Pulse for Good Uses ML to Turn Cumbersome System into Showstopping Product Feature
Pulse for Good is the world’s first Human Services Feedback Management Platform, used by its customers to efficiently collect, analyze, and act on input from vulnerable populations. Pulse’s self-service kiosks are a safe space where people can anonymously share their true feelings and lived experiences.
- Pulse for good uses Nyckel text classification to categorize and filter 100,000+ pieces of feedback data
- Significant reduction in manual review time
- Supports strong customer retention
- AI filter sale team’s “favorite product feature”
About Pulse for Good
Pulse for Good is a startup with a unique mission. The company provides human services providers, like homeless shelters, refugee resettlement agencies, mental and behavioral health facilities, and other similar establishments, with specially-designed kiosks and supporting technology that allows people to leave anonymous feedback in a safe and private space.
“We help social service organizations around the world more effectively listen to their clients,” said Blake Kohler, CEO of Pulse. “So, if you think of the classic reputation management or customer experience system, we do that – but for people experiencing trauma or vulnerabilities.”
The Challenge
The surveys collected by Pulse’s platform include the option to leave open-ended feedback. This feature allows people to record any message they wish for the service providers.
The responses received run the gamut from sincere (“we could use shower curtains and more case management”) to flippant (“can we get more cocaine and hookers?”) to incoherent. With around 100,000 pieces of feedback in their system, the team at Pulse was becoming overwhelmed with manual content moderation and sorting.
Pulse for Good quickly realized it needed some form of natural language processing (NLP) to filter feedback data — separating valuable feedback from spam or comments that was more offensive than constructive. With limited time and resources, Pulse also needed something that didn’t require too steep of a learning curve.
“There are lots of natural language processing platforms out there that you have to do a lot of work to get up and running,” said Blake. “And as a startup, a barrier to entry like that would be way too hard for us to overcome and still get any return on our investment.”
Blake realized that a machine learning API with clear and specific documentation was the way to go. But finding an API that offered great value and a great user experience was another matter altogether.
The Solution
The team at Pulse looked into several options, including products from Microsoft Azure, Amazon, and Google. But with each platform they tried, they inevitably ended up frustrated. The team found that they kept running into the same two problems: either the product was not at all intuitive and required too much work, or it was far too costly. That’s when Blake discovered Nyckel.
“With Nyckel, I didn’t ever run into any roadblocks,” said Blake. “Everywhere else, I’d start implementing something and hit a wall. Nyckel just flowed all the way through!”
When first using Nyckel, Pulse for Good focused primarily on managing spam. But soon after Blake and his team started training their first Nyckel model to review client feedback, they experienced a light bulb moment.
“We started realizing Nyckel could do much more for us,” said Blake. “And so we quickly jumped from just doing spam into breaking things down by sentiment: happy, sad, disgusted, and so on.”
Now, instead of having an overwhelming backlog of feedback that slowed the team’s progress, they could quickly sort feedback into helpful, distinctive categories that offered incredible value to their customers.
And not only that — Pulse’s product feature allowed its customers to see how Nyckel labeled their clients’ feedback. Customers could then verify whether or not Nyckel had categorized it into the correct sentiment (e.g.,” disgust” vs. “anger”) and the correct category (e.g., “suggestion” or “complaint”). This consistent customer input has allowed Pulse to continually re-train Nyckel’s model, resulting in an even more accurate and efficient system.
The Results
For Pulse for Good, the benefits of using Nyckel go directly back to the experience of its customers: the human services providers. Thanks to Nyckel, these providers can now filter and sort feedback, giving them quicker, more accurate insight into the needs of their clients.
“It really makes the feedback we get more valuable. Using machine learning changed it from just data to actual information,” said Blake.
All of this has directly equated to saved time and energy, along with more renewals and customers. And not only that: the Pulse sales team absolutely loved Nyckel. To date, it’s their favorite feature ever. “It took part of our application that was overwhelming and cumbersome and made it into something that is a showstopper during demos,” explained Blake.
Pulse for Good has already enjoyed a lot of success implementing Nyckel to categorize and filter content, but they’re not stopping there. The team hopes to expand Nyckel’s categorization abilities to create a “marketplace of ideas,” where service providers can share their experiences and ideas about improving their organizations. These insights will then be automatically categorized so that they can be easily accessed and shared by Pulse customers.
Blake and his team are also looking into ways to help their customers understand how they are doing in relation to other service providers. Benchmarks for human services providers will help customers identify opportunities for improvement, so they can better serve the vulnerable populations in their areas.
Inspired by Pulse for Good’s success story and curious about what Nyckel might be able to do for you and your specific use case? Check out Nyckel’s product offerings. It’s free and easy to get started. Don’t hesitate to contact us if you have any questions.