# Marcus Reilly, VP of Sales at Claridex (83 people, B2B SaaS, logistics data) — read of Lead Scoring AI, May 6 2026

> 14 years in B2B sales, last 6 managing teams. I have built four different lead scoring models from scratch in spreadsheets and I have broken all four of them.

## How I got here

Searched "HubSpot lead scoring limitations" on a Sunday night after my daughter's soccer game. We are getting 300-400 inbound leads a month now and my reps are picking their favorites, not the best ones. Third or fourth result was a Reddit thread where someone mentioned this. Clicked through without much expectation.

## What I clicked first

"Score leads like your best closer. Every time." That is a good line. It got me because it is framing the problem correctly. My best closer, Dana, has this instinct that I cannot train into the other three reps. If this thing can operationalize whatever Dana is doing, I care. I also clicked "Try it Live" immediately because I wanted to see whether it was a real demo or one of those typeform things that just asks your email.

## Where I paused

"Train the model on your closed deals." That stopped me. How many closed deals do I need? We close maybe 60-70 accounts a quarter. Is that enough? What if my closed deal mix is skewed toward one segment because that is what we happened to focus on last year? I did not see any answer to this on the page, and it is not a small question. A model trained on bad inputs is worse than no model because my reps will trust it.

## What I distrusted

No logos. No numbers. Not one customer quote. "Surface high-probability buyers. Deprioritize tire-kickers before your team wastes cycles." That is three different ways of saying the same thing and none of them is evidence. "Discover which company profiles, buyer signals, and engagement patterns your team closes most consistently" -- okay, but what signals exactly? Firmographic? Website visits? Email opens? I am reading this and I genuinely do not know whether the product is pulling from my CRM activity history or whether it needs a separate tracking pixel or whether it is just doing company-level enrichment from a third-party database. The page does not say. Also: "Built by Wishdeal Studio." I had to Google that. I do not know if this is a funded startup or a two-person agency that built this on the side. That matters to me because if I put this in our stack and it breaks in 8 months, I own that decision.

## What would convince me

One customer, named, with a company I can look up, saying something like "We had 400 inbound leads a month and after 90 days the top quartile was closing at 3x the rate of leads the model ranked below 40." I do not need a polished case study PDF. A tweet, a Loom, a Reddit comment from a real person with a real profile. That is all. I also want to see one screenshot of what the actual score card looks like inside HubSpot. Not a mockup. A real one with a real contact record and the reasoning section they describe. "Size match? Engagement velocity? Industry fit?" Those words are in there and they sound reasonable. Show me what that looks like in practice.

## What I'd ask in an email reply

1. How many historical closed deals do we need in HubSpot before the model is actually useful, and what does the ramp period look like before scores are trustworthy?
2. Is the scoring based only on data inside my CRM, or are you pulling from external sources, and if so, which ones and how do you handle GDPR for EU leads?
3. What happens when I close a new deal type we have never sold before -- does the model degrade gracefully or does it just misfire until we retrain it?

## Verdict: on-the-fence

The problem framing is exactly right and I have not seen many of these that nail it. But the page is so thin on specifics that I cannot tell whether this is a real product or a well-written landing page for something that is still being built. I would reply to the founder if there was any signal at all that a real company was behind it.

---
*Memo by skeptic persona, generated 2026-05-06. Studio breaks own self-grading loop.*