The evidence base
Nothing on this site is opinion dressed as research. This page is what was collected, how, what it can and cannot support, and every claim Kendall could break.
No human has read a word of this. Machine output, published raw so the reasoning can be attacked before anyone polishes it. Judge the chain, not the copy — if a numbered step is wrong, say which.
What was collected
| Layer | Volume | Method |
|---|---|---|
| Fund Recs site | 301 pages | Firecrawl |
| Published assets | 209 pages · 118/118 blog bodies · 1 whitepaper PDF | Firecrawl |
| Client video transcripts | 4 | Northern Trust, Apex, Alchelyst, SuMi Trust — transcribed from open hubfs .mp4 |
| Third-party coverage | 49 pages | Firecrawl + 59 manual fetches |
| Competitor pages | 467 | SmartStream 426, AutoRek 15, Duco 11, Xceptor 8, Gresham 7 |
| Review listings | 21 | G2, Capterra, Gartner, Glassdoor, TrustRadius |
| Gated assets | 10 targeted | 3 full text, 4 partial, 3 blocked. No forms were submitted anywhere |
| LinkedIn feeds | 7 | Six companies + Meaney. Authenticated read |
| LinkedIn ad libraries | 375 ads | Six searches, method validated against a known-positive |
| Structured store | 1,205 records · 7.7 MB | Deduped; 599 duplicate URLs merged |
Every record carries: url · domain · company · title · type · published_date · word_count · themes[] · mentions_ai · named_products[] · named_clients[] · quantified_claims[] · raw_path.
How
Two agents working in parallel with a clean split. One held the argument, one held the evidence store. Five Sonnet sub-agents ran the research legs; a Firecrawl crawler built the artefact store; an authenticated browser session read what no crawler can reach.
The one thing that made everything else possible was specifying quantified_claims as a field before the crawl ran. That single decision is what let us test — rather than assert — the claim that Fund Recs publishes fewer numbers than its competitors. It turned out to be false, and the field is why we found out.
Why it matters commercially
- It's queryable. "Which competitor claims what, with a number, and where" is one query, not a week of reading.
- Every claim traces to a URL. Nothing on this site is unfalsifiable.
- It updates. Re-run monthly and it becomes a competitive monitor rather than a one-off audit.
- Nobody sells them this. Agencies bring opinions. This is an evidence base with the opinions bolted on afterwards.
Kendall — pulling the claims apart
Every number on this site, attacked. If you're asked "how do you know", this is the honest answer.
| Claim | Status | The attack |
|---|---|---|
| depositary 142 : 1 | Weakest claim on the site | Fund Recs has 423 records in the corpus; Duco has 58, only 13 with full text. The ratio is heavily inflated by crawl asymmetry — Duco's site is JS-rendered and under-captured. The direction is safe; the number is not. Never quote 142:1. Say "we found no depositary content from any competitor." |
| $10m ARR, 49% growth, 93% close | Self-published | A LinkedIn marketing post, unaudited. "PoC-to-close" is undefined — if a PoC only starts after heavy qualification, 93% is a selection effect, not a product signal. The whole comms plan rests on it. |
| ACV ~$154k · ~65 clients | Modelled | Built from "$10m ARR ÷ inferred client count", triangulated from "37 EMIR clients" and "40+ clients" across different years. Could be 40 clients at $250k or 100 at $100k — two very different companies. |
| 16 PoCs per year | Modelled on modelled | Depends on the ACV above and an assumed 60/40 expansion split that nothing evidences. If expansion is 80% of growth the number is 7. |
| ~1,545 buying units · 17–25% coverage | Directional | Segment populations are estimates from industry structure, not a count. "50–80 accounts per seller" is a rule of thumb. Both could be out by half. |
| Duco 344 ads, Fund Recs 0 | Measured | Solid — same method returned 31 for Xceptor. But the zeros for SmartStream, Gresham and AutoRek are name-match failures, not confirmed absences. Don't say "only Duco advertises." |
| Xceptor ≈ $217k/year | Derived | Arithmetic on their marketing ROI figures assuming IDC's three-year convention. Appearing in two independent cases raises confidence; it doesn't confirm it. |
| Josh Yeo / Channel Capital | Published, stale | Channel merged with Fidante in June 2026 — Channel Group, A$150bn. The case study describes a company ~7× smaller. He may not be in that seat. |
| Engagement medians | Measured, thin | n=24 for Fund Recs, "Top" sorted not chronological, ~4 months. LinkedIn's sort is not neutral. |
| AI models answer wrongly | Second-hand | Reported from testing, not re-run at time of writing. Model outputs vary by session and phrasing. Re-test before quoting. |
| Rex | Solid | Independently corroborated — absent from a 301-page crawl. The caveat that matters: it's customer support, not reconciliation. Overclaiming it would be worse than not using it. |
| $800k budget | Benchmark | We don't know current spend. At ~10% net margin, $800k is roughly their entire net profit. |
What's missing
- LinkedIn is a manual read, not a scrape. ~12 Fund Recs company posts, ~9 of Meaney's, 2–6 per competitor, "Top"-sorted, 3–6 months deep. There is no denominator for "every post" — LinkedIn won't serve one.
- Three gated assets remain blocked — Gresham's WBR and IDC reports, Xceptor's ETD e-book. No forms were submitted.
- No internal data at all. No CRM, no pipeline, no win/loss, no revenue by module, no pricing.
- SmartStream's blog index is JS-rendered and under-captured relative to the others.
- Ad-library payer search not run for the three zero results.