Founder Reference Calls: From First List to Go or No-Go

Deciding to back a founder is the highest-stakes judgement an early-stage investor makes, and reference calls are where most of the real evidence lives. This guide is the end-to-end program you keep open while working the phones: how to assemble the list, in what order to dial, what to ask each type of source, and how to convert a set of guarded conversations into a go or no-go you can defend to your partners. It is written for partners, platform and talent partners, and angels who need a repeatable process, not a one-off teardown.

The core discipline is simple to state and hard to execute. The founder's own list is positive-biased by construction, so the check that matters is the one you source yourself. And because references lean positive, the signal is in the delivery - the pause, the hedge, the volunteered story - as much as in the words.

How many reference calls does one deal need?

A full founder reference program runs roughly 10 to 20 calls at top-tier firms and 5 to 7 at early stage, weighted heavily toward people the founder did not name. There is no single correct number, but the shape is consistent: a small curated list from the founder, then a much larger set you assemble and dial yourself.

The volume ranges by firm and stage. Top-tier firms like Sequoia, Andreessen Horowitz, and Benchmark run 10 to 20+ calls per deal, targeting both named and unnamed people. One practitioner estimate puts the broad average at roughly 10 calls per closed deal, with top firms in the 15 to 25 range and most calls going to people you were not given. At early stage, CRV advises founders running the mirror process to aim for 5 to 7 conversations, including backchannel references they dig up themselves.

Against those totals, the founder-provided list is tiny. Founders tend to prep and offer 1 to 2 curated referees; some funds, such as 2048 Ventures, ask each founder for 2 to 3 professional references. The math tells the story on its own: if you run 15 calls and the founder gave you three names, four fifths of your evidence comes from sources you found.

Where reference calls sit on the deal clock

Formal reference calls sit between the partner meeting and the term sheet, with light backchannel pings sometimes starting earlier and customer calls held until the end. Placing calls on the clock matters as much as running them, because each call type carries a different cost to the founder and a different value to you.

The documented cadence runs in four phases. Before the first meeting, an investor does a light LinkedIn review and quiet backchannel pings to mutual connections. After the partner meeting, formal reference calls begin, usually 10 to 15 in the first week. Before the term sheet, final calls clear any remaining concerns. Post-term-sheet checks are rare and confirmatory, run only if something surfaces late.

Sources disagree on the earliest point. Some investors reach out to their own network before even taking a first meeting; others hold all formal work until the partner meeting is done. Resolve this by your own conviction level: quiet, low-cost pings early are fine, but the intensive program should wait until the deal is real enough to justify the founder's exposure.

Customer references are the exception to "earlier is better." Expect these late, sometimes post-term-sheet, because customer calls mainly validate rather than educate - and because calling a founder's biggest account can cost them the account.

The six source types and what each one proves

A structured founder reference program spans six source types, each of which proves something different and each of which lies in a different way. Knowing what a source can and cannot tell you is what keeps a call list from becoming a popularity contest.

• Former managers, board members, and mentors speak to integrity, resilience, and decision-making under pressure. They see the founder from above and tend to be generous, so demand specifics.
• Direct reports reveal how the founder leads and how they treat people who work for them. This is where leadership style shows without a filter.
• Former colleagues - engineers, marketers, operators who worked alongside the founder - speak to day-to-day execution and how the founder behaves as a peer.
• Co-investors confirm or contradict cap table, runway, and progress claims. They are the fact-check on the numbers.
• Customers validate product value, sales process, and retention claims. High-signal, high-cost, and therefore held for last.
• Failed-company peers - founders whose companies struggled or failed - are the highest-signal conversation available.

That last category deserves its own weight. A reference from a founder whose company struggled or failed can tell you how the person behaved when the metrics were down, when bridge financing was on the table, and when the board dynamic got tense. Success references describe a person in good conditions; the failed-peer describes them under the conditions that actually test them.

Success references describe a founder in good weather; the failed-company peer describes them in the storm.

Each source type also has a false-positive mode. A big-name former board member who barely worked with the founder gives a glowing, textureless call - the halo reference. A co-investor whose deal went badly may be unfairly bearish. Weigh the source's actual proximity to the founder, not their prestige.

Sourcing the backchannel list

Backchannel references are people you find independently through LinkedIn, portfolio networks, or mutual contacts, and they are the part of the program that separates diligence from theatre. The difficulty of sourcing them is not a constant; it is a market variable that changes how ambitious your program can be.

In Refolk's index of professional profiles, the reachable Founder or Co-Founder pool - the backchannel universe you draw from - is about 606,878 profiles in the United States, 154,325 in the United Kingdom, and 41,227 in Germany. That is roughly a 14.7x gap between the largest and smallest of these markets, and about 3.9x between the US and the UK. A 15-call program that is routine in a US ecosystem is not mechanically portable to a smaller one, where the same six categories draw from far fewer reachable people.

The practical implication: in a thin market, you widen your bucket definitions and lean harder on portfolio-network introductions, because the pool of people who once reported to a given founder or co-founded with them is small enough to enumerate by hand. In a deep market, the constraint is your time, not the supply of names.

The hard part is turning a founder's public history into named, reachable people in each of the six buckets - the former reports, the prior co-founders, the operators from a startup that closed years ago. This is exactly the sourcing problem Refolk exists to remove: you describe the people you want in plain English and get back a list to work, instead of reconstructing a founder's org chart from stale LinkedIn pages.

The reference call program, step by step

The program runs in eight stages from setting notice to writing the recommendation, and each stage has an owner, a rough duration, and a definition of done. Run the early stages in sequence and the formal calls in parallel.

What to ask, and how to read the answer

The most telling question is the binary "Would you work with this person again?", and the most important skill is reading the delivery of the answer, not just its content. References lean positive, so a diplomatic "yes" tells you almost nothing while a pause before that "yes" tells you a great deal.

Build every call around a small set of forcing questions. The single most telling is "Would you work with this person again?" A money-on-the-line variant sharpens it further: "If you had $25,000 of your own money, would you invest it in this company?" Both force a binary that cuts through diplomatic vagueness. Add comparative rankings - is this founder in the top 1%, 5%, 10%, or 25% of people you have worked with? - to convert a vague compliment into a placed bet. And ask the two questions that surface texture: "How does this founder handle bad news or conflict?" and "What is their biggest weakness, and have they improved?"

Then read the delivery. Any hedged, qualified, or carefully worded answer is a functionally negative signal, not a neutral one. A pause before answering carries more weight than it might seem. The strongest positive signal is a reference who proactively volunteers specific stories without prompting; uniformly enthusiastic references with no texture may indicate coached conversations.

1. Context: How did you work with [founder], and for how long?
2. Execution: What did they own, and what actually shipped?
3. Under pressure: How do they handle bad news or conflict? Give me an example.
4. Weakness: What is their biggest weakness, and have they improved on it?
5. Ranking: Top 1%, 5%, 10%, or 25% of people you have worked with - which, and why?
6. Binary: Would you work with this person again?
7. Money on the line: Would you put $25,000 of your own into this company?
8. Open: Is there anything I should have asked but did not?

Notes column: log delivery - pauses, hedges, and volunteered stories.

How this goes wrong

The reference program fails in predictable ways, and most failures share one root cause: treating a single call as a verdict instead of treating patterns across many calls as the verdict. A16z frames the whole exercise as triangulation for exactly this reason. Below are the failure modes to guard against and the check that catches each.

• Curated-only list. Looks thorough, is structurally positive-biased. Every investor gets a list of founders to call, and those conversations are useful only as a baseline. Check: did you source at least one name the founder did not give you, and at least one failed-company peer?
• Halo reference. A big name who barely worked with the founder gives a glowing, textureless call. Check: enthusiasm with no specific story is a coaching or halo signal, not a positive.
• Single-source negative. A bitter ex-co-founder or a burned prior investor blames the founder for a failure with many causes. Roughly 50% of startups experience a founder breakup, and many founders who do not love each other are still highly capable. Check: is the same concern named by two or more independent sources before it counts against the deal?
• The polite "yes." A hedged or delayed answer to "would you back them again" reads as approval but is functionally negative. Check: log the pause and the caveats, then run a third call on that theme.
• Customer-call blowback. Calling the founder's biggest customer signals distress and can cost the founder the account, not just the deal. Check: hold customer references until intent is high.
• Network-mirror bias. Preferential attachment - the Matthew Effect - means success accrues to those already in successful networks, so a glowing reference may reflect the founder's pedigree rather than their performance. Check: separate "who they know" from documented behaviour.
• Legal exposure. Backchannelling with no lawful basis or notice. Check: is legitimate interest documented in a Legitimate Interest Assessment, and did the founder get upfront notice?

The controlling insight is that one negative is data and a pattern is a verdict. Repeated references to the same concern generally carry more weight than one outlier complaint. The value comes from patterns across multiple conversations, which is why the triangulation step is the load-bearing part of the whole program.

Consent, notice, and lawful basis

Backchannelling a founder without a lawful basis or upfront notice is the one failure that can outlast the deal, and the fix is cheap: document your basis and tell the founder before you start. VC-specific legal guidance for backchannelling a founder is not publicly established, so I lean on the transferable body of employment-reference law under GDPR and UK GDPR - if you operate outside those regimes, confirm the equivalent locally.

Three practical rules follow from that guidance. First, do not rely on consent: due to the power imbalance in these settings, consent is rarely considered freely given, and employers typically rely on legitimate interest instead. Second, if you use legitimate interest, document it - the basis requires a Legitimate Interest Assessment ensuring the processing is necessary, proportionate, and does not override the founder's interests. Third, in regulated industries where references are mandatory, the appropriate basis is legal obligation rather than legitimate interest.

The documented best practice for investors is upfront notice. Form Ventures, for example, discloses that its diligence includes reaching out to mutual connections on its "how we invest" page, shared before a first meeting. That single move converts backchannelling from a surprise into an expected part of the process and removes most of the friction.

Converting the calls into a verdict

The decision is a written recommendation that names the load-bearing signals, scores patterns rather than single calls, and flags any lone hedged "yes" for a re-check before it can carry the day. Do not average the calls into a number; identify the two or three signals that actually move your conviction and state whether each is confirmed.

Before you write "go" or "no-go," run this final check.

Keep the program current by treating it as a living checklist, not a memory. Ecosystems change and the reachable backchannel pool shifts, so re-check your bucket coverage each time you enter a new market or stage. The method holds regardless of the numbers; what changes is how many names you can reach and how hard you have to work to find the pivotal failed-peer call. That call, more than any glowing reference, is what earns the check.