Identifying promising companies while searching

Sales reps in ZoomInfo Sales spend a lot of time searching for companies they want to prospect and researching whether those accounts are worth contacting. This project focused on that search workflow.

The product already generated AI signals that could help with that decision, such as signs a company might be showing interest or changes that made an account newly relevant. But most of those signals lived on a separate Copilot homepage instead of inside the search experience reps already used.

The goal was to surface those signals during search, so users could notice promising opportunities while working the way they already worked.

Some time after launch, adoption of Copilot signals was lower than expected. Analytics showed why: users were still following their old habits. They opened Quick Search or Advanced Search to prospect manually, and skipped the Copilot homepage where most signals lived.

Quick Search let users look up a company or contact directly from the nav bar. Advanced Search was a separate page with more filters for researching and narrowing down accounts. Both were already part of users' daily workflow, which made them the obvious place to surface signals.

I ran qualitative research to dig deeper, focusing on Copilot users who used Quick Search and Advanced Search without engaging with signals. The goal was to understand why they weren't utilizing Copilot and to explore how signals could improve their search process.

What I found matched what the data suggested. Users mainly said they don't have time to learn a new interface, that they'd tried the Copilot homepage briefly and hadn't seen immediate value, or that the companies it surfaced weren't the ones they were actively working on right now.

Given these findings, this project aimed to meet users where they work: surface Copilot inside Quick Search and Advanced Search, assuming they would be more valuable while searching due to their relevance to the companies users were searching. As a side effect, we hypothesized this would make the homepage's value obvious from search so more users would eventually try it on its own.

I gathered a few designers into a Crazy 8 brainstorm to generate some quick pre-concept ideas. I presented the problem, the research, and the project's goals. We ran it twice: once for Quick Search and once for Advanced Search. Each time, each of us sketched eight ideas in eight minutes. We presented the strongest sketches and scored them together, and I used the best ones as a basis for my work below.

After opening Quick Search and before typing to search, we showed users their top signals for their target accounts, using the right side of the Quick Search initial state as a promotional space and an entry point for relevant AI signals. While searching, the same promotional space stayed on the right side of Quick Search for users who may have missed it due to typing quickly, at which point the initial state is replaced with results. We hypothesized that a focused set of the top, most relevant AI signals would entice users to click through to find out more.

For results, we showed signals for the first result only, since it's the most likely to match what the user typed, and showing signals for other results will likely be irrelevant for the user, who is obviously interested in the specific company they are searching for. If the top result was not what the user was searching for, signals were revealed for other results on hover. Whether users clicked on a result with signals, or on a company in the promotional space on the right, they would be redirected to the company's profile page as usual, but with the AI signals section in the profile page (which already existed before this project) briefly highlighted to draw attention to it.

In Advanced Search, we added a Signals column to the results table showing the number of signals per company. Here too, we hypothesized that users would be more inclined to research companies with AI signals. Clicking a company row already opened a side panel with its information, so we made it so clicking a signal count would open the same side panel, but highlight the signals section.

Copilot signal adoption increased post-release. When we investigated where exactly, we found it came from Quick Search. We assumed that the new Signals column got lost among the other columns in Advanced Search. Later teams picked this up for further improvements.