Aquent’s Book is a powerful portfolio search engine developed by Aquent’s internal product and staffing technology teams. As the product designer, I worked with product managers, engineers, stakeholders, and staff to conceptualize what the search experience might entail if we were to combine machine learning, broad crawling, and Aquent’s internal database to deliver a single source for talent discovery and pooling.
First, the technology: Aquent’s Book auto-tags portfolios using image classification and object detection. It tags work through the application of more than 24 million object and image labels. Using natural language processing, it combines image search with full-text search, giving equally rich data on creative talent both inside the database and across the web. In less than a second — 500 milliseconds to be exact — it sifts through millions of portfolios, finds the precise type of work you’re looking for, and then lets you see the breadth of the talents’ creative capabilities.
We were tasked with creating solution concepts for how Aquent’s Book’s front-end search experience might evolve to better suit agents’ workflow. I’m sharing with you a successful concept that led to the direction eventually implemented.
Our discovery process revolved around agent interviews and third party feature auditing. We were looking for answers to questions such as:
- What is the the agent’s typical journey for discovering talent? Ideal journey?
- How are the current search results inside our platform failing to meet expectations of agents?
- What signals exist in other platforms that indicate value?
- Do we need to buy or build?
- Feature add or feature iteration?
- What does the ideal talent candidate look like? What information makes this true?
After synthesizing the feedback, a strong pattern emerged. Agents utilize numerous tools for sourcing talent, some of which are found within Aquent’s internal software while others are publicly or privately accessible from third parties (i.e. LinkedIn, ZipRecruiter, Behance, etc.). Each delivers a slice of value, but none deliver the entire pie. The result of using multiple third party destinations in an agent’s workflow?
- Increase in time and effort spent.
- Increase in cost of access to information.
- Increase in communication channels (and info lost in translation).
- Decrease in usable data stored in the database.
- Fragmented sourcing processes across teams.
While agents do use internal software, they also rely heavily on external tools for finding:
- Status information (i.e. location, availability, compensation, etc.)
- Work history information (i.e. years experience, industry experience, etc.)
- Stylistic information (i.e. portfolio imagery)
Our hypothesis was that if we could create our own single destination tool that delivered results combining these types of information, we could impact two of the gold KPIs tied to agents: fill rate and time to fill.
While engineering built the crawlers and ML models needed to aggregate and learn from said information, we began to explore how that it would be delivered to the end user. I conducted a series of design jams with engineers to work through multiple iterations of user flows and UI compositions. We eventually arrived at a concept that garnered great feedback from agents.
It’s a module (to exist set inside Aquent’s platform) consisting of a controls panel and a results list. The controls panel is comprised of the highest value ranked data. Ranking was determined by a combination of agent surveys and by tracking usage of inputs in the previous search experience that led to successful momentum in sourcing. Results are tabulated to differentiate talent found in the database and talent found on the web.
Users are presented results crawled from the web and surfaced from the database. Users can add crawled talent to the database, which effectively sets in motion a process to contact the individual talent and gather more information (conversion).
The results list is comprised of talent cards, each filled with relevant status, work history, and portfolio information. Selecting images on a database result provides a preview of the talents’ work, including descriptive captions and ICLs used in calculating the result. Selecting images on a web result will navigate the user to the external URL the image resides at.
Much of this concept became the basis for what would eventually be version 2.0 of Aquent’s Book. We’ve successfully deployed a portfolio search engine that more accurately delivers high value content to agents to enable them to make faster, more insightful sourcing decisions, as well as discover new talent and seamlessly add them to Aquent’s database. These are processes that previously required multiple tools, and were manual and laborious in nature. While the tool has not yet proven to reduce time to fill, it has reinvigorated agents’ trust in Aquent’s internal software, effectively motivating them to not have to rely on third party platforms to achieve a desired result. This, in turn, has given us the ability to accrue data that we can build models. This will aid in our design decisions in a plethora of in-flight talent initiatives in 2021.