How we test
Methodology
We don't review spec sheets. Every verdict on this site comes from running the actual work. The same controlled rig, the same leads, every single time, so two scores mean the same thing.
- Bench calibrated
- Re-test cadence
- Every 6 months
The bench
There is one test business. A dedicated phone number, a dedicated email, and a single one-page site. Every tool we review is installed against that same, identical setup. Same forms, same plumbing, same fake company. So the scores are comparing tools, not whoever had the nicer test rig.
We do live with each tool too. The real-world Reply First experience, actually waiting on the thing like a customer would, is the gut-check. But the published score comes from the rig, not the vibe.
The lead battery
Ten fixed inbound leads, identical for every tool. We fire the same battery across business hours, after hours, and a weekend, because “fast” at 2pm Tuesday and “fast” at 11pm Saturday are not the same product.
- 4
Easy / in-scope
Plain questions the tool should never miss.
- 3
Messy
Typos, vague asks, several questions crammed into one.
- 2
Off-script curveballs
Outside the script, where the cracks show.
- 1
Hot, ready-to-buy
A live sale on the line. Don't fumble it.
Five scored axes
Each axis is scored 0 to 10. These are the exact labels you'll see on every review's scorecard. Nothing is hidden behind a different name.
- Speed to first reply0 to 10
Median time from lead submission to the first real response. We measure the clock, not the marketing claim.
- Quality of replies0 to 10
Correct, in scope, and not hallucinating. Off-script leads are weighted hard. A confident wrong answer is the worst failure a tool can hand a customer, and we score it that way.
- Setup & onboarding0 to 10
Signup to first working reply. How much code, configuration, and praying it takes to get a tool actually answering leads.
- Hand-off to humans0 to 10
Whether the hot lead escalates to a person, and how fast a human is actually looped in. A bot that traps a buyer is worse than no bot.
- Price vs. value0 to 10
The real total at realistic volume, with overages, seats, and conversation caps included. Not the headline tier on the pricing page.
The Dread Score
The five axes roll up into one number out of 100. No secret weighting, no thumb on the scale. Just the maths:
Dread Score = ( Speed + Quality + Setup + Hand-off + Value ) × 2
Five axes, 0 to 10 each → 0 to 100. No secret weighting.
That score is the Dread Score, our own hands-on, benched verdict. A second score is coming: the Skeptic Score, built from real public reviews and crowd reports. Every verdict on the site today is a Dread Score; the Skeptic Score lands once we've done the public-review reads to back it.
Verdict bands
The score lands in one of three bands. That band is the verdict, the badge at the top of every review.
- 70 to 100Certified
- 40 to 69Mixed
- 0 to 39Slop
Hard disqualifiers
Some failures are bad enough that the raw maths doesn't get the final word. These caps override the axis scores, no exceptions:
Confidently fabricated answers
Quality is capped at 3. The tool cannot exceed a MIXED verdict, no matter how fast or cheap it is.
No human hand-off for a hot lead
Hand-off is capped at 2. A ready-to-buy lead that dead-ends in a bot is a lost sale, full stop.
Fake testimonials or case studies
Automatic SLOP. If a vendor invents proof, nothing else on their site can be trusted either.
Dark-pattern cancellation
Price is capped low. Making it hard to leave is a tax we price in, even when the product works.
Worked example: Chatbase
Chatbase is a genuinely good documentation-search bot, and its raw axes added up to 60, a MIXED score. But it has no structural way to escalate a hot lead to a human, which trips the No Hand-off disqualifier. Cap applied: 60 30, MIXED down to SLOP. A tool that traps a ready-to-buy lead can't be Certified, no matter how good the rest of it is.
When we re-test
Every tool gets back on the bench every six months, or sooner if it ships a major version. The tested-on date stays visible on every review so you always know whether a verdict is fresh or whether the product has moved on since we last looked.
The other half of the promise
This is how we score. If you're wondering whether money ever moves a verdict, it doesn't. Here's exactly how the affiliate side works.