# Matt Piotrowski, Infrastructure Lead at Crawlbase Partners — read of ProxyBox ISP Quality Scorer, June 18, 2026

> 9 years routing proxy traffic for e-commerce intelligence clients. Currently managing 4 ISP feeds across Bright Data, Oxylabs, and two smaller residential suppliers. I have a spreadsheet I update manually every Monday morning to track which ISPs are choking on which target domains. It takes about an hour. I hate it.

## How I got here

Searched "compare ISP success rates proxy fleet dashboard" on Google last week. This showed up on page 2. I bookmarked it and forgot about it. Came back this morning while waiting for a scrape job to finish. My daughter's soccer practice got rained out so I had an extra 20 minutes.

## What I clicked first

"Cut timeouts by 18%." That jumped out immediately because it's a number. Specific. Not "reduce timeouts significantly" or "improve performance." 18%. I wanted to know: 18% compared to what baseline, on which task types, measured over what window. I clicked around looking for that answer and didn't find it.

## Where I paused

The honest disclosure section. Long pause. "We don't have live customers on this idea yet. We shipped the strategy package; you ship the customer conversations." That stopped me cold because the feature copy above it reads like a real product. "Real-Time ISP Scoring," "SLA Compliance Dashboard," "Auto-failover recommendations included." I had to re-read the whole page to understand what I was actually looking at. This is a business idea being sold as a dossier, not a product I can sign up for. The specs table reads like a product sheet but there's no product.

## What I distrusted

The self-scoring feels circular. "71/100 Adoptability" using their own "10 Adoptability axes" is the studio grading its own homework. The "buyer clarity: 10/10" score is notable given that I, the supposed buyer, spent 4 minutes figuring out whether this thing exists yet.

"$-19,760 Year-1 take-home" is interesting because at least they're saying the number out loud. But I don't know if that assumes I'm building this myself or hiring them. The pricing tiers go from $5 to $99 to "custom" and none of those prices map to what I thought I was buying when I landed here.

Also: "Multi-Tenant Ready White-label dashboard. Each reseller sees only their fleet data with enterprise-grade isolation." Enterprise-grade isolation for a product with zero live customers is a spec on a napkin.

## What would convince me

I want to see the actual data behind "18%." A real graph. One real deployment, even anonymized: "a proxy reseller running 3 ISPs, here's their completion rate by ISP before and after we routed by task type." Not a case study PDF. A screenshot of an actual dashboard with real numbers, even blurred company name.

And I need to understand what I'm actually buying. If the $99 tier ships me working code, what does "working" mean? Does it connect to Bright Data's API out of the box? Does it require a database I have to provision? "Code starter" is doing a lot of work in that pricing line.

## What I'd ask in an email reply

1. The 18% timeout reduction -- where does that number come from? Is it from an actual deployment or a modeled estimate?
2. The code starter at $99 -- does it actually integrate with any of the major proxy providers (Bright Data, Oxylabs, Smartproxy) or is it a blank framework I'd still need to wire up myself?
3. The idea page says you can "operate with us, custom" -- what does that actually mean in practice? Are you building this with me or for me, and what's the price range we're talking?

## Verdict: on-the-fence

The underlying pain is real, I feel it every Monday morning. But I can't tell from this page whether I'm buying a strategy doc, a code repo, or a service, and "we don't have live customers yet" after a feature list that sounds like a product in production is a trust problem. If they emailed me a 5-minute Loom showing the dashboard actually running against a real ISP feed, I'd probably reply.

---
*Memo by skeptic persona, generated 2026-06-18. Studio breaks own self-grading loop.*