perfect for the role. Not only was the role filled in 26 days, but both employer and employee were thrilled with the significantly easier process and speedy approach.

In their own words...


“Phase reached out to me about a data analyst role at a toy company. Two weeks after I was introduced to the hiring manager, I accepted their job offer. I’m excited to have a data role that leverages my background as a toy designer. I feel amazing!”


“Phase has been a fantastic talent partner for our company. We hired the first candidate they sent us -- she was experienced in our industry and had a great analytics background. We went from first candidate introduction to first day on the job in 26 days.”

Head of Amazon Business Unit,
Maison Battat



Ready to get started?

© 2021