When we started building Drafted in 2024, the core product question was deceptively simple: what does it mean for a company to have good network coverage for a specific open role? Not network size in the abstract, but the right coverage for the right hire at a particular moment.
We spent a lot of the early months getting this definition wrong before arriving at something we think is actually useful. This post is our current answer, with the caveats that come from still building and learning.
The first wrong answer: raw connection count
The obvious first approach is to count connections. If your twenty employees collectively have 40,000 LinkedIn connections, you have 40,000 potential candidates in your extended network. That is a lot. You should be able to find someone for any role in there.
In practice, this turns out to be nearly meaningless for hiring. Most connections in a large LinkedIn network are shallow: a conference, a cold connection request, a brief professional overlap that did not result in any ongoing relationship. When you ask an employee to reach out to a 2nd-degree connection they barely know, through a 1st-degree connection they barely know, you have constructed a path that looks warm on paper and is cold in practice. The candidate will experience it as outreach from a stranger.
We learned this early when we looked at which introduction attempts in our beta testing actually resulted in a candidate taking a call. The correlation between connection count and successful introduction was weak. The correlation between relational depth indicators and successful introduction was strong. So we stopped thinking about coverage in terms of how many nodes and started thinking about how many real connections.
The second wrong answer: role-agnostic coverage
The next instinct was to define coverage as a company-wide property. This company has good network coverage. This company has gaps. But coverage is not a property of a company. It is a property of a company relative to a specific role type.
A twenty-person company with a founding team that spent years in enterprise SaaS sales and account management might have excellent coverage for sales and customer success hiring and almost no coverage for engineering roles. Telling that company they have "high network coverage" is actively misleading. For their next senior engineer hire, they are starting from scratch. For their next account executive hire, they have warm paths to dozens of qualified candidates through their team's existing relationships.
Coverage needs to be defined per role domain. What looks like a well-connected company can have blind spots that only surface when you try to hire in a specific area.
What we actually measure now
Our current coverage model evaluates network quality against a role profile on four factors:
Connector match: Do any of the company's employees have professional history that puts them in close network proximity to candidates with the right background? We look at employer overlap, functional area alignment, and technical domain overlap between an employee's career history and the target profile.
Connection depth: For the employees who have connector match, how strong are the individual connections? We score this on recency, mutual connection density, and shared employment tenure. A direct connection to someone who worked at the same small company for three years scores very differently from a connection made at a networking event last year.
Path reach: Beyond direct connections, are there strong 2nd-degree paths? A warm 2nd-degree path, where the 1st-degree connector has genuine depth of relationship with both the company employee and the target candidate, is often nearly as effective as a direct introduction. A weak 2nd-degree path through a superficial connection is not warm at all.
Coverage density: For a given role, how many distinct warm paths exist? Having a single strong path is useful. Having five independent strong paths means you have real options, and that one connector being unavailable or declining to help does not block you entirely.
What good coverage looks like in practice
For a given open role, good coverage means: at least two employees have genuine relational depth with candidates who match the target profile, there are at least three distinct warm paths to qualified candidates, and at least one of those paths runs through a direct connection rather than requiring a 2nd-degree hop.
For a company our typical early-access customer size, around forty to one hundred fifty employees, strong coverage on a given role type usually comes from having two to four employees who spent significant time in that professional domain. One is enough to start. Two gives you redundancy. Three or more means that role type is a strength of your network.
Coverage gaps are equally important to know. If a role type has zero connector match, that is your signal to plan for cold sourcing for that role, work the agency or recruiter route, or factor network coverage into how you think about the next hire in that domain. Some of our customers have explicitly made "strengthens a coverage gap" a secondary criterion for certain hires, knowing that the person they bring on will enable the next five hires through their own network.
The part we are still working on
Coverage tells you whether the paths exist. What it cannot fully tell you is whether those paths will convert. An employee with strong connections in a domain still has to be willing to make the introduction, and the candidate still has to be open to the specific role at the specific time. Those factors are softer and harder to model.
We have found that the quality of the introduction request made to the employee matters a lot here. Vague asks ("do you know anyone in ML?") convert poorly even when the underlying network coverage is strong. Specific asks ("we have a principal engineer role focused on recommendation systems, and you know someone who has done exactly that kind of work") convert much better. Getting the specificity right is an operational discipline, not just a product feature, and it is something we are still learning how to support well.
The other open question is temporal. Network coverage changes as employees leave and join, as professional relationships warm and cool, and as role requirements evolve. A static coverage analysis is accurate for perhaps a quarter before it starts to drift. We re-run the analysis on a rolling basis, but the question of how to surface coverage changes to recruiting teams in a useful, actionable way rather than a noise-generating way is ongoing product work.
Our best current answer to "what is good network coverage?" is: enough real connections to qualified candidates in a specific domain that at least one warm introduction is plausible at any given moment, with enough density that you have options rather than a single point of failure. Everything else is optimization around that baseline.