# Ray Kuhlman, VP of GTM at Fieldpath (47 employees, B2B SaaS) — read of Segment AI, May 19 2026

> 11 years in B2B sales and GTM, currently running a 6-person outbound team on HubSpot + Apollo + Clay + Gong. Still get burned occasionally by tools that over-promise on segmentation.

## How I got here

Someone in the GTM Alliance Slack dropped this link under a thread about ICP drift. The comment was just "interesting model here" with no context. I opened it on my phone during my daughter's Saturday soccer practice, skimmed it, then came back to it on my laptop Monday morning because something about it stuck in a weird way. Not sure if it was curiosity or confusion.

## What I clicked first

The hero pulled me in for about 8 seconds. "Your ideal customer is already in your data" is a line I've heard before, but then right underneath: "You'll get a ranked list of next-best prospects based on the signals that predict your closes." That's a more specific promise than I usually see in the hero. Most of these pages say "find your ICP" and leave it there. The word "signals" combined with "predict your closes" at least implies the tool has done something with historical outcome data, not just firmographics.

Then I hit the $5 unlock button and got disoriented.

## Where I paused

The "Honest disclosure" section stopped me cold. It reads: "we don't have live customers on this idea yet. We shipped the strategy package; you ship the customer conversations."

That's when I understood what I was actually looking at. This isn't a product. It's a blueprint for a product someone could build. The $499/mo pricing earlier in the page is the projected price for the thing that doesn't exist yet. The $5 gets you a dossier. The $99 gets you code and copy assets.

That's a legitimately different business model than what the top two-thirds of the page implied. I'm not saying it's dishonest. The disclosure is right there. But I read about four pricing sections before I understood what I was being invited to buy.

## What I distrusted

The stats have no grounding. "Our segmentation model outperforms manual segmentation by 3x on deal velocity." Outperforms whose manual segmentation? In what study? On what customer dataset? This is sitting next to an honest disclosure that there are no live customers. So where did the 3x come from?

Same with "Average customer sees 40% faster sales cycles and 25% larger deal sizes within 90 days." There is no average customer. These numbers are either from the Fermi modeling on the page itself (which it calls estimates) or they're just pattern-matched from other segmentation tools' marketing. Either way, putting them in a "Why Segment AI" feature section without a footnote is the move that costs this page credibility.

Also: "Informed by patterns from publicly-documented segmentation research and 50,000+ B2B companies." What research? Which 50,000 companies? Documented where? This reads like it was written to pass a smell-test rather than to mean anything.

## What would convince me

I'd want to see the actual scoring rubric explained, not just "our segmentation model outperforms." Show me a before/after on a real (or synthetic but documented) dataset. Even a toy example: here's a company with 80 customers, here's what manual analysis found, here's what the model found, here's why the model's version closed faster. Walk me through the logic once.

The Wishdeal Factory scoring at the bottom is actually interesting to me. The "1 in 8 meaningful success odds" is more honest than most pitch decks. But it also makes the "40% faster sales cycles" claim above feel contradictory. You can't have both. Pick a tone.

## What I'd ask in an email reply

1. The page shows $499/mo as a price. Is that what you're planning to charge, or what you think the market will bear? Who set that number and why?

2. The segmentation model is described as using "firmographic, behavioral, and transactional signals." What happens when I connect my HubSpot and half my deals have incomplete company-size data? Does the model degrade gracefully or just return garbage clusters?

3. You're selling the dossier for $5 and the code starter for $99. What does "working code starter" actually mean? Is this a Postgres schema and some Python scripts, or something I can put in front of a customer in its current state?

## Verdict: on-the-fence

The honest disclosure earns some respect, and the hero copy is more specific than average. But the page talks like a live product for most of its length and then reveals it isn't one, and the stats that appear before that disclosure are unsupported and probably made up from averages. I'd read the $5 dossier. I would not forward this to my team yet.

---
*Memo by skeptic persona, generated 2026-05-19. Studio breaks own self-grading loop.*