A Look Inside Vervoe’s Product Evolution

Share on facebook
Share on twitter
Share on linkedin
Share on email

The release of our Select feature is a new milestone for us at Vervoe. Not only is it a powerful feature that we know our customers will love, but because it’s essentially the last piece of the puzzle that completes what we set out to achieve when we started Vervoe four years ago.

From the beginning, our vision for Vervoe was to create a platform that could accurately predict job performance. To remove the false signals and unfairness associated with the current process of hiring, relying on resumes and interviews; and instead make hiring decisions based on merit. By allowing every candidate to prove how well they could do the job and allowing employers to see what the candidate would do if hired, we knew we could make a huge difference.

Making hiring about merit, not background, was (and still is) an ambitious goal. To achieve this goal, we needed to build out our platform in stages, in a particular order to maximize Vervoe’s usefulness and value.

Vervoe's product journey milestones
Vervoe’s Product Timeline

Below is the journey we took to get to where we are now.

Vervoe launches it’s core functionality

In Dec 2016, we built the first version of Vervoe, which focused on our core functionality, creating and completing assessments. This version was the first iteration of what we call Create today. 

The objective was to provide a flexible and powerful content authoring tool to simulate the job candidates are applying for. To truly predict job performance, you had to start by putting every candidate into realistic scenarios that they would encounter on the job. 

We included everything from embedded spreadsheet functionality for accounting and management roles, code testing for developers, presentations for sales roles, audio for customer service, etc. We knew that if you really wanted to predict performance, there was no point asking multiple-choice questions or getting candidates to play a game.

We built a clean and easy to use interface where candidates could complete the assessments in a controlled and consistent manner. Focused on the candidate journey from the outset, we understood that candidates had no idea what Vervoe was. Here they were being sent into an assessment by a potential employer. We did everything we could to help the employer create an environment that looked and felt like their brand, including having the interface employer branded and not requiring candidates to register or log in.

AI-grading of candidate assessments added

In July 2018, we released our industry-changing AI-grading.  From the beginning, having every candidate complete a skills assessment (or “auditions,” as we used to call them) presented one obvious problem. They required a lot of work by the employer to review and manually grade every single assessment. From the start, we intended to use machine learning to grade the assessments automatically. To do so, we needed a large and clean data set to train our grading models, which we didn’t have initially. 

Because the first version required every candidate response to be graded (out of 10), it provided us with a super clean data set to train our models. A data set built on real data. We only had employers who paid to use the product to hire real candidates for real jobs. 

By the time we launched our second version, we had built our data set using over a million manually graded questions. We had now formed a solid foundation for our extremely accurate automated grading.

A look inside vervoe’s product evolution 1
Vervoe’s UI from first product launch to present.

Machine learning content creation

Next, we focused our attention on content. 

We always had an assessment library, but for the same reasons that no two jobs are the same, it doesn’t make sense to have just one assessment for a role. For example, say you wanted to hire a sales manager.  Would you use the same assessment for a sales manager at an ice cream shop and a sales manager for a software company? Of course not. Some skills and questions may overlap; however, the overall assessment would be different.

Instead of merely adding more assessments, we decided to create an interface where users could start with our assessments as a base and then easily add/remove questions to suit their unique requirements.  After all, our users are the only ones who know what skills they want a successful candidate to have.

We created a powerful question suggestion feature with over 100,000 questions. The suggestion engine, driven by machine learning,  suggests questions we think are relevant based on what other users have successfully used for similar roles. This feature has been hugely popular as it gives our users confidence they are asking the right questions in their assessments while also giving them a choice to create an assessment bespoke to their needs.

With this release, in January 2020, customers now had multiple ways to create an assessment, building one from scratch  (with embedded docs, spreadsheets, coding, video, audio, etc), choosing one from the assessment library, or mixing and matching to get the perfect assessment.

Ability to select the importance of specific questions and skills

Listen to Vervoe Co-Founder and Chief Product Officer, David Weinberg talk about the exciting new feature, Select.

Our entire team is extremely proud of the product we have created. It’s a product that has helped companies hire thousands of deserving, successful employees and allowed millions of candidates to put their best foot forward. To be noticed and evaluated in a fair and unbiased way.

Until now, the “output” of Vervoe was a list of candidates, automatically ranked based on our machine learning, from best to worst. We have made this ranking as accurate as possible,  and every time our customers use Vervoe, the list becomes more accurate. However, we had always strived to achieve the next level.

At Vervoe, we know no algorithm or test can tell you exactly who you should hire. Nobody knows the best person to hire better than you. Our objective was always to provide our customers with deep insights to give them the confidence to make the best, most informed decision based on accurate, unbiased data.

That’s why this release is a game-changer for our users. Select goes beyond a prioritized list to help you decide who to actually hire. You can run different scenarios by excluding specific skills or questions and see the list instantly change. If you think question 2 is no longer relevant – excluded it. Care more about “communication skills,” and less about “javascript” remove the whole Javascript skill and have all candidates instantly re-ranked. You can also compare candidate scores and responses side-by-side, and share shortlists, including full report cards, easily with stakeholders. Once you’ve made your decision, mark the winning candidates as “hired”  so their data is easily accessible to track them with post-hire analytics.

This release is about making it easier for you to make the right decision. It’s about leveraging everything we have built over the last four years, all the powerful predictive features, to deliver a simple interface that gives our users the confidence to make the best hiring decision. That’s why we call it Select. 

Don’t worry; this certainly doesn’t mean our product backlog is empty. We have a massive roadmap of features and releases that will continue to make Vervoe better and better. However, from a product workflow perspective, this was the missing piece that we have waited so long to establish. 

Our team has been working super hard on this release and we can’t wait for you all to try it.

Related articles

Inside the Product
Anatolii Kryvko

Introducing Active Learning for Vervoe

Right out-of-the-box, our AI grades your candidates automatically with 80% accuracy. But we also recognize while building Vervoe that not