Turning Eureka Moments Into Reality Through Data

By driving product decisions with data, these organizations have effectively determined what clients need.
Written by Avery Komlofske
January 19, 2022Updated: January 19, 2022

Eureka! You have the next big idea. 

Not only do you know it’s brilliant, but you have the data to back it up. You’ve been collecting and analyzing numbers in your intended market, and the data practically rearranged itself and smacked you in the face with inspiration. This product could change the world.

It takes more than an idea to make a successful product — you need to build it out, market it and keep making it better. While this may seem daunting, you already have the skill you need to start the process: data collection and analysis. 

It’s the first step toward fruitful decision-making. Gathering accurate data and reading it effectively can show you where, when and how to make your product work for your clients. Applying data-driven planning to your business will help you get your brilliant idea out into the right places, like it has with other great products and services.

Take it from Criteria Corp’s Danny Duncan, Advertise Purple’s Rowland Hazard and Ace Metrix’s Harsha Manjunatha. These professionals are successfully using their data to create and refine products that fill a need in their market. Built In LA sat down with these data analysts to learn about how they make that happen every day.

 

professionals pointing at data
Shutterstock
Danny Duncan
Sr. Data Engineer • Criteria Corp

 

Criteria Corp’s platform offers data-backed, objective assessments for job candidates to reduce bias and improve efficiency in the hiring process.

 

When did you first realize that your data may have some untapped value?

They say that money can’t make you happy — but money can remove many of the obstacles that can make you unhappy. So too, our approach to data is not to think of it as an augury of success, but rather a tool for removing the obstacles that prevent employers from having a talented, diverse and value-aligned workforce.

In many sectors, employers are finding it difficult to hire the talent needed to grow their business. When faced with these obstacles, they may not be sure about how to proceed. One method is chasing the same talent that every other firm is trying to get — but are there any other tools in the toolbox?

With data, our clients can broaden the pool of potential employees. The signals we provide in our assessments allow recruiters to compare quantitative measurements across factors such as aptitude, attitude and employee personality. With this information, employers can not only hire candidates who show a high likelihood for success, but also who may have been overlooked by their competitors. In this way, we give our clients a competitive edge in the hiring market.

This doesn’t just benefit employers. Candidates benefit from being assessed in ways beyond traditional evaluations like resumes and interviews — which have been shown to be poor predictors of job performance. This is particularly true for candidates who are new to the field, or for those from marginalized communities where we know there exists a measurable bias in hiring outcomes. We’re really proud of the work we’re doing by bringing equity to the process of hiring and talent management.

 

How did you bring this product to life? And how did you collaborate with other teams to do it?

While utilizing data is part of our DNA, right now we’re working to build out a data services team that furthers our data efforts to bring actionable insight into hiring and developing talent. There are a couple of things that make this exciting.

We do not accept inertia as a reason for using any data technologies. We spend time evaluating technologies that will help us disrupt the industry. As an example, we took considerable time evaluating potential data integration workflows, looking into both partners, open source technologies and in-house methods. In addition, everyone in our data team learns a range of technologies to address the many ways we help internal and external stakeholders. For example, we provide services for integration, orchestration, transformations, documentation, programmatic interfaces and internal user interfaces, as well as supporting a utilities package.

It is vital to balance research, product planning and execution.”

 

What’s the biggest technical challenge you faced along the way? And how did you overcome it?

The biggest challenge for the team is balancing the range of tasks that we perform to support our stakeholders. Right now we are experiencing a Cambrian explosion of data technologies and resources, some of which present significant value. It is therefore vital to balance research, product planning and execution. One way we work to address this challenge is using tools that help with the process of ideation and implementation.

 

 

Advertise Purple office
Advertise Purple
Rowland Hazard (he/him/his)
Vice President of Product • Advertise Purple

 

Advertise Purple offers a full suite of affiliate management tools for online brands and e-commerce companies.

 

When did you first realize that your data may have some untapped value? 

As soon as we began to aggregate our data at scale, we knew there was untapped value. Before Purply, data aggregation and analysis was time-consuming to the point that it became prohibitively expensive. As soon as we were able to leverage our platform, we had unparalleled access to insights in our and our clients’ verticals as well as any other index of our choosing.

 

How did you bring this product to life? And how did you collaborate with other teams to do it? 

Initially, we focused on identifying any and all pain points that Purply could solve; a focus on utility for end users, managers and executives enabled us to build a comprehensive suite of services that added value and synergy to each of our departments and the company as a whole. Once our MVP was live, a continuous feedback and prioritization process allowed us to quickly improve the platform and deliver high-value features.

As soon as we began to aggregate our data at scale, we knew there was untapped value.”

 

What’s the biggest technical challenge you faced along the way? And how did you overcome it?

The diversity of data sources was far and away our most difficult technical hurdle. Not only did we have to deal with a multitude of integrations, but we had to map each data source correctly across the entire platform to ensure 100 percent accuracy on both micro and macro analysis. A rigorous initial QA process — plus a series of daily and hourly automated data accuracy checks and notifications — allows us to be confident in our data and proactively diagnose and solve any issue as they arise.

 

 

two coworkers working together
Ace Metrix
Harsha Manjunatha
Senior Data Scientist • Ace Metrix an iSpot.tv company

 

Ace Metrix measures the performance of video advertisements and turns that data into actionable metrics.

 

When did you first realize that your data may have some untapped value? 

Our optional, open-ended responses provide us a treasure trove of information about each creative. Respondents often feel obligated to disclose their thoughts and feelings about the creative they just watched in our survey — the emotional sentiments that we classify and extract from these open-ended responses using natural language processing (NLP) models allow us to help brands accurately identify what emotions resonated the most among a target demographic and what didn't. This is a huge win for our clients in helping them improve their overall targeting and spending.

 

How did you bring this product to life? And how did you collaborate with other teams to do it? 

Working on the research team involves trying out different ideas and keeping up with the state-of-the-art technology in the industry to consistently improve our insights. When we have a working prototype that might be of importance to clients, we often disclose the results to internal stakeholders asking for feedback. When there is sufficient traction, we collaborate with internal stakeholders to iteratively improve the prototype, shaping it for a viable release.

Research never ends, and we often have to draw a line to ship a product. This usually involves the research team building an internal library for using the product. We collaborate with the corresponding engineering teams to expose a service or wrapper for it to be incorporated into the rest of our UI product suite facing our clients.

Working on the research team involves trying keeping up with the state-of-the-art technology in the industry to consistently improve our insights.”

 

What’s the biggest technical challenge you faced along the way? And how did you overcome it?

Oftentimes it’s difficult to replicate the same results from a state-of-the-art (SOTA) conference paper to your domain. Apart from the exorbitant computing required to train such large SOTA models from scratch, there are often a multitude of undisclosed tricks involved in fine-tuning such models to adapt and perform comparatively on one’s own data. This requires running several training experiments, trying different tricks and tracking the performance of those models on our data.

Apart from training, these models also require several thousand examples of human annotated ground truth — which gets increasingly expensive with the 57 different emotions that we measure. One solution that we came up with allowed us to scale up the number of training examples as we introduce new emotions. We were able to bootstrap our previous emotional sentiment classification system to discover and generate candidate examples of each emotion, which we call hits. We then use these hits to auto-generate millions of misses using a clustering approach in vector space. We have gotten much better results when the misses outnumber the hits since there are more ways to not say an emotion than to say it.

 

Jobs from companies in this blog

Los Angeles startup guides

LOCAL GUIDE
Best Companies to Work for in Los Angeles
LOCAL GUIDE
Coolest Tech Offices in Los Angeles
LOCAL GUIDE
Best Perks at Los Angeles Tech Companies
LOCAL GUIDE
Women in Los Angeles Tech