The Off-List Reference List: Who to Call Before You Wire

You have a term sheet out on a founder, and the reference list they handed you is the list they wanted you to have. This guide is for early-stage investors, platform and talent partners, and angels who need to reconstruct that founder's real working history and produce their own call list before wiring. It carries one anonymized founder all the way through: the actual searches, the counts each returns, the forks, and the wrong turns.

The deliverable is concrete. In a single sitting you will turn one name into a ranked list of 8 to 15 off-list references - prior co-founders, early employees, and direct reports - each with a one-line reason it was chosen and a route to reach it. Existing guides give you question scripts for the call. None show the search work of building the list without the founder's help. That is the gap this fills.

Why the founder's list is the wrong place to start

The founder's reference list is a curated artifact, and the most useful sources are the ones it structurally omits. The disciplined move is to stop treating that list as complete and reconstruct the founder's employment history yourself, then pull each prior company's alumni and co-founders.

The reason is not paranoia, it is where the signal lives. As one framing puts it, the best investor diligence happens off-list, founder to founder, where people are willing to explain how someone behaved during a missed quarter, a co-founder issue, or a down-round conversation. Those people almost never appear on a list a founder assembles, because a founder assembles a list of people who will be kind. A warm, glowing peer who surfaces on their own is often a personal ally, not an independent witness.

Reference work sits late in the deal. For the sell side, the right time for reference calls is after receiving a term sheet and before signing it. On the buy side, backchannel calls typically sit after the partner meeting but before the term sheet, and formal calls often run 10 to 15 in the first week. There is an order dispute worth knowing: some firms start backchannel work earlier, during the first or second meeting, before the founder knows. Either way, the search work below is the same.

The worked example: one founder, one sitting

Take an anonymized founder, call her the subject. Her public profile shows a current startup at the seed stage and two prior companies. The whole point is that her handed-over list names three people, all from her most recent and most successful role. Nobody from the earlier venture that quietly wound down. That absence is the tell.

Reference volume benchmarks give the target size of the finished list. Documented practice clusters between 10 and 25 calls per deal.

Aim for a ranked list of 8 to 15 names. That gives enough to complete the highest-signal calls in a first-week sprint even when a third decline or go unreachable. Note the honest limit up front: firms run these volumes, but a direct published correlation between the number of back references run and fund returns is not established. Outcomes are dominated by founder traits, where a meta-study found business performance correlating with founder personality at multiple R around .31. References work as a downside and fraud filter, not an alpha generator. Run them to avoid a bad wire, not to find a great one.

Step by step: name to ranked call list

The procedure is one sitting of roughly two to three hours. Each step has a clear "done" state so you never wonder whether to move on.

Step 1 in practice: the spine

The subject's timeline resolves to three windows. Her earliest role was a mid-size company for two years, then the venture that wound down where she was a co-founder for three years, then the current startup. The wound-down venture is the one her list ignores entirely, which makes it the priority target. Write the timeline with months, not years, because month-level dates are the only way step 3 works.

Step 2 in practice: the pulls and their counts

Run the past-company filter on each employer. The mid-size early company returns a raw count well over 1,000, which means it is silently truncated. This is the first fork. Rather than accept a truncated list, batch by title and location so each sub-query stays under the ceiling and the counts add up.

The wound-down venture is small and returns 47 alumni total, no truncation. That is the list that matters most, and it is small enough to read by hand. The wrong turn here is spending an hour perfecting the batching on the 1,000-plus employer before touching the 47. Do the small, high-signal venture first.

Reading the pool sizes so absence does not fool you

A sparse alumni trail is usually a base-rate artifact, not evidence of a weak network. Founding titles are rare and inconsistently self-applied, while "co-founder" is a durable label people keep for years.

The numbers make this concrete. In Refolk's index of professional profiles, the US pool of people titled "Founding Engineer" or "Founding Team Member" is 4,520, while the "Co-Founder" pool is 194,582.

Co-founder alumni are 43 times more abundant than founding-engineer alumni. So if the subject's wound-down venture surfaces two co-founders but almost no one who self-titles "founding engineer," that is exactly what the base rate predicts. Do not read it as a hollow team.

Geography narrows the pull sharply. Early-stage talent clusters in a metro, so a location filter is one of the fastest cuts you can make.

The US pool is 7.3 times the UK pool, and within each, one metro dominates. In the UK sample, London held an indicative 13 of 25 profiles; in the US, San Francisco led. For the subject, whose earlier venture was US-based, adding "San Francisco or New York" to the alumni query cut the reachable-and-relevant list by more than half in one move.

This is where a plain-English search removes the friction the last three steps described. Reconstructing a spine, pulling alumni, and filtering to tenure overlap is mechanical work across public LinkedIn and Crunchbase records and the public web. Refolk collapses it into one ask: name the prior company, the years, and the metros, and it returns the people who overlapped rather than everyone who ever held a badge. You still make the ranking calls, but the list assembly stops eating the sitting.

Bucketing and ranking: what each source proves

Sort every surviving name into the five documented reference categories, then rank so the frankest sources rise. Character references, customer calls, former colleague interviews, co-investor checks, and backchannel conversations each reveal a different facet.

Ranking is where judgment enters. Prioritize direct reports and colleagues from ventures that failed or struggled over people from famous wins. A direct report at the wound-down venture saw the missed quarter and the co-founder friction from the inside. A peer from a celebrated exit mostly saw the win, and success papers over a lot.

For the subject, the top of the ranked list becomes: her co-founder from the wound-down venture, two direct reports from that same venture, an early customer who churned, and only then the peers from the successful role her own list already covered. Eleven names in total, each with a one-line reason.

The backchannel is where deals die quietly, so the reference a founder omits is the one worth the call.

Finding reachable paths and clearing consent

For each ranked name, resolve a route to reach them, preferring a warm mutual connection over cold outreach. Then run a consent and exposure pass before you dial, because the way you reach a source can itself do harm.

The matrix below is the judgment call on each name: how strong the signal is against how exposed the outreach is.

The exposure risk is asymmetric. Reaching a source at their current employer signals to that employer the person may be leaving, which is both a privacy problem and a reputational one. Under GDPR the practical legal basis for a reference is consent, with legitimate interest as an alternative only if questions stay strictly job-relevant and respect reasonable privacy expectations. Favoring past-company alumni is therefore not just cleaner data, it is lower legal and reputational exposure.

Subject: Quick question about your time at [prior venture]

Hi [name], I saw you overlapped with [founder] at [prior venture] between [years]. I'm doing background work on a professional matter and would value ten minutes on what they were like to work with day to day. Nothing confidential, and I'm happy to keep it entirely off the record. Would a short call this week work? Thanks either way.

How this goes wrong

The failure modes here are specific and most of them produce a false positive that reads as fact. Treat this section as the checklist you run against your own work before trusting the list.

• Tenure-overlap error. Pulling all alumni without checking dates yields people who never met the founder. The false positive is a "colleague" who joined after the founder left. Compare month-level dates on both profiles before keeping a name.
• The 1,000-result cliff. A large prior employer truncates silently, so you think the list is complete when it is missing hundreds. Batch by title and location and confirm the counts add up.
• Curated-list contamination. Names that surface on their own may be the founder's friends. A glowing peer who is a personal ally reads as an independent witness. Weight failed-venture and direct-report sources higher and treat uniform praise as low signal.
• Negative-data over-weighting. A lukewarm former-employer call gets read as a founder flaw when it reflects a side-hustle dynamic. Many founders did not love their corporate jobs and were hustling on the side, and those answers are often not predictive. Separate personal conflict from professional capability before scoring.
• Absence read as guilt. A thin network gets treated as a red flag when the referral was the reference, because early hires often came through the founder's own network or board. Ask how the early hires arrived before concluding anything.
• Consent and tip-off exposure. Contacting a current employer of a still-employed source outs them and creates privacy exposure. Flag current-employer overlaps and keep questions to job-relevant facts.
• Presence bias in synthesis. You write up who you reached and skip who you could not, missing the most telling gap. Presence bias is the systematic tendency to reason from what is present more readily than from what is absent, and it stays dormant until you deliberately trigger it.

Verify before you call

Run this checklist against the finished list. It catches the failure modes above before they reach a phone call.

Keeping the list honest as the deal moves

Off-list reference work is not a one-time artifact; it moves with the deal, so timestamp your list and re-check it if the round stalls. Diligence runs 2 to 6 weeks - seed rounds 2 to 3, Series A and later 4 to 6 - and a founder's public trail can shift within that window as people change jobs and old ventures get restructured on Crunchbase.

Two things go stale fastest. First, a source's current employer, which changes the exposure calculation from the matrix above; re-check it the day before you call. Second, newly surfaced alumni from a truncated employer, which appear as people update or clean up their profiles. When a deal drags past its expected envelope, re-run the alumni pull on the highest-signal venture once more, because the 47-person list that was complete three weeks ago may now have a name that was not there before.

Budget the effort honestly. A VC can count on spending at least 20 hours of due diligence per deal, and the reference sprint is a slice of that, not the whole thing. Do the list assembly in one sitting, spread the 8 to 15 calls across the first week, and keep the absence line updated as sources come in. The list is done when you can answer, for every name you did not reach, why - and when the names you did reach include the ones the founder never wanted you to find.