Candidates choose their own references. Of course they pick people who will say nice things. That’s not a flaw in the system — it’s a feature of human nature. But it means listed references are systematically biased toward positive feedback.
The backchannel reference graph is our way of finding signal outside the candidate’s curated list.
Before we explain how it works, a note on ethics: we take privacy seriously. The backchannel graph surfaces potential connections. It doesn’t contact anyone automatically. It doesn’t scrape data. Every outreach decision is made by a human. We built this feature to be useful without being creepy, and the line between those two things is one we think about constantly.
The problem with listed references
Listed references are a curated sample. The candidate has pre-screened these people. They’ve probably had a conversation: “Hey, you might get a call. Here’s what the role is about.” The reference knows they’re being evaluated, and they know they’re expected to be positive.
This doesn’t make listed references useless. A good interviewer can still extract real signal from a listed reference. But it does mean the information is filtered. You’re getting the highlight reel, not the behind-the-scenes footage.
The studies on this are clear: listed references have almost zero predictive validity for job performance. They’re better than nothing, but not by much. The question is whether there’s a way to get unfiltered signal without violating trust.
How it works
Rob maps the LinkedIn connections of your existing team and cross-references them with the candidate’s professional network. If someone on your team worked at the same company during the same period as the candidate, that’s a potential backchannel reference.
Rob surfaces these connections and suggests reaching out. The hiring manager decides whether to make the call. It’s opt-in, not automatic — we believe in transparency about the process.
The graph is more sophisticated than a simple overlap check. Rob considers the department, the seniority level, and the duration of overlap. A two-year overlap in the same engineering team is a much stronger connection than a one-month overlap in different departments. Rob ranks the potential references by relevance and proximity.
Each potential reference comes with context: “Your engineer Alex worked at Stripe from 2021–2023, overlapping with the candidate for 18 months. They were both on the infrastructure team.” This gives the hiring manager enough information to decide whether the outreach would be valuable.
Why backchannel references matter
Backchannel references are more honest because the candidate didn’t select them. They’re also more relevant because they come from people your team already trusts.
The best backchannel conversations happen when the reference doesn’t know it’s a reference call. A casual “Hey, I see you worked with Sarah at Stripe. What was she like?” over coffee surfaces information that a formal reference call never would.
In our data, backchannel references have significantly higher predictive validity than listed references. Hiring decisions informed by at least one backchannel reference have a 23% higher 12-month retention rate. The unfiltered signal makes a real difference.
The ethics of backchannel references
Backchannel references exist in a gray area. On one hand, people have always informally asked around about candidates. That’s just networking. On the other hand, systematizing it with software feels different. It raises legitimate questions about consent, privacy, and power dynamics.
Our approach is opt-in at every level. Rob surfaces the connections but never reaches out. The hiring manager decides whether to act on them. And we recommend that hiring managers inform the candidate that they’ll be reaching out to mutual connections. Transparency builds trust.
We’ve also built in safeguards. Rob won’t surface connections through the candidate’s current employer, because that could jeopardize their existing position. It won’t surface connections that might reveal protected information. And it always reminds the hiring manager that backchannel references should supplement, not replace, the formal reference process.
What we’ve learned
After a year of the backchannel graph being live, here’s what we’ve learned: about 60% of roles have at least one potential backchannel reference. Of those, hiring managers follow up on about 40%. And of the follow-ups, about 75% produce information that influenced the hiring decision.
The most common outcome is confirmation: the backchannel reference validates what the listed references and interviews already showed. That’s valuable because it increases confidence. The second most common outcome is nuance: the backchannel reference adds context that changes how the hiring manager interprets other signals.
The rarest but most impactful outcome is a red flag that nobody else surfaced. This happens in about 5% of backchannel conversations, and in every case, the hiring manager told us the information was critical. Without the graph, they would have made a different — and worse — decision.