How can The Cole Group mitigate inefficiencies in the executive recruitment process?

Example of pipeline view for a given job/candidate search.

Context and Background

The Cole Group is an executive recruiting firm that places CMOs and CROs at high-growth, VC-backed companies. I was hired to build what was essentially an internal CRM—which we called “Rolodex” internally—plus some additional features. Building it began with an extensive discovery process and was heavy on data and process management. Using Rolodex, recruiters were able to close searches 13% faster and take on 25% more concurrent searches. However, it originally started with a relatively simple problem and a narrower objective.

When I was approached about a role, there was one particular challenge that the founders initially wanted to address: clients were slow to make a hire. Few CEOs have been in the position to make such an important hire before… understandably, they want to make the “right” decision.

We used house-hunting as an analogy:

Purchasing a home is one of the biggest decisions most people will make. As a result, few feel comfortable making an offer on the first or second home they see—especially first-time buyers. They want to “get a feel for the market” before deciding.

For homebuyers, though, new tools like Redfin and Zillow offer access to far more data than ever before: price estimates, detailed descriptions, rooms, square footage, walk scores, school districts, etc. As a result, homebuyers can understand the market—and the realistic options available—better and faster. That, in turn, instills the confidence to make an offer far more quickly.

In short, one of my first objectives was to help build a tool that could similarly instill a sense of market understanding and confidence in our clients.

Research & Discovery

Task flow identifying core recruiting process from start to finish,
highlighting points of friction

After discussions with the founders, I started by interviewing all the recruiters, executive assistants, and accounting. I went to client meetings and went through historical Salesforce data. After identifying common patterns, I created a task flow that mapped out the process and challenges from start to finish. This went through several iterations of feedback from the founders and recruiters.

Example of a “job spec,” where recruiters indicate which criteria are being sought in terms of ideal candidates. The search results page (SRP) enables further filtering and segmentation.

Cohort Analysis Tool

In order to address the issue of time-to-conviction, the first feature we focused on was a cohort analysis tool. The goal was to communicate to customers, “here’s how companies similar to yours have hired,” with ‘similar’ being defined primarily by industry and similar revenue at time of hire, plus the specifics of the job spec.

For example, say you’re a FinTech company doing $50MM ARR looking for a CMO with experience managing a large budget and affiliate and search marketing. Based on the job spec—the desired or required attributes in the hiring profile—the cohort analysis tool would show us the distribution of candidates that were hired and how many of the criteria they had, on average. For example, you might find that out of the 58 hires made, on average they met 7.4 out of 13 criteria.

Once built, this tool would eventually help clients understand details like whether hires were more typically junior people coming from larger companies and stepping into a more senior role, or more senior but from smaller companies.

Example of a high-level cohort analysis, which shows how similar companies have hired (matched by criteria like size, industry, business model, growth trajectory).

Foundational Data Need

Before we could create this kind of cohort analysis tool, we obviously needed quite a lot of company and candidate data. And we needed some of the core functionality of a CRM. (Recruiters used Salesforce until we built Rolodex.)

Internally, we had twenty years of prior placement data, which was a great start (although it wasn’t organized or structured, unfortunately!). We also had a lot of data from Venture Capital partners, since hiring talent for portfolio companies is a big part of VC world and we had a long history of placing candidates at those companies. We also used Mattermark’s API and combed through public sources for additional details of executive hiring.

We worked with mostly objective data, like the size, vertical, or GTM approach of the hiring company, since subjective data is extremely difficult to apply at scale or be consistent about in terms of methodology. We even created formulas to build on that data, like estimating seniority based on normalized titles and company type/size. We developed a remarkably accurate revenue estimation algorithm as well, which required very few data points, to estimate size of companies we didn’t know (and we knew a lot).

We also had to develop and test hypotheses around what made for—and even what constituted—a “good hire.”

Prioritization & Roadmap

The challenges identified in the discovery process helped to inform our product goals. Aside from the cohort analysis tool, we knew there was significant value in leveraging the foundational data for the search process. I segmented the goals into product, process, and research (mostly dependencies) goals. I then prioritize them, comparing the value to Cole and the level of difficulty/effort, and of course getting feedback from the founders and recruiters.

Once prioritized, the goals were put into a high-level roadmap. Roadmap conversations took place in a simple Google sheet:

Sorry for the blur, some things can’t be shared!

Individual tasks related to each goal were scored (in an attempt) to remove bias and subjectivity from more granular prioritization. (these frameworks are very helpful in deciding what to approach first, determining criteria, and thinking through dependencies.) All items were then split into PRDs in Confluence and put into epics/tickets in Jira.

Execution

In terms of execution, I collaborated with our designer using Sketch and then Figma—working within an existing but somewhat loose design system—and worked with our engineering team in two-week sprints, at the end of which we had demos and then released new features. Some features would make it into production sooner, if urgent.

We worked on an extremely diverse set of features beyond cohort analysis, including:

Candidate profiles, including assessment pages
Company profiles
Search & results pages
Candidate reference forms
Email automation/drip campaign management
Pipeline overview, lead funnel/conversion tracking
Analytics

Integrations

We used a number of third-party integrations like RocketReach and even facial recognition (while it’s a touchy subject, a number of companies wanted to be sure to include “diverse” candidates, and FR helped code women and minority candidates—it had a 99% accuracy rate, although we were able to manually update assessments). And we had data, processes, and integrations to “auto-assess” as much as possible, to minimize the manual assessment otherwise required.

Candidate profiles. We incorporated a certain amount of “auto-assessment” based on the companies they worked for and the role they were in, for example.

Results

The investment paid off.

Most importantly, revenue increased 16%.

Recruiters were closing searches 13% faster, could take on 20-25% more concurrent searches, and already-high customer satisfaction was 7% higher.

Recruiters also rolled completely off of Salesforce, since we eventually baked the features they needed into Rolodex (besides, the SFDC lightning integration had occasional issues, which were nice to resolve). The cost savings there alone was in the six figures.