Backed by Y CombinatorThree documents and two sheets that synthesize reference calls by theme instead of call by call, and track on-list responses separately from off-list ones.
Free download · No account needed
What you get
Pack of 5 files: 3 documents, 2 spreadsheets
Investor Reference Call Questions TemplateSynthesis Note
Every response is logged against the theme it addressed, on-list and off-list marked separately, then read down each theme's column across every call that touched it.
| Theme | On-list | Off-list | Gap |
|---|---|---|---|
| — | — | — | — |
| — | — | — | — |
| — | — | — | — |
The theme with the largest gap between on-list and off-list mentions is usually the finding worth calling about directly.
Eight reference calls on one CEO candidate ahead of a Series B: five names the CEO supplied, three found independently. Logged call by call, each reads as one more opinion next to four others. Logged by theme instead, with on-list mentions kept separate from off-list ones, a number appears. On-list references returned zero negative mentions across twenty-two theme mentions. Off-list references returned five negative mentions out of thirteen, 38.5 per cent, and four of those five land on the same two related themes.
The two negative accounts came from the CEO's prior company: a former direct report and a former peer, contacted separately, who did not know about each other's calls. A third off-list reference, a current employee the CEO did not think to list, raised nothing negative, so the gap tracks one specific period rather than off-list sourcing generally. A reference call feeding an investment memo is only as good as how it gets read; summarized call by call, the same eight calls would have produced eight positive-leaning paragraphs and no visible pattern.
Candidates nominate the references likely to speak well of them, and those references predominantly do; that is the finding of structured research on reference-check content, not a cynical guess. Off-list, or backchannel, calls exist for exactly this reason, and Bain Capital Ventures' own field guide to running them notes the value shows up in patterns across several independent accounts rather than any single call. The same discipline that finds a gap in a first meeting's question set applies here, one call later in the process.
Synthesis NoteDuskfield Robotics, illustrative. Eight completed calls on the CEO ahead of a Series B: five on-list, three off-list.
| List type | Calls | Mentions | Negative | Rate |
|---|---|---|---|---|
| On-list | 5 | 22 | 0 | 0.0% |
| Off-list | 3 | 13 | 5 | 38.5% |
| Theme | On-list | Off-list | Gap |
|---|---|---|---|
| Decisiveness under pressure | 5/5 pos | 3/3 pos | None |
| Handles disagreement without becoming defensive | 5/5 pos | 2/3 neg | Diverges |
| Follow-through on commitments | 4 pos, 1 neu | 3/3 pos | None |
| Proactively surfaces bad news | 3/3 pos | 2/2 neg | Diverges |
| Direct-report morale and retention | 3 pos, 1 neu | 1 neg, 1 neu | Weak signal |
All five negative mentions trace to two references, not three. Both work at the CEO's prior company. The third off-list reference, an unlisted current employee, read exactly like the on-list team.
Response RegisterOne row per theme a call actually addressed, not one row per call. 8 of 35 rows shown.
| Reference | List | Theme | Rating | Example |
|---|---|---|---|---|
| L1 | On-list | Handles disagreement | Positive | Took pushback in a team meeting, no defensiveness |
| L3 | On-list | Proactively surfaces bad news | Positive | Flagged churn risk before being asked |
| O1 | Off-list | Handles disagreement | Negative | Framed a technical objection as disloyalty, in front of the exec team |
| O2 | Off-list | Handles disagreement | Negative | Same meeting, different seat: objection was shut down |
| O3 | Off-list | Handles disagreement | Positive | Took pushback on a budget request without issue |
| O1 | Off-list | Proactively surfaces bad news | Negative | A delay surfaced only when asked directly, three weeks late |
| O2 | Off-list | Proactively surfaces bad news | Negative | Confirms: bad news surfaced only under direct questioning |
| TOTAL | 8 refs | 5 themes | 22 on / 13 off | 0 neg on, 5 neg off |
O3, the unlisted current employee, is off-list but reads like the on-list team on this theme. The gap tracks the prior company, not off-list sourcing itself.
Reference RegisterEvery call with its relationship, list type and status.
| Ref | Relationship | List type | Status |
|---|---|---|---|
| L1–L2 | Direct reports (current) | On-list | Completed |
| L3 | Board observer | On-list | Completed |
| L4–L5 | Customers | On-list | Completed |
| O1 | Former VP Eng, prior co. | Off-list | Completed |
| O2 | Former peer, prior co. | Off-list | Completed |
| O3 | Direct report, unlisted | Off-list | Completed |
| TOTAL | 8 references | 5 on / 3 off | 8/8 done |
A register with zero off-list rows cannot produce the comparison this pack is built around; it can only produce a longer version of the candidate's own pitch.
Call GuideMatched to what each relationship can actually see, not one list asked of everyone.
How much did they let you run versus check in? Describe a specific disagreement and what happened when you raised it.
Describe a time a plan slipped. When did you find out, and from whom? How many comparable CEOs have you worked with, and where does this one rank?
Was what got committed during the sales process delivered on the stated timeline? If something slipped, how did you find out?
Every question asks for a dated, specific instance. A rating with no example attached does not make it into the Synthesis Note.
Every call tracked with a list-type column, on-list or off-list, so the comparison the whole pack is built around exists from the first row rather than as an afterthought
A different question set for a direct report, a peer, a board member and a customer, since each vantage point can honestly answer a different set of questions
How to find and reach references nobody supplied, transparently and without routing the request back through the candidate first
One row per theme a call actually addressed, not one row per call, which is what lets a pattern spanning three unconnected calls become visible at all
Read down each theme's column across the whole set, on-list mentions compared against off-list ones, with the gap between them stated as its own fact
The standing rule: no per-call summary stands as the space's primary output, and on-list and off-list mentions live in one sheet, never two
Whatever reference list already exists, plus what you know about the candidate's background independently, so off-list candidates can be identified against it.
Every call logged as on-list or off-list before it happens. A prior role with nobody on the list is the clearest place to look for an independent contact.
Each response filed under the theme it addressed, not the call it came from, with the specific example attached to the rating rather than the word alone.
Read down each theme's column across every call that touched it. Lead with whichever theme shows the largest gap between on-list and off-list, not the first one alphabetically.
Because the question is not what to ask, it's who answers. Candidates can only offer people who already lean toward speaking well of them; that isn't dishonesty, it's how anyone builds a list. Splitting the two and comparing them by theme is what turns a curated set of five positive calls into a finding, rather than into eight paragraphs that all sound reassuring.
At least one for every two on-list references is the working minimum in this pack; fewer than that and a genuine pattern can look like one person's grudge. In the worked example, three off-list calls against five on-list ones was enough for two independent, unconnected accounts to describe the same incident from different seats.
No, it changes how the same calls get read. Most processes already make calls and write a paragraph per call, the same way a data room gets worked through document by document. This pack asks the calls to also get logged by theme, on-list separated from off-list, so a pattern spanning three calls becomes visible instead of sitting unconnected across three summaries.
Then that is the finding, and the Synthesis Note says so plainly rather than manufacturing a gap that is not there. A flat, unanimous result across both groups is real evidence the candidate's story holds up independently, which is different from five positive on-list calls that were never tested against anyone the candidate did not choose.
Yes, done transparently. The pack's own Off-list Reference Approach doc covers exactly this: name the process plainly when reaching out, give an easy way to decline, and never tell one reference what another one said. Reaching a former colleague independently is a routine, disclosed part of diligence, not a covert one.
It installs the pack as your own private space, then helps build the Reference Register from whatever list and work history you send. As call notes come in, it helps sort each response into the right theme rather than filing it under the call, and drafts the Synthesis Note once enough calls are in to read a real pattern.
The Call Guide, Off-list Reference Approach and Synthesis Note download as Word documents; the Reference Register and Response Register download as CSV. Word opens in Word, Pages or Google Docs, and the CSVs open in Excel, Numbers or Sheets with the columns intact.
Written by
Co-founder & CEO, River
Chandler has spent the last six years working in AI, building tools for millions of users and Fortune 100 companies. In 2025, he co-founded River to build a new kind of operating system for AI agents.
Send your reference list and what you know of the candidate's background. The first pass builds the register and flags which prior roles have no independent reference yet.
Build my reference register