Automated referral requests 88 outcomes · Aug 2026
80 declines, 8 accepts.
A read of the comments on 88 referral request outcomes from August 2026.
QUESTION 1
Why are they being rejected?
Twelve buckets behind 80 declines, and how they split between requests stopped internally and requests that reach a Group Leader.
QUESTION 2
What do the accepts share?
Five traits recur across the accepted requests, and two of them appear in the comments rather than the research.
QUESTION 3
What could we check before submission?
A rules check, three thresholds, and a source check would cover most of the stated decline reasons.
Click through the findings. →
Where Declines Happen
Most declines happen before a Group Leader sees the request.
Every square is one referral request. The status field records who declined: GD - Declined Request is the Group Leader, and any other status means a research coordinator or MDA resolved it internally.
Resolved internally — research / MDA (64)
Declined by the Group Leader (16)
Accepted / request sent (8)
0%
Resolved internally
0%
Reached a GL, declined
0%
Accepted
0
Member groups represented
The buckets in this deck are coded from the free-text comments on each request, across 25 member groups and 21 Group Directors.
Question 1 · The Buckets
Eighty declines, twelve reasons.
Each request coded to its primary stated reason. Bars split by who said no.
Resolved internally
Declined by GL
Three buckets account for 36 of 80: the shared affiliation is too large, the claim does not hold up on checking, or the prospect is not a fit for the group.
Bucket 1 · Largest
The shared affiliation is often too large to imply a relationship.
Thirteen declines note that the shared group is too large to establish a connection. Bar length is the number of people who share the affiliation cited as evidence.
Declined — the shared group
Accepted — the shared group
Mark, on the NAM board: "NAM board of directors is REALLY big. I wouldn't assume that John knows any of those people super-well." Bahar, on the CNBC CFO Council: "the list has 139 CFOs on it. Way too large to make a referral off of."
Bucket 4 · Co-appearance
Speaking at the same event is not the same as sharing a panel.
Eight declines note that both people spoke at the event but were not on the same panel. Six comments describe it in nearly the same terms.
Submitted · one conference, eight sessions
✗ No evidence of direct interaction.
Accepted · one session, four seats
✓ Same session, four participants.
Wendy, on this pattern: "We can't assume that just because a member and a prospect sat on a panel together means they have a relationship outside of the event." The accepted requests pair a same-panel find with a second connection point.
Bucket 2 · Verification
Thirteen requests were declined after source verification.
In each case a coordinator opened the cited source and could not find the stated fact. Four are labelled "Hallucination" in the comments.
✕Not on the agenda"Megha Yethadka is not listed on the agenda nor listed under the same panel."
✕Not on the board roster"Neither Soph Sophocles nor Stephen Lawniczak is currently listed on the official leadership board."
✕Not in the transcript"Gina Steed is not mentioned in the transcript of the episode."
✕Not on their LinkedIn"No evidence of overlap. T. Rowe Price not listed on Audra['s] LI."
✕Not in attendance"She's nowhere to be found on the actual agenda… Graham spoke with his predecessor."
✕Labelled hallucination ×4A single unverifiable event — a "joint banking coalition comment letter" with the member CC'd — generated four separate requests.
Each of these is a check against the cited source, and each can be run before a request is submitted.
Outside the Connection
31 declines are unrelated to the connection itself.
These requests are declined on a fact about the prospect, the member, or a standing rule — each of them knowable before the connection research begins.
0
Not a group fit
Vendor, company too small, already a member, wrong group entirely.
0
Stale role data
"No longer in role." The prospect has left the job on record.
0
Member-side
Member is disengaged, retired, or the ask is inappropriate to make.
0
Intragroup rule
"Not risk to risk." Cross-function asks that violate an unwritten routing rule.
0
Competitor / block list
Coke & Pepsi in one group; IBM on the Accenture block list; a Lowe's → Home Depot ask.
3 of the 16 GL declines are member-side: the member was disengaged, or the ask was not appropriate to make of them. These are routing questions rather than research quality.
Question 2 · The Accepts
Accepted requests quantify the connection.
Each one states how many people share the connection point, and gives a specific, dated artifact. Most carry all five traits below.
🔢
A small, stated denominator
A 4-person panel. An 11-member board. 14 speakers at the event. The size of the group is stated.
🧱
Two or more connection points
Shared board plus a joint article. Co-panel plus a WSJ quote plus a photo.
⏱️
Months, not years
June 2026 roundtables. April 2026 board overlap. Recency measured in weeks.
📸
Visual proof
A LinkedIn post with both executives tagged or pictured together.
🪞
Role parity
Tax→Tax. CAO→CAO. Supply chain→supply chain. Never cross-function.
Two process traits appear on the accept side: a named validation stamp ("Validated by: Virginia Kuester") and a Prospect History paragraph giving the GD context on the ask.
One caveat: 1 of the 8 came back Member Declined. A 5-person committee and a documented overlap, but the member replied "I don't know him well." A small group size alone does not carry the ask.
Question 3 · Proposed Checks
Three checks before a request is submitted.
Each one is a count, a date, or a lookup against the cited source.
🤖
Candidate found
Agent proposes a member → prospect connection.
📋
Check 1 · Rules
Role currency, group fit, intragroup rule, competitor blocks, member engagement.
📏
Check 2 · Thresholds
< 15 people share it
≤ 18 mo since the evidence
≥ 2 connection points
🔍
Check 3 · Sources
Open the cited source and confirm both names appear in it.
🧑💼
To the Group Leader
With the group size, the date, the connection points, and prospect history attached.
0
Covered by check 1
0
Covered by check 2
0
Covered by check 3
0/80
Stated reasons covered
Accepted requests state the size of the shared group. Most declines do not.
Checking group size, recency, and connection count before submission would cover 70 of the 80 stated decline reasons.